2023-04-27 12:19:53,966 INFO [train.py:976] (2/8) Training started 2023-04-27 12:19:53,966 INFO [train.py:986] (2/8) Device: cuda:2 2023-04-27 12:19:53,967 INFO [train.py:995] (2/8) {'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 12:19:53,968 INFO [train.py:997] (2/8) About to create model 2023-04-27 12:19:54,652 INFO [zipformer.py:178] (2/8) 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] (2/8) Number of model parameters: 70369391 2023-04-27 12:19:57,230 INFO [train.py:1016] (2/8) Using DDP 2023-04-27 12:19:58,268 INFO [multidataset.py:46] (2/8) About to get multidataset train cuts 2023-04-27 12:19:58,269 INFO [multidataset.py:49] (2/8) Loading LibriSpeech in lazy mode 2023-04-27 12:19:58,289 INFO [multidataset.py:65] (2/8) Loading GigaSpeech 1998 splits in lazy mode 2023-04-27 12:20:00,752 INFO [multidataset.py:72] (2/8) Loading CommonVoice in lazy mode 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:230] (2/8) Enable MUSAN 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:231] (2/8) About to get Musan cuts 2023-04-27 12:20:03,039 INFO [asr_datamodule.py:255] (2/8) Enable SpecAugment 2023-04-27 12:20:03,039 INFO [asr_datamodule.py:256] (2/8) Time warp factor: 80 2023-04-27 12:20:03,039 INFO [asr_datamodule.py:266] (2/8) Num frame mask: 10 2023-04-27 12:20:03,039 INFO [asr_datamodule.py:279] (2/8) About to create train dataset 2023-04-27 12:20:03,039 INFO [asr_datamodule.py:306] (2/8) Using DynamicBucketingSampler. 2023-04-27 12:20:07,553 INFO [asr_datamodule.py:321] (2/8) About to create train dataloader 2023-04-27 12:20:07,554 INFO [asr_datamodule.py:435] (2/8) About to get dev-clean cuts 2023-04-27 12:20:07,555 INFO [asr_datamodule.py:442] (2/8) About to get dev-other cuts 2023-04-27 12:20:07,556 INFO [asr_datamodule.py:352] (2/8) About to create dev dataset 2023-04-27 12:20:07,802 INFO [asr_datamodule.py:369] (2/8) About to create dev dataloader 2023-04-27 12:20:25,620 INFO [train.py:904] (2/8) Epoch 1, batch 0, loss[loss=7.45, simple_loss=6.745, pruned_loss=7.032, over 17212.00 frames. ], tot_loss[loss=7.45, simple_loss=6.745, pruned_loss=7.032, over 17212.00 frames. ], batch size: 45, lr: 2.50e-02, grad_scale: 2.0 2023-04-27 12:20:25,620 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 12:20:32,875 INFO [train.py:938] (2/8) Epoch 1, validation: loss=6.911, simple_loss=6.238, pruned_loss=6.721, over 944034.00 frames. 2023-04-27 12:20:32,876 INFO [train.py:939] (2/8) Maximum memory allocated so far is 12891MB 2023-04-27 12:20:36,366 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:20:52,194 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:21:17,135 INFO [train.py:904] (2/8) Epoch 1, batch 50, loss[loss=1.241, simple_loss=1.104, pruned_loss=1.23, over 16426.00 frames. ], tot_loss[loss=2.177, simple_loss=1.97, pruned_loss=1.98, over 741926.40 frames. ], batch size: 146, lr: 2.75e-02, grad_scale: 2.0 2023-04-27 12:21:46,533 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:22:02,795 WARNING [train.py:894] (2/8) Grad scale is small: 0.001953125 2023-04-27 12:22:02,795 INFO [train.py:904] (2/8) Epoch 1, batch 100, loss[loss=1.081, simple_loss=0.9256, pruned_loss=1.231, over 16729.00 frames. ], tot_loss[loss=1.634, simple_loss=1.453, pruned_loss=1.622, over 1320472.06 frames. ], batch size: 124, lr: 3.00e-02, grad_scale: 0.00390625 2023-04-27 12:22:13,668 INFO [optim.py:368] (2/8) 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 12:22:20,962 WARNING [optim.py:388] (2/8) Scaling gradients by 0.0112030990421772, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:21,072 INFO [optim.py:450] (2/8) 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 12:22:43,766 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:22:46,126 WARNING [optim.py:388] (2/8) Scaling gradients by 0.0022801109589636326, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:46,234 INFO [optim.py:450] (2/8) 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 12:22:49,648 WARNING [optim.py:388] (2/8) Scaling gradients by 0.04246773198246956, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:49,754 INFO [optim.py:450] (2/8) 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 12:22:51,349 WARNING [optim.py:388] (2/8) Scaling gradients by 0.000716241542249918, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:51,459 INFO [optim.py:450] (2/8) 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 12:22:52,995 INFO [train.py:904] (2/8) Epoch 1, batch 150, loss[loss=1.141, simple_loss=0.9735, pruned_loss=1.217, over 11990.00 frames. ], tot_loss[loss=1.394, simple_loss=1.222, pruned_loss=1.447, over 1755767.98 frames. ], batch size: 246, lr: 3.25e-02, grad_scale: 0.00390625 2023-04-27 12:22:53,728 WARNING [optim.py:388] (2/8) Scaling gradients by 0.049951765686273575, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:53,833 INFO [optim.py:450] (2/8) 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 12:22:58,542 WARNING [optim.py:388] (2/8) Scaling gradients by 0.00609818659722805, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:58,648 INFO [optim.py:450] (2/8) 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 12:23:08,234 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=5.47 vs. limit=2.0 2023-04-27 12:23:12,122 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=5.92 vs. limit=2.0 2023-04-27 12:23:16,870 WARNING [optim.py:388] (2/8) Scaling gradients by 0.059935860335826874, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:16,977 INFO [optim.py:450] (2/8) 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 12:23:20,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7443, 3.7922, 3.7995, 3.7884, 3.7429, 3.7129, 3.7296, 3.8094], device='cuda:2'), covar=tensor([0.0264, 0.0130, 0.0148, 0.0088, 0.0164, 0.0268, 0.0377, 0.0205], device='cuda:2'), in_proj_covar=tensor([0.0010, 0.0010, 0.0010, 0.0010, 0.0010, 0.0010, 0.0010, 0.0010], device='cuda:2'), out_proj_covar=tensor([1.0342e-05, 1.0199e-05, 1.0330e-05, 1.0003e-05, 1.0090e-05, 1.0082e-05, 9.9195e-06, 1.0489e-05], device='cuda:2') 2023-04-27 12:23:25,521 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=6.62 vs. limit=2.0 2023-04-27 12:23:28,341 WARNING [optim.py:388] (2/8) Scaling gradients by 0.060559310019016266, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:28,450 INFO [optim.py:450] (2/8) 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 12:23:33,578 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=8.47 vs. limit=2.0 2023-04-27 12:23:42,817 WARNING [train.py:894] (2/8) Grad scale is small: 0.00390625 2023-04-27 12:23:42,817 INFO [train.py:904] (2/8) Epoch 1, batch 200, loss[loss=0.9291, simple_loss=0.7871, pruned_loss=0.9529, over 16641.00 frames. ], tot_loss[loss=1.264, simple_loss=1.096, pruned_loss=1.325, over 2086000.86 frames. ], batch size: 134, lr: 3.50e-02, grad_scale: 0.0078125 2023-04-27 12:23:44,218 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=8.40 vs. limit=2.0 2023-04-27 12:23:48,241 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=10.13 vs. limit=2.0 2023-04-27 12:23:51,003 INFO [optim.py:368] (2/8) 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 12:23:51,003 WARNING [optim.py:388] (2/8) Scaling gradients by 0.002041660714894533, model_norm_threshold=541.4743041992188 2023-04-27 12:23:51,112 INFO [optim.py:450] (2/8) 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 12:24:00,574 WARNING [optim.py:388] (2/8) Scaling gradients by 0.02974529005587101, model_norm_threshold=541.4743041992188 2023-04-27 12:24:00,688 INFO [optim.py:450] (2/8) 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] (2/8) Scaling gradients by 0.01955481991171837, model_norm_threshold=541.4743041992188 2023-04-27 12:24:01,584 INFO [optim.py:450] (2/8) 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 12:24:30,805 INFO [train.py:904] (2/8) Epoch 1, batch 250, loss[loss=0.921, simple_loss=0.7822, pruned_loss=0.8824, over 15403.00 frames. ], tot_loss[loss=1.165, simple_loss=1.003, pruned_loss=1.213, over 2356921.59 frames. ], batch size: 190, lr: 3.75e-02, grad_scale: 0.0078125 2023-04-27 12:24:33,598 WARNING [optim.py:388] (2/8) Scaling gradients by 0.057925041764974594, model_norm_threshold=541.4743041992188 2023-04-27 12:24:33,707 INFO [optim.py:450] (2/8) 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 12:24:39,563 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=4.46 vs. limit=2.0 2023-04-27 12:25:00,568 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=18.96 vs. limit=2.0 2023-04-27 12:25:16,071 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:25:20,568 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:25:21,111 WARNING [train.py:894] (2/8) Grad scale is small: 0.0078125 2023-04-27 12:25:21,112 INFO [train.py:904] (2/8) Epoch 1, batch 300, loss[loss=0.897, simple_loss=0.7533, pruned_loss=0.851, over 16890.00 frames. ], tot_loss[loss=1.095, simple_loss=0.9361, pruned_loss=1.124, over 2579588.92 frames. ], batch size: 109, lr: 4.00e-02, grad_scale: 0.015625 2023-04-27 12:25:28,245 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=4.48 vs. limit=2.0 2023-04-27 12:25:29,642 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6970, 5.9848, 6.0493, 6.0612, 6.0419, 6.0569, 5.9272, 5.7772], device='cuda:2'), covar=tensor([0.0668, 0.0051, 0.0021, 0.0028, 0.0032, 0.0034, 0.0059, 0.0225], device='cuda:2'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:2'), out_proj_covar=tensor([9.0155e-06, 8.9208e-06, 8.8574e-06, 8.8470e-06, 8.8131e-06, 9.0190e-06, 8.8093e-06, 9.0936e-06], device='cuda:2') 2023-04-27 12:25:30,079 INFO [optim.py:368] (2/8) 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 12:25:45,861 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=4.77 vs. limit=2.0 2023-04-27 12:26:12,607 INFO [train.py:904] (2/8) Epoch 1, batch 350, loss[loss=0.9014, simple_loss=0.7451, pruned_loss=0.8597, over 16524.00 frames. ], tot_loss[loss=1.043, simple_loss=0.8857, pruned_loss=1.055, over 2749208.52 frames. ], batch size: 68, lr: 4.25e-02, grad_scale: 0.015625 2023-04-27 12:26:18,689 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:26:18,890 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=30.37 vs. limit=5.0 2023-04-27 12:26:25,784 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=4.63 vs. limit=2.0 2023-04-27 12:26:52,676 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:27:06,004 INFO [train.py:904] (2/8) Epoch 1, batch 400, loss[loss=0.9551, simple_loss=0.7811, pruned_loss=0.8994, over 17028.00 frames. ], tot_loss[loss=1.009, simple_loss=0.8506, pruned_loss=1.005, over 2879657.10 frames. ], batch size: 50, lr: 4.50e-02, grad_scale: 0.03125 2023-04-27 12:27:17,608 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=2.70 vs. limit=2.0 2023-04-27 12:27:17,878 INFO [optim.py:368] (2/8) 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 12:27:41,163 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=31.21 vs. limit=5.0 2023-04-27 12:27:46,061 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:27:56,210 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:27:59,548 INFO [train.py:904] (2/8) Epoch 1, batch 450, loss[loss=0.8457, simple_loss=0.6898, pruned_loss=0.7713, over 16451.00 frames. ], tot_loss[loss=0.9854, simple_loss=0.8251, pruned_loss=0.9651, over 2975165.58 frames. ], batch size: 75, lr: 4.75e-02, grad_scale: 0.03125 2023-04-27 12:28:51,183 INFO [train.py:904] (2/8) Epoch 1, batch 500, loss[loss=0.9157, simple_loss=0.7393, pruned_loss=0.8267, over 17122.00 frames. ], tot_loss[loss=0.9651, simple_loss=0.8025, pruned_loss=0.929, over 3053861.14 frames. ], batch size: 49, lr: 4.99e-02, grad_scale: 0.0625 2023-04-27 12:28:59,802 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=31.12 vs. limit=5.0 2023-04-27 12:29:01,368 INFO [optim.py:368] (2/8) 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 12:29:14,976 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=15.10 vs. limit=5.0 2023-04-27 12:29:44,925 INFO [train.py:904] (2/8) Epoch 1, batch 550, loss[loss=0.95, simple_loss=0.7604, pruned_loss=0.846, over 17220.00 frames. ], tot_loss[loss=0.9517, simple_loss=0.7862, pruned_loss=0.9, over 3104173.89 frames. ], batch size: 45, lr: 4.98e-02, grad_scale: 0.0625 2023-04-27 12:29:58,105 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:03,823 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:30:12,146 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-04-27 12:30:26,926 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:38,237 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:30:38,744 INFO [train.py:904] (2/8) Epoch 1, batch 600, loss[loss=0.8913, simple_loss=0.7153, pruned_loss=0.7652, over 17191.00 frames. ], tot_loss[loss=0.9428, simple_loss=0.7742, pruned_loss=0.8745, over 3155982.71 frames. ], batch size: 44, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:30:48,386 INFO [optim.py:368] (2/8) 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 12:31:02,546 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:31:07,526 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:31:17,792 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7253, 5.7268, 5.7233, 5.7199, 5.7128, 5.7226, 5.7118, 5.7177], device='cuda:2'), covar=tensor([0.0085, 0.0081, 0.0074, 0.0099, 0.0095, 0.0100, 0.0186, 0.0129], device='cuda:2'), in_proj_covar=tensor([0.0012, 0.0013, 0.0012, 0.0013, 0.0012, 0.0012, 0.0013, 0.0012], device='cuda:2'), out_proj_covar=tensor([1.2249e-05, 1.2341e-05, 1.1901e-05, 1.2008e-05, 1.2011e-05, 1.2028e-05, 1.2120e-05, 1.1928e-05], device='cuda:2') 2023-04-27 12:31:25,745 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.20 vs. limit=5.0 2023-04-27 12:31:28,003 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:31:30,996 INFO [train.py:904] (2/8) Epoch 1, batch 650, loss[loss=0.8238, simple_loss=0.6676, pruned_loss=0.6741, over 12163.00 frames. ], tot_loss[loss=0.9312, simple_loss=0.7613, pruned_loss=0.8462, over 3187736.40 frames. ], batch size: 248, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:31:31,342 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:31:32,142 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:31:42,418 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=6.15 vs. limit=5.0 2023-04-27 12:32:08,124 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=2.69 vs. limit=2.0 2023-04-27 12:32:22,400 INFO [train.py:904] (2/8) Epoch 1, batch 700, loss[loss=0.8802, simple_loss=0.7168, pruned_loss=0.6955, over 17218.00 frames. ], tot_loss[loss=0.918, simple_loss=0.7495, pruned_loss=0.8132, over 3214482.11 frames. ], batch size: 45, lr: 4.98e-02, grad_scale: 0.25 2023-04-27 12:32:31,774 INFO [optim.py:368] (2/8) 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 12:33:00,783 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:33:05,889 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:33:14,007 INFO [train.py:904] (2/8) Epoch 1, batch 750, loss[loss=0.8318, simple_loss=0.6925, pruned_loss=0.6149, over 16742.00 frames. ], tot_loss[loss=0.9024, simple_loss=0.7379, pruned_loss=0.7762, over 3238686.13 frames. ], batch size: 62, lr: 4.97e-02, grad_scale: 0.25 2023-04-27 12:33:51,298 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:34:03,296 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=2.20 vs. limit=2.0 2023-04-27 12:34:06,256 INFO [train.py:904] (2/8) Epoch 1, batch 800, loss[loss=0.7073, simple_loss=0.5963, pruned_loss=0.5005, over 15903.00 frames. ], tot_loss[loss=0.8758, simple_loss=0.7197, pruned_loss=0.7288, over 3263661.71 frames. ], batch size: 35, lr: 4.97e-02, grad_scale: 0.5 2023-04-27 12:34:17,530 INFO [optim.py:368] (2/8) 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 12:34:18,070 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-27 12:34:54,097 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:34:58,583 INFO [train.py:904] (2/8) Epoch 1, batch 850, loss[loss=0.6814, simple_loss=0.5757, pruned_loss=0.4725, over 16744.00 frames. ], tot_loss[loss=0.8404, simple_loss=0.6951, pruned_loss=0.6767, over 3270806.09 frames. ], batch size: 124, lr: 4.96e-02, grad_scale: 0.5 2023-04-27 12:35:04,821 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 2023-04-27 12:35:20,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5733, 3.8251, 3.6166, 3.8471, 2.9128, 3.2353, 3.1224, 3.9989], device='cuda:2'), covar=tensor([0.5741, 0.4701, 0.5197, 0.3668, 0.6371, 0.6093, 0.6194, 0.3460], device='cuda:2'), in_proj_covar=tensor([0.0058, 0.0061, 0.0052, 0.0051, 0.0061, 0.0068, 0.0062, 0.0049], device='cuda:2'), out_proj_covar=tensor([5.0397e-05, 5.4381e-05, 5.1805e-05, 4.3365e-05, 5.7672e-05, 5.5537e-05, 5.3184e-05, 5.5476e-05], device='cuda:2') 2023-04-27 12:35:47,995 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=8.09 vs. limit=2.0 2023-04-27 12:35:50,488 INFO [train.py:904] (2/8) Epoch 1, batch 900, loss[loss=0.6854, simple_loss=0.5785, pruned_loss=0.4692, over 16845.00 frames. ], tot_loss[loss=0.8059, simple_loss=0.672, pruned_loss=0.6267, over 3276362.58 frames. ], batch size: 42, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:35:57,044 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:36:00,422 INFO [optim.py:368] (2/8) 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 12:36:08,847 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:36:14,430 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:36:33,029 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7702, 2.5269, 2.5835, 2.6177, 2.6154, 2.3533, 2.4639, 2.6761], device='cuda:2'), covar=tensor([0.3156, 0.2523, 0.2721, 0.2691, 0.2852, 0.3630, 0.3476, 0.2646], device='cuda:2'), in_proj_covar=tensor([0.0035, 0.0029, 0.0033, 0.0032, 0.0032, 0.0034, 0.0035, 0.0033], device='cuda:2'), out_proj_covar=tensor([3.1423e-05, 2.8577e-05, 3.4273e-05, 3.1401e-05, 3.1060e-05, 3.4380e-05, 3.3617e-05, 3.2590e-05], device='cuda:2') 2023-04-27 12:36:37,790 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:36:39,482 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([6.0419, 4.8432, 5.0182, 5.7132, 5.0708, 5.2555, 5.9647, 5.8937], device='cuda:2'), covar=tensor([0.0521, 0.5005, 0.4145, 0.1488, 0.3082, 0.3054, 0.0660, 0.0748], device='cuda:2'), in_proj_covar=tensor([0.0061, 0.0072, 0.0073, 0.0068, 0.0067, 0.0069, 0.0062, 0.0062], device='cuda:2'), out_proj_covar=tensor([5.5211e-05, 6.8235e-05, 7.3876e-05, 5.7021e-05, 6.9796e-05, 6.0752e-05, 5.3844e-05, 5.8188e-05], device='cuda:2') 2023-04-27 12:36:42,345 INFO [train.py:904] (2/8) Epoch 1, batch 950, loss[loss=0.7169, simple_loss=0.6123, pruned_loss=0.4739, over 12129.00 frames. ], tot_loss[loss=0.7747, simple_loss=0.6516, pruned_loss=0.5821, over 3271843.58 frames. ], batch size: 247, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:36:44,103 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:49,971 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([6.2860, 5.7945, 6.2180, 6.2882, 6.2863, 6.2878, 6.2874, 6.2867], device='cuda:2'), covar=tensor([0.0385, 0.2597, 0.0921, 0.0376, 0.0330, 0.0406, 0.0491, 0.0380], device='cuda:2'), in_proj_covar=tensor([0.0038, 0.0053, 0.0047, 0.0038, 0.0036, 0.0039, 0.0042, 0.0038], device='cuda:2'), out_proj_covar=tensor([3.3534e-05, 4.7556e-05, 4.0870e-05, 3.1226e-05, 3.2238e-05, 3.2220e-05, 3.8304e-05, 3.2038e-05], device='cuda:2') 2023-04-27 12:37:03,518 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.61 vs. limit=5.0 2023-04-27 12:37:04,390 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 2023-04-27 12:37:35,474 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:37:35,994 INFO [train.py:904] (2/8) Epoch 1, batch 1000, loss[loss=0.5833, simple_loss=0.5212, pruned_loss=0.3497, over 17209.00 frames. ], tot_loss[loss=0.7386, simple_loss=0.6269, pruned_loss=0.5365, over 3287404.23 frames. ], batch size: 44, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:37:46,287 INFO [optim.py:368] (2/8) 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 12:37:51,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5875, 3.8862, 3.9421, 3.8755, 4.0080, 3.6340, 3.9381, 3.1242], device='cuda:2'), covar=tensor([0.3494, 0.4969, 0.6368, 0.4335, 0.6224, 0.5131, 0.6559, 1.4193], device='cuda:2'), in_proj_covar=tensor([0.0034, 0.0027, 0.0028, 0.0034, 0.0024, 0.0022, 0.0030, 0.0028], device='cuda:2'), out_proj_covar=tensor([2.7379e-05, 2.0721e-05, 2.1525e-05, 2.6987e-05, 1.8940e-05, 1.8032e-05, 2.4445e-05, 2.5341e-05], device='cuda:2') 2023-04-27 12:38:21,634 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:38:29,482 INFO [train.py:904] (2/8) Epoch 1, batch 1050, loss[loss=0.5939, simple_loss=0.5448, pruned_loss=0.3355, over 17131.00 frames. ], tot_loss[loss=0.7075, simple_loss=0.6064, pruned_loss=0.4968, over 3288563.21 frames. ], batch size: 47, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:38:52,746 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=2.21 vs. limit=2.0 2023-04-27 12:39:12,047 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:39:21,796 INFO [train.py:904] (2/8) Epoch 1, batch 1100, loss[loss=0.6241, simple_loss=0.5678, pruned_loss=0.357, over 17150.00 frames. ], tot_loss[loss=0.6782, simple_loss=0.5875, pruned_loss=0.4603, over 3302019.87 frames. ], batch size: 49, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:39:31,162 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 12:39:33,202 INFO [optim.py:368] (2/8) 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 12:40:17,536 INFO [train.py:904] (2/8) Epoch 1, batch 1150, loss[loss=0.5749, simple_loss=0.5345, pruned_loss=0.3141, over 17170.00 frames. ], tot_loss[loss=0.6534, simple_loss=0.5722, pruned_loss=0.4291, over 3301675.75 frames. ], batch size: 46, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:40:19,549 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:41:10,398 INFO [train.py:904] (2/8) Epoch 1, batch 1200, loss[loss=0.6002, simple_loss=0.5421, pruned_loss=0.3448, over 17068.00 frames. ], tot_loss[loss=0.6293, simple_loss=0.5563, pruned_loss=0.4013, over 3308150.28 frames. ], batch size: 53, lr: 4.93e-02, grad_scale: 2.0 2023-04-27 12:41:12,681 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:41:21,558 INFO [optim.py:368] (2/8) 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 12:41:25,710 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:41:26,775 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-27 12:41:29,956 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:41:34,770 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:41:57,143 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:42:03,009 INFO [train.py:904] (2/8) Epoch 1, batch 1250, loss[loss=0.5316, simple_loss=0.502, pruned_loss=0.2815, over 16635.00 frames. ], tot_loss[loss=0.611, simple_loss=0.5448, pruned_loss=0.3793, over 3309799.43 frames. ], batch size: 62, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:42:17,897 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:42:24,718 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:49,321 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:56,579 INFO [train.py:904] (2/8) Epoch 1, batch 1300, loss[loss=0.4555, simple_loss=0.4415, pruned_loss=0.2299, over 16762.00 frames. ], tot_loss[loss=0.5928, simple_loss=0.5329, pruned_loss=0.3593, over 3298978.82 frames. ], batch size: 39, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:43:07,782 INFO [optim.py:368] (2/8) 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 12:43:51,773 INFO [train.py:904] (2/8) Epoch 1, batch 1350, loss[loss=0.484, simple_loss=0.4534, pruned_loss=0.2596, over 16511.00 frames. ], tot_loss[loss=0.577, simple_loss=0.5237, pruned_loss=0.341, over 3302841.13 frames. ], batch size: 75, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:44:49,666 INFO [train.py:904] (2/8) Epoch 1, batch 1400, loss[loss=0.5147, simple_loss=0.4856, pruned_loss=0.2727, over 17236.00 frames. ], tot_loss[loss=0.563, simple_loss=0.5149, pruned_loss=0.3259, over 3296348.34 frames. ], batch size: 44, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:45:00,101 INFO [optim.py:368] (2/8) 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 12:45:27,734 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:45:46,438 INFO [train.py:904] (2/8) Epoch 1, batch 1450, loss[loss=0.5252, simple_loss=0.4801, pruned_loss=0.2912, over 16909.00 frames. ], tot_loss[loss=0.5483, simple_loss=0.5055, pruned_loss=0.3113, over 3294775.30 frames. ], batch size: 109, lr: 4.90e-02, grad_scale: 2.0 2023-04-27 12:46:34,497 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:46:41,359 INFO [train.py:904] (2/8) Epoch 1, batch 1500, loss[loss=0.4299, simple_loss=0.4202, pruned_loss=0.2158, over 16778.00 frames. ], tot_loss[loss=0.5344, simple_loss=0.4969, pruned_loss=0.2976, over 3311073.55 frames. ], batch size: 39, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:46:44,184 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:46:50,549 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-04-27 12:46:51,040 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:46:52,703 INFO [optim.py:368] (2/8) 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 12:46:54,992 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5815, 4.8035, 5.1122, 5.2950, 5.5125, 5.2100, 4.8995, 5.5377], device='cuda:2'), covar=tensor([0.0171, 0.0545, 0.0922, 0.0415, 0.0426, 0.0585, 0.0571, 0.0283], device='cuda:2'), in_proj_covar=tensor([0.0077, 0.0086, 0.0098, 0.0094, 0.0077, 0.0084, 0.0092, 0.0080], device='cuda:2'), out_proj_covar=tensor([6.3595e-05, 7.9487e-05, 9.7349e-05, 8.0498e-05, 7.4050e-05, 7.3273e-05, 7.7003e-05, 7.2008e-05], device='cuda:2') 2023-04-27 12:47:38,770 INFO [train.py:904] (2/8) Epoch 1, batch 1550, loss[loss=0.5067, simple_loss=0.4749, pruned_loss=0.2707, over 16732.00 frames. ], tot_loss[loss=0.5259, simple_loss=0.4916, pruned_loss=0.289, over 3303119.00 frames. ], batch size: 134, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:47:39,031 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:47:56,140 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:48:21,107 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1541, 4.2311, 3.7466, 4.3546, 4.2096, 4.4681, 4.1745, 4.2069], device='cuda:2'), covar=tensor([0.0568, 0.0518, 0.1122, 0.0931, 0.0986, 0.0600, 0.0818, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0076, 0.0073, 0.0091, 0.0075, 0.0076, 0.0072, 0.0072, 0.0073], device='cuda:2'), out_proj_covar=tensor([6.8164e-05, 6.6864e-05, 9.0373e-05, 7.4084e-05, 7.0336e-05, 7.0175e-05, 6.6244e-05, 6.8280e-05], device='cuda:2') 2023-04-27 12:48:35,336 INFO [train.py:904] (2/8) Epoch 1, batch 1600, loss[loss=0.5862, simple_loss=0.5396, pruned_loss=0.32, over 16138.00 frames. ], tot_loss[loss=0.5169, simple_loss=0.4868, pruned_loss=0.2799, over 3313061.89 frames. ], batch size: 164, lr: 4.88e-02, grad_scale: 4.0 2023-04-27 12:48:39,156 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 12:48:47,163 INFO [optim.py:368] (2/8) 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 12:48:47,465 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:49:05,659 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:49:12,202 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-04-27 12:49:21,415 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-27 12:49:32,287 INFO [train.py:904] (2/8) Epoch 1, batch 1650, loss[loss=0.4588, simple_loss=0.4501, pruned_loss=0.2307, over 17235.00 frames. ], tot_loss[loss=0.5099, simple_loss=0.4834, pruned_loss=0.2726, over 3315978.52 frames. ], batch size: 44, lr: 4.87e-02, grad_scale: 4.0 2023-04-27 12:49:40,711 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0379, 5.0156, 4.7930, 4.8214, 4.7328, 5.1079, 5.0794, 4.6333], device='cuda:2'), covar=tensor([0.0369, 0.0505, 0.0676, 0.0825, 0.0865, 0.0509, 0.0529, 0.1075], device='cuda:2'), in_proj_covar=tensor([0.0101, 0.0118, 0.0116, 0.0127, 0.0126, 0.0101, 0.0095, 0.0126], device='cuda:2'), out_proj_covar=tensor([8.8438e-05, 1.0722e-04, 9.8659e-05, 1.0979e-04, 1.1736e-04, 8.9386e-05, 8.4854e-05, 1.1673e-04], device='cuda:2') 2023-04-27 12:49:49,422 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4232, 3.4296, 3.2104, 3.1785, 3.4521, 3.1169, 3.3129, 3.3012], device='cuda:2'), covar=tensor([0.0441, 0.0338, 0.0654, 0.0494, 0.0438, 0.0698, 0.0526, 0.0432], device='cuda:2'), in_proj_covar=tensor([0.0071, 0.0065, 0.0063, 0.0064, 0.0070, 0.0074, 0.0072, 0.0068], device='cuda:2'), out_proj_covar=tensor([6.7252e-05, 5.9080e-05, 6.2975e-05, 6.0937e-05, 6.5230e-05, 7.1135e-05, 6.5560e-05, 6.2688e-05], device='cuda:2') 2023-04-27 12:49:51,735 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4517, 4.9128, 5.2092, 5.3776, 5.5735, 5.3927, 5.0283, 5.4792], device='cuda:2'), covar=tensor([0.0251, 0.0593, 0.0786, 0.0425, 0.0407, 0.0414, 0.0541, 0.0269], device='cuda:2'), in_proj_covar=tensor([0.0089, 0.0092, 0.0105, 0.0102, 0.0086, 0.0092, 0.0103, 0.0086], device='cuda:2'), out_proj_covar=tensor([7.1118e-05, 8.7991e-05, 1.0770e-04, 8.7229e-05, 8.3994e-05, 8.2914e-05, 9.0190e-05, 7.6355e-05], device='cuda:2') 2023-04-27 12:49:57,848 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:50:03,657 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-04-27 12:50:24,164 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9266, 4.1272, 4.5319, 4.4391, 4.0653, 3.6724, 4.5225, 4.7688], device='cuda:2'), covar=tensor([0.1104, 0.0454, 0.0397, 0.0529, 0.0680, 0.0963, 0.0364, 0.0265], device='cuda:2'), in_proj_covar=tensor([0.0062, 0.0060, 0.0053, 0.0057, 0.0065, 0.0062, 0.0057, 0.0051], device='cuda:2'), out_proj_covar=tensor([5.1689e-05, 4.8914e-05, 4.6607e-05, 4.5340e-05, 5.8117e-05, 5.0452e-05, 4.6404e-05, 4.5090e-05], device='cuda:2') 2023-04-27 12:50:29,818 INFO [train.py:904] (2/8) Epoch 1, batch 1700, loss[loss=0.3952, simple_loss=0.4113, pruned_loss=0.1835, over 16791.00 frames. ], tot_loss[loss=0.505, simple_loss=0.4818, pruned_loss=0.2671, over 3305937.28 frames. ], batch size: 39, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:50:41,826 INFO [optim.py:368] (2/8) 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 12:50:43,271 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7673, 3.3464, 3.7737, 4.0176, 2.3689, 3.8731, 3.4786, 3.7443], device='cuda:2'), covar=tensor([0.0480, 0.0294, 0.0280, 0.0328, 0.0571, 0.0233, 0.0270, 0.0207], device='cuda:2'), in_proj_covar=tensor([0.0040, 0.0036, 0.0030, 0.0028, 0.0025, 0.0030, 0.0037, 0.0031], device='cuda:2'), out_proj_covar=tensor([3.7624e-05, 3.0016e-05, 2.4987e-05, 2.4356e-05, 2.3851e-05, 2.5443e-05, 3.1898e-05, 2.4985e-05], device='cuda:2') 2023-04-27 12:51:19,573 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-27 12:51:28,762 INFO [train.py:904] (2/8) Epoch 1, batch 1750, loss[loss=0.46, simple_loss=0.4727, pruned_loss=0.2187, over 16711.00 frames. ], tot_loss[loss=0.4972, simple_loss=0.478, pruned_loss=0.2598, over 3314318.52 frames. ], batch size: 57, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:52:15,501 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:52:27,926 INFO [train.py:904] (2/8) Epoch 1, batch 1800, loss[loss=0.4819, simple_loss=0.4894, pruned_loss=0.2338, over 17068.00 frames. ], tot_loss[loss=0.4919, simple_loss=0.4764, pruned_loss=0.2543, over 3312111.64 frames. ], batch size: 53, lr: 4.85e-02, grad_scale: 4.0 2023-04-27 12:52:37,327 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:52:38,890 INFO [optim.py:368] (2/8) 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 12:53:10,948 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:25,087 INFO [train.py:904] (2/8) Epoch 1, batch 1850, loss[loss=0.4339, simple_loss=0.4394, pruned_loss=0.2121, over 16858.00 frames. ], tot_loss[loss=0.4833, simple_loss=0.4726, pruned_loss=0.247, over 3322519.24 frames. ], batch size: 42, lr: 4.84e-02, grad_scale: 4.0 2023-04-27 12:53:32,841 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:32,955 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:53:48,607 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:54:21,365 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:54:23,725 INFO [train.py:904] (2/8) Epoch 1, batch 1900, loss[loss=0.4578, simple_loss=0.4574, pruned_loss=0.2279, over 15697.00 frames. ], tot_loss[loss=0.4741, simple_loss=0.4681, pruned_loss=0.2396, over 3318733.33 frames. ], batch size: 190, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:54:36,242 INFO [optim.py:368] (2/8) 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 12:54:44,522 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:54:49,110 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:55:00,072 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:55:22,815 INFO [train.py:904] (2/8) Epoch 1, batch 1950, loss[loss=0.3835, simple_loss=0.4169, pruned_loss=0.1738, over 17227.00 frames. ], tot_loss[loss=0.465, simple_loss=0.4635, pruned_loss=0.2325, over 3302413.72 frames. ], batch size: 44, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:55:42,112 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:55:53,396 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7583, 3.8030, 3.5317, 4.0246, 3.8348, 4.1165, 3.9051, 3.8172], device='cuda:2'), covar=tensor([0.0426, 0.0492, 0.0832, 0.0622, 0.0610, 0.0384, 0.0555, 0.0562], device='cuda:2'), in_proj_covar=tensor([0.0095, 0.0095, 0.0124, 0.0100, 0.0095, 0.0096, 0.0087, 0.0100], device='cuda:2'), out_proj_covar=tensor([9.0226e-05, 9.1946e-05, 1.3017e-04, 1.0592e-04, 9.3855e-05, 9.5339e-05, 8.5127e-05, 1.0105e-04], device='cuda:2') 2023-04-27 12:56:23,561 INFO [train.py:904] (2/8) Epoch 1, batch 2000, loss[loss=0.4038, simple_loss=0.4309, pruned_loss=0.1883, over 17216.00 frames. ], tot_loss[loss=0.4581, simple_loss=0.4599, pruned_loss=0.2275, over 3312787.81 frames. ], batch size: 44, lr: 4.82e-02, grad_scale: 8.0 2023-04-27 12:56:34,070 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7155, 3.9434, 4.1701, 4.0852, 2.7733, 4.1405, 3.9283, 4.1677], device='cuda:2'), covar=tensor([0.0687, 0.0115, 0.0157, 0.0216, 0.0549, 0.0160, 0.0215, 0.0114], device='cuda:2'), in_proj_covar=tensor([0.0052, 0.0037, 0.0031, 0.0027, 0.0025, 0.0030, 0.0038, 0.0032], device='cuda:2'), out_proj_covar=tensor([5.1382e-05, 3.1708e-05, 2.6306e-05, 2.3937e-05, 2.4227e-05, 2.5808e-05, 3.3528e-05, 2.5472e-05], device='cuda:2') 2023-04-27 12:56:36,802 INFO [optim.py:368] (2/8) 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 12:57:27,604 INFO [train.py:904] (2/8) Epoch 1, batch 2050, loss[loss=0.3814, simple_loss=0.4233, pruned_loss=0.1697, over 17096.00 frames. ], tot_loss[loss=0.4468, simple_loss=0.4531, pruned_loss=0.2198, over 3311926.05 frames. ], batch size: 47, lr: 4.81e-02, grad_scale: 8.0 2023-04-27 12:58:19,032 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:58:32,223 INFO [train.py:904] (2/8) Epoch 1, batch 2100, loss[loss=0.4354, simple_loss=0.4458, pruned_loss=0.2125, over 16858.00 frames. ], tot_loss[loss=0.4367, simple_loss=0.4471, pruned_loss=0.2128, over 3314815.53 frames. ], batch size: 116, lr: 4.80e-02, grad_scale: 16.0 2023-04-27 12:58:45,933 INFO [optim.py:368] (2/8) 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 12:59:19,751 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:59:35,891 INFO [train.py:904] (2/8) Epoch 1, batch 2150, loss[loss=0.3911, simple_loss=0.4408, pruned_loss=0.1707, over 17132.00 frames. ], tot_loss[loss=0.4335, simple_loss=0.4461, pruned_loss=0.2102, over 3310105.75 frames. ], batch size: 49, lr: 4.79e-02, grad_scale: 16.0 2023-04-27 13:00:32,070 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:40,252 INFO [train.py:904] (2/8) Epoch 1, batch 2200, loss[loss=0.3863, simple_loss=0.4136, pruned_loss=0.1795, over 16104.00 frames. ], tot_loss[loss=0.4256, simple_loss=0.4423, pruned_loss=0.2043, over 3312567.50 frames. ], batch size: 35, lr: 4.78e-02, grad_scale: 16.0 2023-04-27 13:00:45,182 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2624, 3.4257, 3.2467, 4.0219, 4.1977, 1.4884, 3.8050, 4.0493], device='cuda:2'), covar=tensor([0.5444, 0.1881, 0.3177, 0.0809, 0.1835, 0.9108, 0.0954, 0.0315], device='cuda:2'), in_proj_covar=tensor([0.0060, 0.0035, 0.0055, 0.0038, 0.0033, 0.0072, 0.0035, 0.0023], device='cuda:2'), out_proj_covar=tensor([5.7772e-05, 3.5176e-05, 5.1144e-05, 2.9494e-05, 3.3524e-05, 6.2783e-05, 2.9289e-05, 2.1815e-05], device='cuda:2') 2023-04-27 13:00:53,195 INFO [optim.py:368] (2/8) 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 13:00:56,685 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:01,850 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3059, 4.8560, 5.2952, 5.3372, 4.6276, 5.1120, 5.4116, 4.7660], device='cuda:2'), covar=tensor([0.0381, 0.0501, 0.0217, 0.0127, 0.0787, 0.0366, 0.0201, 0.0363], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0071, 0.0094, 0.0072, 0.0098, 0.0081, 0.0072, 0.0078], device='cuda:2'), 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:2') 2023-04-27 13:01:07,617 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 13:01:11,879 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:44,909 INFO [train.py:904] (2/8) Epoch 1, batch 2250, loss[loss=0.5314, simple_loss=0.5108, pruned_loss=0.276, over 11620.00 frames. ], tot_loss[loss=0.4177, simple_loss=0.4378, pruned_loss=0.1986, over 3315221.64 frames. ], batch size: 246, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:02:05,812 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:02:08,814 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:02:49,153 INFO [train.py:904] (2/8) Epoch 1, batch 2300, loss[loss=0.3271, simple_loss=0.3787, pruned_loss=0.1377, over 16339.00 frames. ], tot_loss[loss=0.4129, simple_loss=0.4349, pruned_loss=0.1953, over 3313867.38 frames. ], batch size: 36, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:03:01,864 INFO [optim.py:368] (2/8) 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 13:03:07,889 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:20,378 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2084, 3.2624, 3.2427, 2.8193, 2.3839, 2.8095, 3.5291, 3.2942], device='cuda:2'), covar=tensor([0.2043, 0.1656, 0.0929, 0.1859, 0.2321, 0.2337, 0.0831, 0.0887], device='cuda:2'), in_proj_covar=tensor([0.0025, 0.0026, 0.0033, 0.0034, 0.0035, 0.0037, 0.0022, 0.0023], device='cuda:2'), out_proj_covar=tensor([2.3865e-05, 2.1730e-05, 2.8541e-05, 2.9286e-05, 2.9065e-05, 3.0961e-05, 1.8353e-05, 2.0801e-05], device='cuda:2') 2023-04-27 13:03:39,679 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:03:53,137 INFO [train.py:904] (2/8) Epoch 1, batch 2350, loss[loss=0.4044, simple_loss=0.4187, pruned_loss=0.195, over 16876.00 frames. ], tot_loss[loss=0.4082, simple_loss=0.4312, pruned_loss=0.1925, over 3311415.17 frames. ], batch size: 96, lr: 4.76e-02, grad_scale: 16.0 2023-04-27 13:04:40,340 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 13:04:54,413 INFO [train.py:904] (2/8) Epoch 1, batch 2400, loss[loss=0.39, simple_loss=0.4221, pruned_loss=0.179, over 16857.00 frames. ], tot_loss[loss=0.4057, simple_loss=0.4307, pruned_loss=0.1902, over 3319154.36 frames. ], batch size: 39, lr: 4.75e-02, grad_scale: 16.0 2023-04-27 13:04:55,593 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:04:58,697 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3665, 3.3655, 3.5204, 3.3339, 3.3599, 3.3421, 3.7960, 3.9179], device='cuda:2'), covar=tensor([0.0298, 0.0327, 0.0235, 0.0591, 0.0222, 0.0392, 0.0200, 0.0159], device='cuda:2'), in_proj_covar=tensor([0.0036, 0.0036, 0.0034, 0.0036, 0.0036, 0.0033, 0.0034, 0.0031], device='cuda:2'), out_proj_covar=tensor([3.0442e-05, 3.1251e-05, 2.7580e-05, 3.0555e-05, 3.0747e-05, 2.7882e-05, 2.7656e-05, 2.6099e-05], device='cuda:2') 2023-04-27 13:05:07,295 INFO [optim.py:368] (2/8) 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 13:05:26,923 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1073, 3.9983, 4.0838, 4.3775, 3.5777, 4.3245, 3.1919, 4.6109], device='cuda:2'), covar=tensor([0.0947, 0.1248, 0.0866, 0.0920, 0.4982, 0.0898, 0.2220, 0.1560], device='cuda:2'), in_proj_covar=tensor([0.0045, 0.0048, 0.0044, 0.0040, 0.0077, 0.0046, 0.0054, 0.0038], device='cuda:2'), out_proj_covar=tensor([3.7213e-05, 4.0360e-05, 3.4842e-05, 3.9927e-05, 6.9649e-05, 3.8064e-05, 4.8540e-05, 3.9103e-05], device='cuda:2') 2023-04-27 13:05:55,891 INFO [train.py:904] (2/8) Epoch 1, batch 2450, loss[loss=0.3679, simple_loss=0.4084, pruned_loss=0.1637, over 17240.00 frames. ], tot_loss[loss=0.4012, simple_loss=0.4289, pruned_loss=0.1867, over 3325071.75 frames. ], batch size: 45, lr: 4.74e-02, grad_scale: 16.0 2023-04-27 13:05:57,891 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1421, 4.2258, 4.3067, 4.3492, 4.6660, 4.2267, 4.1836, 4.5999], device='cuda:2'), covar=tensor([0.0331, 0.0299, 0.0470, 0.0395, 0.0320, 0.0310, 0.0467, 0.0231], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0101, 0.0120, 0.0118, 0.0112, 0.0107, 0.0112, 0.0095], device='cuda:2'), out_proj_covar=tensor([1.0487e-04, 1.1674e-04, 1.3016e-04, 1.1804e-04, 1.2028e-04, 1.1359e-04, 1.1292e-04, 9.0632e-05], device='cuda:2') 2023-04-27 13:06:51,079 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:06:58,990 INFO [train.py:904] (2/8) Epoch 1, batch 2500, loss[loss=0.4288, simple_loss=0.4634, pruned_loss=0.1971, over 16662.00 frames. ], tot_loss[loss=0.3961, simple_loss=0.4259, pruned_loss=0.1831, over 3316587.80 frames. ], batch size: 57, lr: 4.73e-02, grad_scale: 16.0 2023-04-27 13:07:06,589 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-27 13:07:11,499 INFO [optim.py:368] (2/8) 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 13:07:14,822 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:29,302 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:01,513 INFO [train.py:904] (2/8) Epoch 1, batch 2550, loss[loss=0.32, simple_loss=0.3762, pruned_loss=0.1319, over 16796.00 frames. ], tot_loss[loss=0.3936, simple_loss=0.4247, pruned_loss=0.1811, over 3328855.26 frames. ], batch size: 39, lr: 4.72e-02, grad_scale: 16.0 2023-04-27 13:08:06,999 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-27 13:08:15,979 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:32,316 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:09:07,692 INFO [train.py:904] (2/8) Epoch 1, batch 2600, loss[loss=0.3588, simple_loss=0.415, pruned_loss=0.1512, over 17113.00 frames. ], tot_loss[loss=0.3897, simple_loss=0.4222, pruned_loss=0.1786, over 3329140.35 frames. ], batch size: 47, lr: 4.71e-02, grad_scale: 16.0 2023-04-27 13:09:20,088 INFO [optim.py:368] (2/8) 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 13:10:11,887 INFO [train.py:904] (2/8) Epoch 1, batch 2650, loss[loss=0.3958, simple_loss=0.4208, pruned_loss=0.1854, over 16850.00 frames. ], tot_loss[loss=0.3846, simple_loss=0.4195, pruned_loss=0.1748, over 3328753.30 frames. ], batch size: 102, lr: 4.70e-02, grad_scale: 16.0 2023-04-27 13:10:55,323 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:11:09,970 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:11:15,039 INFO [train.py:904] (2/8) Epoch 1, batch 2700, loss[loss=0.4064, simple_loss=0.4404, pruned_loss=0.1862, over 17033.00 frames. ], tot_loss[loss=0.3818, simple_loss=0.4182, pruned_loss=0.1727, over 3328811.72 frames. ], batch size: 53, lr: 4.69e-02, grad_scale: 16.0 2023-04-27 13:11:29,642 INFO [optim.py:368] (2/8) 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 13:11:40,568 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 13:12:13,608 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:12:20,232 INFO [train.py:904] (2/8) Epoch 1, batch 2750, loss[loss=0.3486, simple_loss=0.401, pruned_loss=0.148, over 17190.00 frames. ], tot_loss[loss=0.3761, simple_loss=0.415, pruned_loss=0.1686, over 3332629.84 frames. ], batch size: 44, lr: 4.68e-02, grad_scale: 16.0 2023-04-27 13:13:23,779 INFO [train.py:904] (2/8) Epoch 1, batch 2800, loss[loss=0.374, simple_loss=0.4058, pruned_loss=0.1711, over 16871.00 frames. ], tot_loss[loss=0.3758, simple_loss=0.4146, pruned_loss=0.1685, over 3333262.03 frames. ], batch size: 96, lr: 4.67e-02, grad_scale: 16.0 2023-04-27 13:13:24,477 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-27 13:13:35,358 INFO [optim.py:368] (2/8) 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 13:13:54,296 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:14:25,636 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 2023-04-27 13:14:25,999 INFO [train.py:904] (2/8) Epoch 1, batch 2850, loss[loss=0.3476, simple_loss=0.3907, pruned_loss=0.1523, over 17245.00 frames. ], tot_loss[loss=0.3739, simple_loss=0.4128, pruned_loss=0.1675, over 3326911.44 frames. ], batch size: 45, lr: 4.66e-02, grad_scale: 16.0 2023-04-27 13:14:58,140 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-04-27 13:15:09,598 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:15:27,286 INFO [train.py:904] (2/8) Epoch 1, batch 2900, loss[loss=0.3576, simple_loss=0.3765, pruned_loss=0.1693, over 16547.00 frames. ], tot_loss[loss=0.3719, simple_loss=0.4105, pruned_loss=0.1667, over 3331339.96 frames. ], batch size: 146, lr: 4.65e-02, grad_scale: 16.0 2023-04-27 13:15:40,739 INFO [optim.py:368] (2/8) 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 13:16:13,973 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:16:32,273 INFO [train.py:904] (2/8) Epoch 1, batch 2950, loss[loss=0.3205, simple_loss=0.3739, pruned_loss=0.1335, over 17233.00 frames. ], tot_loss[loss=0.3688, simple_loss=0.4077, pruned_loss=0.1649, over 3331780.57 frames. ], batch size: 44, lr: 4.64e-02, grad_scale: 16.0 2023-04-27 13:17:29,914 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:17:32,178 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 13:17:35,318 INFO [train.py:904] (2/8) Epoch 1, batch 3000, loss[loss=0.3742, simple_loss=0.4017, pruned_loss=0.1734, over 16901.00 frames. ], tot_loss[loss=0.3673, simple_loss=0.4063, pruned_loss=0.1641, over 3329475.70 frames. ], batch size: 109, lr: 4.63e-02, grad_scale: 16.0 2023-04-27 13:17:35,319 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 13:17:45,048 INFO [train.py:938] (2/8) Epoch 1, validation: loss=0.2847, simple_loss=0.3895, pruned_loss=0.08992, over 944034.00 frames. 2023-04-27 13:17:45,048 INFO [train.py:939] (2/8) Maximum memory allocated so far is 15888MB 2023-04-27 13:17:59,339 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 13:17:59,838 INFO [optim.py:368] (2/8) 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 13:18:09,234 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:18:36,921 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:18:41,576 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:18:50,094 INFO [train.py:904] (2/8) Epoch 1, batch 3050, loss[loss=0.4189, simple_loss=0.4388, pruned_loss=0.1995, over 16721.00 frames. ], tot_loss[loss=0.3643, simple_loss=0.4037, pruned_loss=0.1624, over 3331860.75 frames. ], batch size: 134, lr: 4.62e-02, grad_scale: 16.0 2023-04-27 13:19:09,102 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4606, 5.7125, 5.2943, 5.8285, 5.2046, 5.4113, 5.3751, 5.8792], device='cuda:2'), covar=tensor([0.0394, 0.0592, 0.0596, 0.0295, 0.0600, 0.0339, 0.0435, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0152, 0.0129, 0.0100, 0.0131, 0.0114, 0.0134, 0.0098], device='cuda:2'), out_proj_covar=tensor([1.1846e-04, 1.4134e-04, 1.1235e-04, 8.0473e-05, 1.1306e-04, 9.7849e-05, 1.1995e-04, 9.1642e-05], device='cuda:2') 2023-04-27 13:19:27,589 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 13:19:53,976 INFO [train.py:904] (2/8) Epoch 1, batch 3100, loss[loss=0.3721, simple_loss=0.4249, pruned_loss=0.1597, over 17082.00 frames. ], tot_loss[loss=0.3636, simple_loss=0.4028, pruned_loss=0.1622, over 3321346.68 frames. ], batch size: 55, lr: 4.61e-02, grad_scale: 16.0 2023-04-27 13:20:06,842 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3391, 6.0794, 5.9021, 5.7784, 6.0034, 6.1918, 6.2845, 5.7634], device='cuda:2'), covar=tensor([0.0453, 0.0601, 0.0580, 0.0989, 0.1069, 0.0413, 0.0482, 0.0939], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0179, 0.0147, 0.0159, 0.0190, 0.0130, 0.0128, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 13:20:07,649 INFO [optim.py:368] (2/8) 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 13:20:10,047 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 13:20:48,765 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5773, 4.4193, 4.7720, 4.8173, 5.0318, 4.7401, 4.5805, 4.8821], device='cuda:2'), covar=tensor([0.0291, 0.0261, 0.0403, 0.0376, 0.0318, 0.0263, 0.0451, 0.0170], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0100, 0.0129, 0.0128, 0.0127, 0.0112, 0.0120, 0.0097], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:2') 2023-04-27 13:21:00,201 INFO [train.py:904] (2/8) Epoch 1, batch 3150, loss[loss=0.353, simple_loss=0.407, pruned_loss=0.1495, over 17053.00 frames. ], tot_loss[loss=0.3622, simple_loss=0.4019, pruned_loss=0.1613, over 3310471.68 frames. ], batch size: 50, lr: 4.60e-02, grad_scale: 16.0 2023-04-27 13:21:07,904 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-04-27 13:21:38,517 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:22:05,296 INFO [train.py:904] (2/8) Epoch 1, batch 3200, loss[loss=0.2959, simple_loss=0.3519, pruned_loss=0.12, over 17020.00 frames. ], tot_loss[loss=0.3567, simple_loss=0.398, pruned_loss=0.1577, over 3323445.23 frames. ], batch size: 41, lr: 4.59e-02, grad_scale: 16.0 2023-04-27 13:22:17,347 INFO [optim.py:368] (2/8) 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 13:22:57,040 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 13:23:05,492 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3714, 4.3543, 3.8707, 4.4292, 4.2034, 4.4367, 4.2417, 4.3667], device='cuda:2'), covar=tensor([0.0777, 0.1099, 0.1532, 0.0568, 0.1023, 0.0641, 0.0931, 0.0726], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0171, 0.0138, 0.0109, 0.0146, 0.0124, 0.0144, 0.0101], device='cuda:2'), out_proj_covar=tensor([1.3031e-04, 1.5914e-04, 1.2098e-04, 8.9903e-05, 1.2700e-04, 1.0759e-04, 1.3276e-04, 9.7194e-05], device='cuda:2') 2023-04-27 13:23:07,319 INFO [train.py:904] (2/8) Epoch 1, batch 3250, loss[loss=0.3741, simple_loss=0.4087, pruned_loss=0.1697, over 16468.00 frames. ], tot_loss[loss=0.3533, simple_loss=0.3957, pruned_loss=0.1555, over 3329486.65 frames. ], batch size: 68, lr: 4.58e-02, grad_scale: 16.0 2023-04-27 13:23:21,489 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1847, 4.7717, 3.6711, 4.9782, 4.6164, 1.9284, 5.3681, 4.8900], device='cuda:2'), covar=tensor([0.0725, 0.0164, 0.0597, 0.0053, 0.0441, 0.1546, 0.0030, 0.0029], device='cuda:2'), in_proj_covar=tensor([0.0114, 0.0064, 0.0096, 0.0053, 0.0059, 0.0099, 0.0053, 0.0029], device='cuda:2'), out_proj_covar=tensor([1.1595e-04, 7.0858e-05, 9.4834e-05, 4.9739e-05, 7.2446e-05, 9.4360e-05, 5.2833e-05, 3.2691e-05], device='cuda:2') 2023-04-27 13:23:51,485 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-27 13:23:56,319 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:24:01,842 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:12,251 INFO [train.py:904] (2/8) Epoch 1, batch 3300, loss[loss=0.3404, simple_loss=0.3855, pruned_loss=0.1476, over 16766.00 frames. ], tot_loss[loss=0.3543, simple_loss=0.3972, pruned_loss=0.1557, over 3326098.24 frames. ], batch size: 83, lr: 4.57e-02, grad_scale: 16.0 2023-04-27 13:24:25,122 INFO [optim.py:368] (2/8) 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 13:25:03,543 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:25:16,109 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 13:25:17,759 INFO [train.py:904] (2/8) Epoch 1, batch 3350, loss[loss=0.3283, simple_loss=0.3767, pruned_loss=0.14, over 17257.00 frames. ], tot_loss[loss=0.3546, simple_loss=0.3975, pruned_loss=0.1559, over 3314342.08 frames. ], batch size: 45, lr: 4.56e-02, grad_scale: 16.0 2023-04-27 13:25:49,835 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:26:09,263 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:26:24,812 INFO [train.py:904] (2/8) Epoch 1, batch 3400, loss[loss=0.3561, simple_loss=0.3903, pruned_loss=0.161, over 16866.00 frames. ], tot_loss[loss=0.3506, simple_loss=0.3951, pruned_loss=0.1531, over 3318041.01 frames. ], batch size: 116, lr: 4.55e-02, grad_scale: 16.0 2023-04-27 13:26:39,219 INFO [optim.py:368] (2/8) 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 13:27:32,305 INFO [train.py:904] (2/8) Epoch 1, batch 3450, loss[loss=0.3369, simple_loss=0.3762, pruned_loss=0.1488, over 16792.00 frames. ], tot_loss[loss=0.3479, simple_loss=0.3922, pruned_loss=0.1518, over 3317038.07 frames. ], batch size: 83, lr: 4.54e-02, grad_scale: 16.0 2023-04-27 13:28:12,487 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:28:39,711 INFO [train.py:904] (2/8) Epoch 1, batch 3500, loss[loss=0.3275, simple_loss=0.3844, pruned_loss=0.1352, over 16703.00 frames. ], tot_loss[loss=0.3447, simple_loss=0.3898, pruned_loss=0.1498, over 3320950.90 frames. ], batch size: 57, lr: 4.53e-02, grad_scale: 16.0 2023-04-27 13:28:53,813 INFO [optim.py:368] (2/8) 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 13:29:16,289 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:29:40,076 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:29:46,554 INFO [train.py:904] (2/8) Epoch 1, batch 3550, loss[loss=0.3372, simple_loss=0.3988, pruned_loss=0.1378, over 16716.00 frames. ], tot_loss[loss=0.3418, simple_loss=0.3871, pruned_loss=0.1483, over 3326503.99 frames. ], batch size: 57, lr: 4.51e-02, grad_scale: 16.0 2023-04-27 13:30:44,918 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:30:45,522 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-04-27 13:30:54,497 INFO [train.py:904] (2/8) Epoch 1, batch 3600, loss[loss=0.3345, simple_loss=0.3702, pruned_loss=0.1494, over 16799.00 frames. ], tot_loss[loss=0.3391, simple_loss=0.3845, pruned_loss=0.1468, over 3322397.33 frames. ], batch size: 102, lr: 4.50e-02, grad_scale: 16.0 2023-04-27 13:31:04,175 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:31:08,701 INFO [optim.py:368] (2/8) 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 13:31:51,882 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:31:56,475 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:32:06,055 INFO [train.py:904] (2/8) Epoch 1, batch 3650, loss[loss=0.3, simple_loss=0.3429, pruned_loss=0.1285, over 16445.00 frames. ], tot_loss[loss=0.3362, simple_loss=0.3813, pruned_loss=0.1456, over 3323261.22 frames. ], batch size: 75, lr: 4.49e-02, grad_scale: 16.0 2023-04-27 13:32:13,707 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4815, 4.8306, 4.5413, 4.5851, 4.6992, 4.9846, 5.0097, 4.4924], device='cuda:2'), covar=tensor([0.0659, 0.0820, 0.0794, 0.1254, 0.1593, 0.0687, 0.0575, 0.1573], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0181, 0.0148, 0.0157, 0.0188, 0.0136, 0.0134, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 13:32:42,344 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:32:43,510 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:32:50,821 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2964, 3.6965, 3.6755, 3.3509, 3.6474, 3.6190, 3.7894, 3.8297], device='cuda:2'), covar=tensor([0.0249, 0.0147, 0.0115, 0.0183, 0.0134, 0.0150, 0.0144, 0.0117], device='cuda:2'), in_proj_covar=tensor([0.0044, 0.0031, 0.0030, 0.0040, 0.0032, 0.0035, 0.0041, 0.0036], device='cuda:2'), out_proj_covar=tensor([6.1634e-05, 4.6189e-05, 4.3856e-05, 5.3043e-05, 4.5563e-05, 5.7252e-05, 5.6398e-05, 4.9205e-05], device='cuda:2') 2023-04-27 13:33:20,376 INFO [train.py:904] (2/8) Epoch 1, batch 3700, loss[loss=0.3731, simple_loss=0.3922, pruned_loss=0.177, over 11198.00 frames. ], tot_loss[loss=0.3339, simple_loss=0.3771, pruned_loss=0.1453, over 3293752.24 frames. ], batch size: 246, lr: 4.48e-02, grad_scale: 16.0 2023-04-27 13:33:30,691 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 13:33:35,072 INFO [optim.py:368] (2/8) 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 13:33:54,062 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:33:58,474 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 13:34:15,079 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:34:20,077 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 13:34:33,710 INFO [train.py:904] (2/8) Epoch 1, batch 3750, loss[loss=0.3124, simple_loss=0.358, pruned_loss=0.1334, over 16420.00 frames. ], tot_loss[loss=0.3344, simple_loss=0.3766, pruned_loss=0.1461, over 3282032.59 frames. ], batch size: 68, lr: 4.47e-02, grad_scale: 16.0 2023-04-27 13:34:36,179 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-04-27 13:35:46,283 INFO [train.py:904] (2/8) Epoch 1, batch 3800, loss[loss=0.3433, simple_loss=0.3786, pruned_loss=0.154, over 16466.00 frames. ], tot_loss[loss=0.3352, simple_loss=0.3766, pruned_loss=0.1469, over 3287775.75 frames. ], batch size: 146, lr: 4.46e-02, grad_scale: 16.0 2023-04-27 13:36:00,501 INFO [optim.py:368] (2/8) 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 13:36:57,573 INFO [train.py:904] (2/8) Epoch 1, batch 3850, loss[loss=0.3458, simple_loss=0.3843, pruned_loss=0.1536, over 15674.00 frames. ], tot_loss[loss=0.3332, simple_loss=0.3748, pruned_loss=0.1458, over 3277176.86 frames. ], batch size: 190, lr: 4.45e-02, grad_scale: 16.0 2023-04-27 13:37:08,044 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 13:37:25,142 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6696, 5.0424, 4.6623, 4.8445, 4.9482, 5.1838, 5.1264, 4.7678], device='cuda:2'), covar=tensor([0.0594, 0.0699, 0.0694, 0.0943, 0.1238, 0.0507, 0.0567, 0.1228], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0176, 0.0141, 0.0156, 0.0184, 0.0132, 0.0136, 0.0192], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 13:38:06,602 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1468, 4.8976, 5.0478, 5.0708, 4.4046, 4.9812, 4.5319, 4.6676], device='cuda:2'), covar=tensor([0.0150, 0.0105, 0.0100, 0.0092, 0.0719, 0.0120, 0.0167, 0.0142], device='cuda:2'), in_proj_covar=tensor([0.0089, 0.0069, 0.0112, 0.0086, 0.0139, 0.0092, 0.0083, 0.0090], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:2') 2023-04-27 13:38:09,800 INFO [train.py:904] (2/8) Epoch 1, batch 3900, loss[loss=0.3216, simple_loss=0.3625, pruned_loss=0.1403, over 16878.00 frames. ], tot_loss[loss=0.3299, simple_loss=0.3722, pruned_loss=0.1438, over 3277135.19 frames. ], batch size: 102, lr: 4.44e-02, grad_scale: 16.0 2023-04-27 13:38:11,790 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:38:24,734 INFO [optim.py:368] (2/8) 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 13:38:46,385 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:39:11,910 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:39:21,370 INFO [train.py:904] (2/8) Epoch 1, batch 3950, loss[loss=0.2754, simple_loss=0.3422, pruned_loss=0.1043, over 17229.00 frames. ], tot_loss[loss=0.3284, simple_loss=0.3702, pruned_loss=0.1432, over 3285644.15 frames. ], batch size: 43, lr: 4.43e-02, grad_scale: 16.0 2023-04-27 13:39:33,290 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3512, 3.3579, 2.8406, 2.1658, 2.4991, 1.7474, 3.0362, 3.5070], device='cuda:2'), covar=tensor([0.0373, 0.0283, 0.0462, 0.1767, 0.1115, 0.1594, 0.0535, 0.0257], device='cuda:2'), in_proj_covar=tensor([0.0046, 0.0042, 0.0065, 0.0109, 0.0097, 0.0098, 0.0057, 0.0036], device='cuda:2'), out_proj_covar=tensor([6.4876e-05, 5.6680e-05, 7.2778e-05, 1.1728e-04, 1.0628e-04, 1.0380e-04, 7.0770e-05, 5.0131e-05], device='cuda:2') 2023-04-27 13:39:56,418 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1253, 3.5880, 2.8858, 4.9792, 4.8329, 4.4393, 2.8898, 4.4461], device='cuda:2'), covar=tensor([0.2001, 0.0263, 0.1030, 0.0033, 0.0053, 0.0184, 0.0616, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0060, 0.0099, 0.0040, 0.0037, 0.0054, 0.0079, 0.0057], device='cuda:2'), out_proj_covar=tensor([1.5003e-04, 6.5121e-05, 1.1428e-04, 5.1555e-05, 5.2594e-05, 7.1360e-05, 8.9227e-05, 6.3258e-05], device='cuda:2') 2023-04-27 13:40:14,002 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:20,465 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:40:36,999 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 13:40:37,500 INFO [train.py:904] (2/8) Epoch 1, batch 4000, loss[loss=0.3047, simple_loss=0.3526, pruned_loss=0.1284, over 16876.00 frames. ], tot_loss[loss=0.325, simple_loss=0.3675, pruned_loss=0.1413, over 3296584.53 frames. ], batch size: 96, lr: 4.42e-02, grad_scale: 16.0 2023-04-27 13:40:52,170 INFO [optim.py:368] (2/8) 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 13:41:24,036 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:41:30,968 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 13:41:50,496 INFO [train.py:904] (2/8) Epoch 1, batch 4050, loss[loss=0.2476, simple_loss=0.3222, pruned_loss=0.08646, over 16722.00 frames. ], tot_loss[loss=0.3198, simple_loss=0.3649, pruned_loss=0.1373, over 3277412.13 frames. ], batch size: 89, lr: 4.41e-02, grad_scale: 16.0 2023-04-27 13:42:24,569 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1431, 4.1870, 3.9639, 4.0235, 4.1792, 4.5764, 4.4811, 4.0644], device='cuda:2'), covar=tensor([0.0855, 0.0984, 0.0918, 0.1223, 0.1623, 0.0633, 0.0739, 0.1601], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0181, 0.0147, 0.0156, 0.0187, 0.0138, 0.0142, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 13:43:04,570 INFO [train.py:904] (2/8) Epoch 1, batch 4100, loss[loss=0.3301, simple_loss=0.3808, pruned_loss=0.1397, over 16691.00 frames. ], tot_loss[loss=0.3147, simple_loss=0.3629, pruned_loss=0.1332, over 3268981.75 frames. ], batch size: 134, lr: 4.40e-02, grad_scale: 32.0 2023-04-27 13:43:15,971 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6098, 2.3692, 2.6947, 2.9558, 3.0431, 2.6210, 3.1616, 2.9007], device='cuda:2'), covar=tensor([0.0133, 0.0537, 0.0274, 0.0138, 0.0166, 0.0244, 0.0123, 0.0181], device='cuda:2'), in_proj_covar=tensor([0.0028, 0.0047, 0.0037, 0.0035, 0.0031, 0.0034, 0.0035, 0.0033], device='cuda:2'), out_proj_covar=tensor([4.0767e-05, 7.3641e-05, 5.6378e-05, 4.3582e-05, 4.1953e-05, 4.7611e-05, 4.4500e-05, 4.3149e-05], device='cuda:2') 2023-04-27 13:43:18,962 INFO [optim.py:368] (2/8) 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 13:44:05,000 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:44:20,299 INFO [train.py:904] (2/8) Epoch 1, batch 4150, loss[loss=0.4366, simple_loss=0.4482, pruned_loss=0.2125, over 11475.00 frames. ], tot_loss[loss=0.3273, simple_loss=0.3738, pruned_loss=0.1404, over 3219945.80 frames. ], batch size: 247, lr: 4.39e-02, grad_scale: 32.0 2023-04-27 13:44:44,804 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4092, 2.0746, 2.0171, 3.0343, 3.3035, 2.8464, 2.5903, 2.7138], device='cuda:2'), covar=tensor([0.0039, 0.0402, 0.0201, 0.0164, 0.0057, 0.0074, 0.0294, 0.0190], device='cuda:2'), in_proj_covar=tensor([0.0022, 0.0036, 0.0027, 0.0027, 0.0022, 0.0026, 0.0026, 0.0029], device='cuda:2'), out_proj_covar=tensor([2.3623e-05, 3.8744e-05, 2.7782e-05, 2.7434e-05, 1.9061e-05, 2.2494e-05, 2.4232e-05, 2.6582e-05], device='cuda:2') 2023-04-27 13:44:56,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4707, 2.1517, 2.1852, 2.6804, 2.8398, 2.3469, 2.7006, 2.6316], device='cuda:2'), covar=tensor([0.0100, 0.0605, 0.0349, 0.0125, 0.0075, 0.0276, 0.0135, 0.0168], device='cuda:2'), in_proj_covar=tensor([0.0029, 0.0050, 0.0039, 0.0035, 0.0031, 0.0035, 0.0036, 0.0033], device='cuda:2'), out_proj_covar=tensor([4.0246e-05, 7.9982e-05, 6.0153e-05, 4.4327e-05, 4.3067e-05, 4.9112e-05, 4.6638e-05, 4.3282e-05], device='cuda:2') 2023-04-27 13:45:37,094 INFO [train.py:904] (2/8) Epoch 1, batch 4200, loss[loss=0.3623, simple_loss=0.4146, pruned_loss=0.155, over 17048.00 frames. ], tot_loss[loss=0.334, simple_loss=0.3818, pruned_loss=0.1431, over 3192795.29 frames. ], batch size: 50, lr: 4.38e-02, grad_scale: 16.0 2023-04-27 13:45:39,655 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:45:39,695 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:45:52,579 INFO [optim.py:368] (2/8) 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 13:46:49,599 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:46:50,856 INFO [train.py:904] (2/8) Epoch 1, batch 4250, loss[loss=0.3302, simple_loss=0.3838, pruned_loss=0.1383, over 17166.00 frames. ], tot_loss[loss=0.3345, simple_loss=0.3836, pruned_loss=0.1427, over 3157450.19 frames. ], batch size: 46, lr: 4.36e-02, grad_scale: 16.0 2023-04-27 13:47:08,127 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7780, 4.0854, 3.7297, 3.0962, 3.7522, 3.8320, 3.8883, 3.6772], device='cuda:2'), covar=tensor([0.0472, 0.0067, 0.0092, 0.0166, 0.0083, 0.0102, 0.0113, 0.0136], device='cuda:2'), in_proj_covar=tensor([0.0051, 0.0027, 0.0028, 0.0038, 0.0029, 0.0030, 0.0035, 0.0034], device='cuda:2'), out_proj_covar=tensor([7.9109e-05, 4.4007e-05, 4.4085e-05, 5.7272e-05, 4.6944e-05, 5.1606e-05, 5.4102e-05, 5.3038e-05], device='cuda:2') 2023-04-27 13:47:36,655 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:45,250 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:48:04,726 INFO [train.py:904] (2/8) Epoch 1, batch 4300, loss[loss=0.3523, simple_loss=0.4124, pruned_loss=0.1461, over 16762.00 frames. ], tot_loss[loss=0.3323, simple_loss=0.3841, pruned_loss=0.1402, over 3171054.04 frames. ], batch size: 89, lr: 4.35e-02, grad_scale: 16.0 2023-04-27 13:48:21,408 INFO [optim.py:368] (2/8) 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 13:48:52,507 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:49:17,726 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:20,274 INFO [train.py:904] (2/8) Epoch 1, batch 4350, loss[loss=0.3384, simple_loss=0.3922, pruned_loss=0.1424, over 17192.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.3879, pruned_loss=0.1416, over 3180957.27 frames. ], batch size: 46, lr: 4.34e-02, grad_scale: 16.0 2023-04-27 13:50:04,950 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:50:36,405 INFO [train.py:904] (2/8) Epoch 1, batch 4400, loss[loss=0.3496, simple_loss=0.4057, pruned_loss=0.1467, over 16753.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3887, pruned_loss=0.141, over 3198343.46 frames. ], batch size: 124, lr: 4.33e-02, grad_scale: 16.0 2023-04-27 13:50:51,960 INFO [optim.py:368] (2/8) 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 13:51:48,408 INFO [train.py:904] (2/8) Epoch 1, batch 4450, loss[loss=0.3415, simple_loss=0.3972, pruned_loss=0.1429, over 16730.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.3907, pruned_loss=0.1402, over 3202042.90 frames. ], batch size: 124, lr: 4.32e-02, grad_scale: 16.0 2023-04-27 13:52:57,209 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:53:02,593 INFO [train.py:904] (2/8) Epoch 1, batch 4500, loss[loss=0.3096, simple_loss=0.376, pruned_loss=0.1216, over 16475.00 frames. ], tot_loss[loss=0.3329, simple_loss=0.3886, pruned_loss=0.1386, over 3171965.29 frames. ], batch size: 68, lr: 4.31e-02, grad_scale: 8.0 2023-04-27 13:53:20,026 INFO [optim.py:368] (2/8) 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 13:54:14,091 INFO [train.py:904] (2/8) Epoch 1, batch 4550, loss[loss=0.3585, simple_loss=0.4065, pruned_loss=0.1553, over 15258.00 frames. ], tot_loss[loss=0.3309, simple_loss=0.3874, pruned_loss=0.1372, over 3179029.87 frames. ], batch size: 191, lr: 4.30e-02, grad_scale: 8.0 2023-04-27 13:54:58,127 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:55:14,500 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8005, 5.7140, 5.5974, 5.5543, 5.5415, 5.9211, 5.5682, 5.3812], device='cuda:2'), covar=tensor([0.0479, 0.0652, 0.0435, 0.0842, 0.1253, 0.0386, 0.0455, 0.1056], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0180, 0.0140, 0.0151, 0.0187, 0.0131, 0.0136, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 13:55:25,714 INFO [train.py:904] (2/8) Epoch 1, batch 4600, loss[loss=0.3216, simple_loss=0.3865, pruned_loss=0.1283, over 17249.00 frames. ], tot_loss[loss=0.3302, simple_loss=0.3877, pruned_loss=0.1363, over 3179662.48 frames. ], batch size: 45, lr: 4.29e-02, grad_scale: 8.0 2023-04-27 13:55:43,313 INFO [optim.py:368] (2/8) 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 13:56:07,429 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:27,638 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:36,566 INFO [train.py:904] (2/8) Epoch 1, batch 4650, loss[loss=0.3123, simple_loss=0.3732, pruned_loss=0.1257, over 16414.00 frames. ], tot_loss[loss=0.3251, simple_loss=0.3833, pruned_loss=0.1335, over 3181614.39 frames. ], batch size: 146, lr: 4.28e-02, grad_scale: 8.0 2023-04-27 13:57:50,180 INFO [train.py:904] (2/8) Epoch 1, batch 4700, loss[loss=0.3462, simple_loss=0.3774, pruned_loss=0.1576, over 11689.00 frames. ], tot_loss[loss=0.3232, simple_loss=0.3809, pruned_loss=0.1327, over 3170556.99 frames. ], batch size: 247, lr: 4.27e-02, grad_scale: 8.0 2023-04-27 13:58:07,984 INFO [optim.py:368] (2/8) 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 13:58:30,596 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6769, 2.5878, 2.5758, 1.7271, 2.5545, 2.6378, 2.5930, 2.6043], device='cuda:2'), covar=tensor([0.0152, 0.0247, 0.0211, 0.1500, 0.0207, 0.0197, 0.0145, 0.0226], device='cuda:2'), in_proj_covar=tensor([0.0056, 0.0059, 0.0057, 0.0125, 0.0054, 0.0055, 0.0053, 0.0068], device='cuda:2'), out_proj_covar=tensor([7.2750e-05, 7.8210e-05, 8.2009e-05, 1.6811e-04, 8.0800e-05, 7.5029e-05, 8.0726e-05, 8.8605e-05], device='cuda:2') 2023-04-27 13:59:02,499 INFO [train.py:904] (2/8) Epoch 1, batch 4750, loss[loss=0.2745, simple_loss=0.3346, pruned_loss=0.1072, over 16636.00 frames. ], tot_loss[loss=0.3185, simple_loss=0.3765, pruned_loss=0.1303, over 3186966.02 frames. ], batch size: 57, lr: 4.26e-02, grad_scale: 8.0 2023-04-27 14:00:11,413 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:00:16,558 INFO [train.py:904] (2/8) Epoch 1, batch 4800, loss[loss=0.2823, simple_loss=0.3477, pruned_loss=0.1085, over 16593.00 frames. ], tot_loss[loss=0.314, simple_loss=0.3726, pruned_loss=0.1277, over 3195580.90 frames. ], batch size: 57, lr: 4.25e-02, grad_scale: 8.0 2023-04-27 14:00:34,036 INFO [optim.py:368] (2/8) 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 14:01:01,616 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4526, 3.0796, 2.6040, 2.8872, 2.4311, 1.9523, 3.0407, 3.3026], device='cuda:2'), covar=tensor([0.1080, 0.0395, 0.0809, 0.0175, 0.1211, 0.1275, 0.0189, 0.0082], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0100, 0.0138, 0.0081, 0.0120, 0.0117, 0.0078, 0.0044], device='cuda:2'), out_proj_covar=tensor([1.7672e-04, 1.2095e-04, 1.4282e-04, 8.7163e-05, 1.4742e-04, 1.2625e-04, 8.8284e-05, 5.3121e-05], device='cuda:2') 2023-04-27 14:01:18,231 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9820, 2.4927, 2.1123, 3.0699, 3.1867, 3.2796, 2.1734, 2.9073], device='cuda:2'), covar=tensor([0.1891, 0.0225, 0.1199, 0.0107, 0.0117, 0.0170, 0.0513, 0.0178], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0071, 0.0126, 0.0042, 0.0044, 0.0061, 0.0093, 0.0068], device='cuda:2'), out_proj_covar=tensor([1.6927e-04, 8.7767e-05, 1.4846e-04, 5.8913e-05, 6.4333e-05, 9.2488e-05, 1.1552e-04, 8.6828e-05], device='cuda:2') 2023-04-27 14:01:22,952 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:31,748 INFO [train.py:904] (2/8) Epoch 1, batch 4850, loss[loss=0.2905, simple_loss=0.3648, pruned_loss=0.1081, over 16907.00 frames. ], tot_loss[loss=0.3141, simple_loss=0.3734, pruned_loss=0.1274, over 3185041.30 frames. ], batch size: 109, lr: 4.24e-02, grad_scale: 8.0 2023-04-27 14:02:17,679 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 14:02:49,104 INFO [train.py:904] (2/8) Epoch 1, batch 4900, loss[loss=0.2754, simple_loss=0.3535, pruned_loss=0.09865, over 16900.00 frames. ], tot_loss[loss=0.3113, simple_loss=0.3723, pruned_loss=0.1251, over 3188837.15 frames. ], batch size: 96, lr: 4.23e-02, grad_scale: 8.0 2023-04-27 14:03:07,683 INFO [optim.py:368] (2/8) 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 14:03:17,313 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0463, 3.7732, 3.9362, 4.1294, 3.4216, 3.9913, 3.8334, 3.7201], device='cuda:2'), covar=tensor([0.0273, 0.0150, 0.0200, 0.0101, 0.0760, 0.0219, 0.0333, 0.0193], device='cuda:2'), in_proj_covar=tensor([0.0081, 0.0060, 0.0107, 0.0081, 0.0131, 0.0086, 0.0074, 0.0085], device='cuda:2'), out_proj_covar=tensor([1.4093e-04, 9.8864e-05, 1.8388e-04, 1.2910e-04, 1.8735e-04, 1.5607e-04, 1.2766e-04, 1.4930e-04], device='cuda:2') 2023-04-27 14:03:55,022 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:04,716 INFO [train.py:904] (2/8) Epoch 1, batch 4950, loss[loss=0.3638, simple_loss=0.4053, pruned_loss=0.1612, over 12377.00 frames. ], tot_loss[loss=0.311, simple_loss=0.3723, pruned_loss=0.1248, over 3201135.91 frames. ], batch size: 246, lr: 4.21e-02, grad_scale: 8.0 2023-04-27 14:04:10,097 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 14:04:15,253 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6568, 5.7182, 5.4339, 5.4725, 5.5966, 6.0031, 5.7776, 5.4816], device='cuda:2'), covar=tensor([0.0509, 0.0662, 0.0552, 0.0981, 0.1420, 0.0461, 0.0467, 0.1093], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0185, 0.0147, 0.0156, 0.0192, 0.0139, 0.0141, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 14:04:26,161 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1513, 4.5618, 4.3808, 1.7331, 4.1256, 4.4113, 3.9847, 3.9915], device='cuda:2'), covar=tensor([0.0085, 0.0117, 0.0148, 0.2223, 0.0145, 0.0133, 0.0099, 0.0232], device='cuda:2'), in_proj_covar=tensor([0.0055, 0.0061, 0.0057, 0.0129, 0.0055, 0.0056, 0.0056, 0.0071], device='cuda:2'), out_proj_covar=tensor([7.2532e-05, 8.3132e-05, 8.2847e-05, 1.7696e-04, 8.3356e-05, 7.8622e-05, 8.7323e-05, 9.3116e-05], device='cuda:2') 2023-04-27 14:05:04,813 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:05:17,360 INFO [train.py:904] (2/8) Epoch 1, batch 5000, loss[loss=0.2895, simple_loss=0.3548, pruned_loss=0.1122, over 16803.00 frames. ], tot_loss[loss=0.3116, simple_loss=0.3735, pruned_loss=0.1249, over 3216141.35 frames. ], batch size: 39, lr: 4.20e-02, grad_scale: 8.0 2023-04-27 14:05:35,392 INFO [optim.py:368] (2/8) 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 14:06:06,313 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5041, 4.5480, 4.7669, 4.8005, 5.0571, 4.4054, 4.6750, 4.7933], device='cuda:2'), covar=tensor([0.0250, 0.0197, 0.0427, 0.0383, 0.0337, 0.0228, 0.0463, 0.0165], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0098, 0.0125, 0.0123, 0.0134, 0.0107, 0.0130, 0.0097], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:2') 2023-04-27 14:06:31,099 INFO [train.py:904] (2/8) Epoch 1, batch 5050, loss[loss=0.3156, simple_loss=0.373, pruned_loss=0.1291, over 16999.00 frames. ], tot_loss[loss=0.3092, simple_loss=0.3721, pruned_loss=0.1232, over 3231403.80 frames. ], batch size: 41, lr: 4.19e-02, grad_scale: 8.0 2023-04-27 14:06:43,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6072, 3.8047, 4.0745, 2.9649, 3.7592, 4.0118, 3.7038, 3.5501], device='cuda:2'), covar=tensor([0.0737, 0.0117, 0.0071, 0.0225, 0.0093, 0.0105, 0.0246, 0.0169], device='cuda:2'), in_proj_covar=tensor([0.0069, 0.0030, 0.0032, 0.0043, 0.0032, 0.0032, 0.0037, 0.0041], device='cuda:2'), out_proj_covar=tensor([1.1529e-04, 5.4382e-05, 5.6791e-05, 7.3613e-05, 5.7654e-05, 6.0915e-05, 6.5824e-05, 7.1079e-05], device='cuda:2') 2023-04-27 14:07:42,590 INFO [train.py:904] (2/8) Epoch 1, batch 5100, loss[loss=0.3104, simple_loss=0.3601, pruned_loss=0.1303, over 16275.00 frames. ], tot_loss[loss=0.3074, simple_loss=0.3704, pruned_loss=0.1222, over 3228354.98 frames. ], batch size: 35, lr: 4.18e-02, grad_scale: 8.0 2023-04-27 14:07:59,831 INFO [optim.py:368] (2/8) 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 14:08:04,024 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8397, 3.5778, 3.3619, 3.0480, 3.6155, 3.4007, 3.3489, 3.4926], device='cuda:2'), covar=tensor([0.0100, 0.0102, 0.0134, 0.0386, 0.0083, 0.0214, 0.0117, 0.0155], device='cuda:2'), in_proj_covar=tensor([0.0048, 0.0037, 0.0055, 0.0073, 0.0043, 0.0055, 0.0055, 0.0051], device='cuda:2'), out_proj_covar=tensor([9.9777e-05, 7.3345e-05, 1.2166e-04, 1.4170e-04, 7.8382e-05, 1.0957e-04, 1.1483e-04, 1.1469e-04], device='cuda:2') 2023-04-27 14:08:31,338 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=2.02 vs. limit=2.0 2023-04-27 14:08:58,108 INFO [train.py:904] (2/8) Epoch 1, batch 5150, loss[loss=0.3621, simple_loss=0.4114, pruned_loss=0.1564, over 12333.00 frames. ], tot_loss[loss=0.3061, simple_loss=0.3706, pruned_loss=0.1208, over 3226591.13 frames. ], batch size: 246, lr: 4.17e-02, grad_scale: 8.0 2023-04-27 14:09:02,776 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8190, 3.8162, 3.2780, 3.3493, 2.9732, 2.1651, 3.7960, 4.1156], device='cuda:2'), covar=tensor([0.1454, 0.0475, 0.0746, 0.0357, 0.1192, 0.1335, 0.0235, 0.0051], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0115, 0.0149, 0.0088, 0.0134, 0.0125, 0.0085, 0.0049], device='cuda:2'), out_proj_covar=tensor([1.8829e-04, 1.3986e-04, 1.5749e-04, 9.8182e-05, 1.6334e-04, 1.3792e-04, 9.8950e-05, 5.9711e-05], device='cuda:2') 2023-04-27 14:09:22,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2181, 1.2834, 2.0429, 1.9176, 1.9439, 2.0436, 1.9683, 2.2309], device='cuda:2'), covar=tensor([0.0100, 0.0518, 0.0131, 0.0212, 0.0114, 0.0191, 0.0161, 0.0198], device='cuda:2'), in_proj_covar=tensor([0.0027, 0.0049, 0.0031, 0.0030, 0.0029, 0.0033, 0.0029, 0.0031], device='cuda:2'), out_proj_covar=tensor([2.9075e-05, 6.4651e-05, 3.5218e-05, 3.5933e-05, 3.0060e-05, 3.4002e-05, 3.2899e-05, 3.4616e-05], device='cuda:2') 2023-04-27 14:10:12,902 INFO [train.py:904] (2/8) Epoch 1, batch 5200, loss[loss=0.3096, simple_loss=0.3735, pruned_loss=0.1229, over 16392.00 frames. ], tot_loss[loss=0.3078, simple_loss=0.3711, pruned_loss=0.1222, over 3216086.72 frames. ], batch size: 146, lr: 4.16e-02, grad_scale: 8.0 2023-04-27 14:10:30,245 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 1, batch 5250, loss[loss=0.2793, simple_loss=0.352, pruned_loss=0.1033, over 16946.00 frames. ], tot_loss[loss=0.3063, simple_loss=0.3684, pruned_loss=0.1221, over 3206330.24 frames. ], batch size: 109, lr: 4.15e-02, grad_scale: 8.0 2023-04-27 14:12:16,605 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 14:12:37,169 INFO [train.py:904] (2/8) Epoch 1, batch 5300, loss[loss=0.261, simple_loss=0.3285, pruned_loss=0.09677, over 16502.00 frames. ], tot_loss[loss=0.3019, simple_loss=0.3637, pruned_loss=0.12, over 3208276.76 frames. ], batch size: 75, lr: 4.14e-02, grad_scale: 8.0 2023-04-27 14:12:50,994 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7674, 3.1944, 3.3351, 4.2378, 3.3341, 3.9435, 3.5220, 3.0528], device='cuda:2'), covar=tensor([0.0266, 0.0431, 0.0338, 0.0196, 0.0865, 0.0230, 0.0443, 0.0927], device='cuda:2'), in_proj_covar=tensor([0.0073, 0.0079, 0.0067, 0.0075, 0.0136, 0.0077, 0.0095, 0.0087], device='cuda:2'), out_proj_covar=tensor([9.0237e-05, 9.2295e-05, 7.8100e-05, 9.5018e-05, 1.5981e-04, 9.0905e-05, 1.0219e-04, 1.1061e-04], device='cuda:2') 2023-04-27 14:12:54,710 INFO [optim.py:368] (2/8) 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 14:13:22,474 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 14:13:49,538 INFO [train.py:904] (2/8) Epoch 1, batch 5350, loss[loss=0.3534, simple_loss=0.3913, pruned_loss=0.1578, over 12406.00 frames. ], tot_loss[loss=0.2983, simple_loss=0.3604, pruned_loss=0.1181, over 3189353.27 frames. ], batch size: 248, lr: 4.13e-02, grad_scale: 8.0 2023-04-27 14:15:00,987 INFO [train.py:904] (2/8) Epoch 1, batch 5400, loss[loss=0.3351, simple_loss=0.3955, pruned_loss=0.1374, over 15372.00 frames. ], tot_loss[loss=0.3026, simple_loss=0.3649, pruned_loss=0.1202, over 3194315.68 frames. ], batch size: 190, lr: 4.12e-02, grad_scale: 8.0 2023-04-27 14:15:18,319 INFO [optim.py:368] (2/8) 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 14:15:30,164 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4622, 1.1719, 1.2822, 1.4310, 1.4228, 1.4604, 1.4915, 1.6755], device='cuda:2'), covar=tensor([0.0109, 0.0361, 0.0161, 0.0160, 0.0136, 0.0154, 0.0126, 0.0119], device='cuda:2'), in_proj_covar=tensor([0.0027, 0.0050, 0.0033, 0.0030, 0.0030, 0.0035, 0.0028, 0.0030], device='cuda:2'), out_proj_covar=tensor([2.9564e-05, 6.8127e-05, 3.8046e-05, 3.7488e-05, 3.1923e-05, 3.7773e-05, 3.3334e-05, 3.4968e-05], device='cuda:2') 2023-04-27 14:16:19,546 INFO [train.py:904] (2/8) Epoch 1, batch 5450, loss[loss=0.3452, simple_loss=0.4055, pruned_loss=0.1424, over 16690.00 frames. ], tot_loss[loss=0.3093, simple_loss=0.3703, pruned_loss=0.1242, over 3192575.19 frames. ], batch size: 89, lr: 4.11e-02, grad_scale: 8.0 2023-04-27 14:16:39,199 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 14:16:48,523 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 14:17:37,167 INFO [train.py:904] (2/8) Epoch 1, batch 5500, loss[loss=0.3787, simple_loss=0.4157, pruned_loss=0.1708, over 16793.00 frames. ], tot_loss[loss=0.3243, simple_loss=0.3817, pruned_loss=0.1335, over 3190862.77 frames. ], batch size: 39, lr: 4.10e-02, grad_scale: 8.0 2023-04-27 14:17:43,324 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4717, 3.2335, 3.0964, 2.3578, 3.1645, 3.1868, 3.3024, 2.8361], device='cuda:2'), covar=tensor([0.1134, 0.0098, 0.0105, 0.0267, 0.0096, 0.0106, 0.0109, 0.0189], device='cuda:2'), in_proj_covar=tensor([0.0085, 0.0035, 0.0038, 0.0051, 0.0036, 0.0036, 0.0042, 0.0047], device='cuda:2'), out_proj_covar=tensor([1.4917e-04, 6.6137e-05, 7.1209e-05, 8.8780e-05, 6.6604e-05, 7.2683e-05, 7.9484e-05, 8.6061e-05], device='cuda:2') 2023-04-27 14:17:56,197 INFO [optim.py:368] (2/8) 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 14:18:08,197 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0396, 3.8910, 3.5513, 3.2769, 3.8664, 3.6488, 3.6547, 3.8954], device='cuda:2'), covar=tensor([0.0216, 0.0114, 0.0151, 0.0448, 0.0100, 0.0260, 0.0118, 0.0149], device='cuda:2'), in_proj_covar=tensor([0.0050, 0.0036, 0.0055, 0.0073, 0.0041, 0.0058, 0.0053, 0.0051], device='cuda:2'), out_proj_covar=tensor([1.1079e-04, 7.6744e-05, 1.2589e-04, 1.4923e-04, 8.2438e-05, 1.2246e-04, 1.1860e-04, 1.2423e-04], device='cuda:2') 2023-04-27 14:18:18,710 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9763, 3.9616, 3.9057, 4.3488, 4.1878, 4.1639, 4.2822, 4.0666], device='cuda:2'), covar=tensor([0.0334, 0.0337, 0.1088, 0.0290, 0.0478, 0.0338, 0.0299, 0.0313], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0143, 0.0235, 0.0160, 0.0140, 0.0142, 0.0125, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 14:18:35,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5015, 3.4921, 3.3156, 1.7408, 2.7798, 1.7973, 3.1848, 3.6080], device='cuda:2'), covar=tensor([0.0277, 0.0270, 0.0299, 0.2144, 0.0924, 0.1428, 0.0682, 0.0163], device='cuda:2'), in_proj_covar=tensor([0.0082, 0.0058, 0.0093, 0.0139, 0.0129, 0.0126, 0.0113, 0.0052], device='cuda:2'), out_proj_covar=tensor([1.3520e-04, 1.0035e-04, 1.3084e-04, 1.7959e-04, 1.7938e-04, 1.6763e-04, 1.7366e-04, 9.2128e-05], device='cuda:2') 2023-04-27 14:18:49,591 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 14:18:57,387 INFO [train.py:904] (2/8) Epoch 1, batch 5550, loss[loss=0.4114, simple_loss=0.4459, pruned_loss=0.1885, over 16863.00 frames. ], tot_loss[loss=0.3404, simple_loss=0.393, pruned_loss=0.144, over 3171720.70 frames. ], batch size: 42, lr: 4.09e-02, grad_scale: 8.0 2023-04-27 14:20:17,844 INFO [train.py:904] (2/8) Epoch 1, batch 5600, loss[loss=0.5098, simple_loss=0.4925, pruned_loss=0.2636, over 11155.00 frames. ], tot_loss[loss=0.35, simple_loss=0.3994, pruned_loss=0.1503, over 3142308.60 frames. ], batch size: 248, lr: 4.08e-02, grad_scale: 8.0 2023-04-27 14:20:37,625 INFO [optim.py:368] (2/8) 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 14:21:33,228 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9447, 3.8745, 3.2996, 3.7440, 2.9721, 2.1818, 4.1819, 4.5075], device='cuda:2'), covar=tensor([0.1740, 0.0529, 0.0978, 0.0306, 0.1714, 0.1451, 0.0205, 0.0039], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0123, 0.0161, 0.0094, 0.0156, 0.0131, 0.0095, 0.0052], device='cuda:2'), out_proj_covar=tensor([2.0465e-04, 1.5089e-04, 1.7078e-04, 1.0700e-04, 1.8745e-04, 1.4840e-04, 1.1240e-04, 6.3510e-05], device='cuda:2') 2023-04-27 14:21:39,870 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:21:40,597 INFO [train.py:904] (2/8) Epoch 1, batch 5650, loss[loss=0.3868, simple_loss=0.4378, pruned_loss=0.1679, over 16547.00 frames. ], tot_loss[loss=0.3597, simple_loss=0.4064, pruned_loss=0.1565, over 3138635.11 frames. ], batch size: 75, lr: 4.07e-02, grad_scale: 8.0 2023-04-27 14:22:57,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6507, 2.1488, 1.9347, 2.9036, 2.9810, 3.2138, 2.7044, 3.1466], device='cuda:2'), covar=tensor([0.0107, 0.0657, 0.0423, 0.0173, 0.0109, 0.0113, 0.0162, 0.0116], device='cuda:2'), in_proj_covar=tensor([0.0034, 0.0078, 0.0056, 0.0043, 0.0037, 0.0037, 0.0043, 0.0037], device='cuda:2'), out_proj_covar=tensor([5.4648e-05, 1.3762e-04, 9.9561e-05, 6.7500e-05, 5.9745e-05, 5.8962e-05, 6.3720e-05, 6.0384e-05], device='cuda:2') 2023-04-27 14:22:59,470 INFO [train.py:904] (2/8) Epoch 1, batch 5700, loss[loss=0.4806, simple_loss=0.476, pruned_loss=0.2426, over 11758.00 frames. ], tot_loss[loss=0.3655, simple_loss=0.4096, pruned_loss=0.1607, over 3102553.58 frames. ], batch size: 248, lr: 4.06e-02, grad_scale: 8.0 2023-04-27 14:23:15,448 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:23:17,958 INFO [optim.py:368] (2/8) 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 14:23:29,084 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7445, 3.6162, 4.0387, 4.0455, 4.1893, 3.7149, 3.8221, 3.9110], device='cuda:2'), covar=tensor([0.0251, 0.0268, 0.0429, 0.0405, 0.0343, 0.0289, 0.0500, 0.0259], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0103, 0.0131, 0.0127, 0.0140, 0.0110, 0.0141, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-27 14:23:37,942 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6879, 2.5067, 2.4358, 2.1557, 2.5372, 2.5015, 2.5936, 2.2546], device='cuda:2'), covar=tensor([0.0894, 0.0134, 0.0133, 0.0289, 0.0114, 0.0130, 0.0127, 0.0263], device='cuda:2'), in_proj_covar=tensor([0.0087, 0.0035, 0.0038, 0.0055, 0.0035, 0.0035, 0.0041, 0.0049], device='cuda:2'), out_proj_covar=tensor([1.5566e-04, 6.9021e-05, 7.2501e-05, 9.8893e-05, 6.6097e-05, 7.3090e-05, 7.7741e-05, 9.1313e-05], device='cuda:2') 2023-04-27 14:24:21,182 INFO [train.py:904] (2/8) Epoch 1, batch 5750, loss[loss=0.4025, simple_loss=0.44, pruned_loss=0.1825, over 15464.00 frames. ], tot_loss[loss=0.3692, simple_loss=0.4125, pruned_loss=0.163, over 3060685.31 frames. ], batch size: 191, lr: 4.05e-02, grad_scale: 8.0 2023-04-27 14:24:25,647 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 14:24:28,295 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:42,931 INFO [train.py:904] (2/8) Epoch 1, batch 5800, loss[loss=0.3029, simple_loss=0.37, pruned_loss=0.1179, over 16865.00 frames. ], tot_loss[loss=0.3689, simple_loss=0.4126, pruned_loss=0.1626, over 3033814.17 frames. ], batch size: 96, lr: 4.04e-02, grad_scale: 8.0 2023-04-27 14:26:01,838 INFO [optim.py:368] (2/8) 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 14:26:06,950 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:38,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4135, 3.3791, 3.7710, 3.8365, 3.8794, 3.5539, 3.5674, 3.8144], device='cuda:2'), covar=tensor([0.0325, 0.0330, 0.0488, 0.0389, 0.0427, 0.0348, 0.0561, 0.0276], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0103, 0.0131, 0.0127, 0.0140, 0.0111, 0.0143, 0.0103], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-27 14:26:55,588 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9174, 3.6340, 3.2990, 2.8657, 3.3896, 3.3547, 3.7125, 2.8799], device='cuda:2'), covar=tensor([0.1145, 0.0099, 0.0156, 0.0309, 0.0155, 0.0174, 0.0135, 0.0305], device='cuda:2'), in_proj_covar=tensor([0.0087, 0.0035, 0.0037, 0.0054, 0.0034, 0.0035, 0.0040, 0.0051], device='cuda:2'), out_proj_covar=tensor([1.5657e-04, 6.9848e-05, 7.3569e-05, 9.9661e-05, 6.4595e-05, 7.5190e-05, 7.6416e-05, 9.5785e-05], device='cuda:2') 2023-04-27 14:27:02,341 INFO [train.py:904] (2/8) Epoch 1, batch 5850, loss[loss=0.3753, simple_loss=0.4356, pruned_loss=0.1575, over 16806.00 frames. ], tot_loss[loss=0.3646, simple_loss=0.41, pruned_loss=0.1596, over 3052875.25 frames. ], batch size: 102, lr: 4.03e-02, grad_scale: 8.0 2023-04-27 14:27:14,360 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:25,813 INFO [train.py:904] (2/8) Epoch 1, batch 5900, loss[loss=0.3223, simple_loss=0.3912, pruned_loss=0.1267, over 16684.00 frames. ], tot_loss[loss=0.3611, simple_loss=0.4078, pruned_loss=0.1572, over 3076214.52 frames. ], batch size: 89, lr: 4.02e-02, grad_scale: 8.0 2023-04-27 14:28:48,065 INFO [optim.py:368] (2/8) 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 14:28:58,986 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:29:49,226 INFO [train.py:904] (2/8) Epoch 1, batch 5950, loss[loss=0.3576, simple_loss=0.4082, pruned_loss=0.1534, over 16704.00 frames. ], tot_loss[loss=0.3586, simple_loss=0.4079, pruned_loss=0.1546, over 3081658.09 frames. ], batch size: 134, lr: 4.01e-02, grad_scale: 8.0 2023-04-27 14:30:52,685 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:31:14,068 INFO [train.py:904] (2/8) Epoch 1, batch 6000, loss[loss=0.2988, simple_loss=0.357, pruned_loss=0.1203, over 16743.00 frames. ], tot_loss[loss=0.3587, simple_loss=0.4076, pruned_loss=0.1549, over 3093330.70 frames. ], batch size: 83, lr: 4.00e-02, grad_scale: 8.0 2023-04-27 14:31:14,069 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 14:31:23,942 INFO [train.py:938] (2/8) Epoch 1, validation: loss=0.2762, simple_loss=0.3752, pruned_loss=0.08863, over 944034.00 frames. 2023-04-27 14:31:23,943 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17361MB 2023-04-27 14:31:31,568 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:31:41,452 INFO [optim.py:368] (2/8) 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 14:32:29,986 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:42,910 INFO [train.py:904] (2/8) Epoch 1, batch 6050, loss[loss=0.4156, simple_loss=0.4256, pruned_loss=0.2029, over 11315.00 frames. ], tot_loss[loss=0.3583, simple_loss=0.4065, pruned_loss=0.1551, over 3076605.33 frames. ], batch size: 246, lr: 3.99e-02, grad_scale: 8.0 2023-04-27 14:32:44,240 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:46,507 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-27 14:33:04,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4473, 5.0853, 4.9430, 5.0707, 5.1197, 5.5403, 5.3355, 4.9775], device='cuda:2'), covar=tensor([0.0725, 0.0993, 0.0745, 0.1143, 0.1725, 0.0647, 0.0687, 0.1531], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0187, 0.0151, 0.0161, 0.0207, 0.0149, 0.0150, 0.0216], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-27 14:33:49,293 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0719, 4.2930, 4.0899, 3.3808, 4.1706, 4.3045, 4.4714, 3.0561], device='cuda:2'), covar=tensor([0.1069, 0.0086, 0.0103, 0.0249, 0.0090, 0.0085, 0.0089, 0.0314], device='cuda:2'), in_proj_covar=tensor([0.0091, 0.0036, 0.0036, 0.0058, 0.0036, 0.0035, 0.0041, 0.0054], device='cuda:2'), out_proj_covar=tensor([1.6743e-04, 7.1735e-05, 7.3472e-05, 1.0946e-04, 7.1857e-05, 7.8629e-05, 7.9800e-05, 1.0536e-04], device='cuda:2') 2023-04-27 14:34:03,525 INFO [train.py:904] (2/8) Epoch 1, batch 6100, loss[loss=0.3085, simple_loss=0.3696, pruned_loss=0.1237, over 16728.00 frames. ], tot_loss[loss=0.3532, simple_loss=0.4039, pruned_loss=0.1512, over 3090980.09 frames. ], batch size: 83, lr: 3.98e-02, grad_scale: 8.0 2023-04-27 14:34:08,755 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:21,376 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:24,018 INFO [optim.py:368] (2/8) 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 14:34:59,830 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:11,059 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4216, 3.7356, 3.5137, 1.4989, 3.6722, 3.7684, 3.4489, 3.6388], device='cuda:2'), covar=tensor([0.0151, 0.0105, 0.0152, 0.2073, 0.0123, 0.0063, 0.0147, 0.0146], device='cuda:2'), in_proj_covar=tensor([0.0067, 0.0063, 0.0060, 0.0138, 0.0059, 0.0053, 0.0062, 0.0076], device='cuda:2'), out_proj_covar=tensor([9.8113e-05, 9.1504e-05, 9.3515e-05, 2.0227e-04, 9.4116e-05, 8.3485e-05, 1.0491e-04, 1.1125e-04], device='cuda:2') 2023-04-27 14:35:18,416 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5068, 2.9435, 2.8733, 3.8012, 2.8927, 3.6427, 2.7318, 2.8156], device='cuda:2'), covar=tensor([0.0279, 0.0403, 0.0334, 0.0214, 0.1060, 0.0183, 0.0624, 0.0874], device='cuda:2'), in_proj_covar=tensor([0.0087, 0.0092, 0.0077, 0.0093, 0.0163, 0.0088, 0.0112, 0.0107], device='cuda:2'), out_proj_covar=tensor([1.1327e-04, 1.1295e-04, 9.6220e-05, 1.2325e-04, 2.0699e-04, 1.0963e-04, 1.2609e-04, 1.4300e-04], device='cuda:2') 2023-04-27 14:35:26,096 INFO [train.py:904] (2/8) Epoch 1, batch 6150, loss[loss=0.3872, simple_loss=0.4204, pruned_loss=0.177, over 11617.00 frames. ], tot_loss[loss=0.3521, simple_loss=0.4025, pruned_loss=0.1509, over 3073094.85 frames. ], batch size: 248, lr: 3.97e-02, grad_scale: 8.0 2023-04-27 14:35:45,901 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7541, 1.4761, 1.4550, 1.5283, 1.9647, 1.6532, 2.0133, 2.0301], device='cuda:2'), covar=tensor([0.0097, 0.0458, 0.0237, 0.0190, 0.0124, 0.0239, 0.0142, 0.0118], device='cuda:2'), in_proj_covar=tensor([0.0031, 0.0057, 0.0037, 0.0034, 0.0036, 0.0038, 0.0030, 0.0031], device='cuda:2'), out_proj_covar=tensor([3.5539e-05, 8.0076e-05, 4.6248e-05, 4.3958e-05, 4.3139e-05, 4.5980e-05, 4.0098e-05, 3.8806e-05], device='cuda:2') 2023-04-27 14:35:57,109 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:39,050 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:46,262 INFO [train.py:904] (2/8) Epoch 1, batch 6200, loss[loss=0.3923, simple_loss=0.4279, pruned_loss=0.1784, over 15315.00 frames. ], tot_loss[loss=0.3493, simple_loss=0.3995, pruned_loss=0.1496, over 3088183.02 frames. ], batch size: 190, lr: 3.96e-02, grad_scale: 8.0 2023-04-27 14:36:49,284 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:03,247 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:05,931 INFO [optim.py:368] (2/8) 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 14:37:08,084 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:33,293 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:02,369 INFO [train.py:904] (2/8) Epoch 1, batch 6250, loss[loss=0.3272, simple_loss=0.3943, pruned_loss=0.13, over 17219.00 frames. ], tot_loss[loss=0.3461, simple_loss=0.3973, pruned_loss=0.1474, over 3105352.83 frames. ], batch size: 45, lr: 3.95e-02, grad_scale: 8.0 2023-04-27 14:38:20,958 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:24,187 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:32,057 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:39:17,017 INFO [train.py:904] (2/8) Epoch 1, batch 6300, loss[loss=0.2842, simple_loss=0.3593, pruned_loss=0.1046, over 16906.00 frames. ], tot_loss[loss=0.344, simple_loss=0.3965, pruned_loss=0.1458, over 3119766.25 frames. ], batch size: 96, lr: 3.94e-02, grad_scale: 8.0 2023-04-27 14:39:25,745 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:39:36,720 INFO [optim.py:368] (2/8) 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 14:39:40,658 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.27 vs. limit=5.0 2023-04-27 14:39:43,507 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1402, 1.6268, 1.7292, 2.3352, 2.5028, 2.2896, 2.0838, 2.3478], device='cuda:2'), covar=tensor([0.0101, 0.0706, 0.0327, 0.0157, 0.0102, 0.0180, 0.0202, 0.0131], device='cuda:2'), in_proj_covar=tensor([0.0036, 0.0081, 0.0060, 0.0044, 0.0038, 0.0038, 0.0047, 0.0037], device='cuda:2'), out_proj_covar=tensor([6.2378e-05, 1.4406e-04, 1.1020e-04, 7.5499e-05, 6.5088e-05, 6.4948e-05, 7.4474e-05, 6.4922e-05], device='cuda:2') 2023-04-27 14:39:59,077 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:28,647 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:36,333 INFO [train.py:904] (2/8) Epoch 1, batch 6350, loss[loss=0.3415, simple_loss=0.3935, pruned_loss=0.1447, over 16388.00 frames. ], tot_loss[loss=0.3492, simple_loss=0.399, pruned_loss=0.1497, over 3099512.26 frames. ], batch size: 146, lr: 3.93e-02, grad_scale: 8.0 2023-04-27 14:40:40,271 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:40:47,927 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2054, 3.1601, 1.3156, 3.2354, 2.1525, 3.2003, 1.7615, 2.4925], device='cuda:2'), covar=tensor([0.0096, 0.0183, 0.1849, 0.0106, 0.0812, 0.0220, 0.1490, 0.0590], device='cuda:2'), in_proj_covar=tensor([0.0056, 0.0057, 0.0120, 0.0058, 0.0100, 0.0056, 0.0129, 0.0090], device='cuda:2'), out_proj_covar=tensor([8.5123e-05, 9.4017e-05, 1.7768e-04, 8.5879e-05, 1.4680e-04, 1.0054e-04, 1.9050e-04, 1.4700e-04], device='cuda:2') 2023-04-27 14:41:36,943 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0087, 3.3796, 2.9797, 4.3159, 3.2316, 4.0896, 3.4169, 3.0301], device='cuda:2'), covar=tensor([0.0206, 0.0346, 0.0339, 0.0156, 0.0935, 0.0139, 0.0426, 0.0815], device='cuda:2'), in_proj_covar=tensor([0.0093, 0.0096, 0.0080, 0.0098, 0.0172, 0.0091, 0.0118, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 14:41:50,537 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:53,486 INFO [train.py:904] (2/8) Epoch 1, batch 6400, loss[loss=0.3121, simple_loss=0.3732, pruned_loss=0.1255, over 16432.00 frames. ], tot_loss[loss=0.35, simple_loss=0.3989, pruned_loss=0.1506, over 3101254.84 frames. ], batch size: 68, lr: 3.92e-02, grad_scale: 8.0 2023-04-27 14:41:56,947 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2798, 4.7590, 4.6501, 4.7594, 4.7267, 5.0873, 4.9028, 4.6464], device='cuda:2'), covar=tensor([0.0815, 0.0945, 0.0671, 0.1159, 0.1604, 0.0621, 0.0699, 0.1512], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0194, 0.0158, 0.0167, 0.0210, 0.0157, 0.0151, 0.0220], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-27 14:42:08,524 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:42:10,410 INFO [optim.py:368] (2/8) 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 14:42:13,058 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8913, 3.1296, 3.3634, 3.3191, 3.3922, 3.0851, 3.2200, 3.3463], device='cuda:2'), covar=tensor([0.0382, 0.0312, 0.0439, 0.0450, 0.0417, 0.0384, 0.0516, 0.0289], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0108, 0.0131, 0.0131, 0.0143, 0.0114, 0.0149, 0.0110], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 14:42:18,914 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 14:43:09,618 INFO [train.py:904] (2/8) Epoch 1, batch 6450, loss[loss=0.3015, simple_loss=0.3685, pruned_loss=0.1173, over 16931.00 frames. ], tot_loss[loss=0.3444, simple_loss=0.3952, pruned_loss=0.1468, over 3109721.63 frames. ], batch size: 116, lr: 3.91e-02, grad_scale: 8.0 2023-04-27 14:43:22,745 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:26,543 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:33,752 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:05,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0176, 4.6122, 4.7894, 4.9261, 4.1453, 4.6718, 4.5339, 4.4137], device='cuda:2'), covar=tensor([0.0216, 0.0131, 0.0164, 0.0100, 0.0758, 0.0180, 0.0152, 0.0169], device='cuda:2'), in_proj_covar=tensor([0.0088, 0.0065, 0.0119, 0.0091, 0.0141, 0.0092, 0.0081, 0.0096], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 14:44:12,249 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:26,976 INFO [train.py:904] (2/8) Epoch 1, batch 6500, loss[loss=0.369, simple_loss=0.3893, pruned_loss=0.1744, over 11244.00 frames. ], tot_loss[loss=0.3395, simple_loss=0.391, pruned_loss=0.1439, over 3117983.31 frames. ], batch size: 246, lr: 3.90e-02, grad_scale: 16.0 2023-04-27 14:44:45,180 INFO [optim.py:368] (2/8) 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 14:44:46,838 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:58,365 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4280, 2.9635, 2.8502, 2.3061, 2.4994, 2.8954, 2.8803, 1.8277], device='cuda:2'), covar=tensor([0.1411, 0.0126, 0.0141, 0.0437, 0.0205, 0.0138, 0.0177, 0.0628], device='cuda:2'), in_proj_covar=tensor([0.0101, 0.0040, 0.0040, 0.0066, 0.0038, 0.0038, 0.0044, 0.0064], device='cuda:2'), out_proj_covar=tensor([1.8952e-04, 8.0779e-05, 8.3261e-05, 1.2858e-04, 7.9379e-05, 8.5333e-05, 8.7983e-05, 1.2800e-04], device='cuda:2') 2023-04-27 14:44:59,743 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:45:04,530 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:07,620 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:42,981 INFO [train.py:904] (2/8) Epoch 1, batch 6550, loss[loss=0.2992, simple_loss=0.3807, pruned_loss=0.1088, over 16665.00 frames. ], tot_loss[loss=0.3425, simple_loss=0.3949, pruned_loss=0.145, over 3135052.14 frames. ], batch size: 76, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:45:55,593 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:00,519 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:06,442 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-27 14:46:08,768 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:59,693 INFO [train.py:904] (2/8) Epoch 1, batch 6600, loss[loss=0.4184, simple_loss=0.4415, pruned_loss=0.1977, over 11406.00 frames. ], tot_loss[loss=0.3456, simple_loss=0.3984, pruned_loss=0.1464, over 3113292.36 frames. ], batch size: 246, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:47:18,178 INFO [optim.py:368] (2/8) 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 14:47:32,783 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:47:43,329 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:10,682 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:18,079 INFO [train.py:904] (2/8) Epoch 1, batch 6650, loss[loss=0.3621, simple_loss=0.4035, pruned_loss=0.1604, over 16375.00 frames. ], tot_loss[loss=0.3491, simple_loss=0.4001, pruned_loss=0.149, over 3081045.18 frames. ], batch size: 35, lr: 3.88e-02, grad_scale: 16.0 2023-04-27 14:49:18,721 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:24,165 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:32,211 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:34,317 INFO [train.py:904] (2/8) Epoch 1, batch 6700, loss[loss=0.4041, simple_loss=0.4204, pruned_loss=0.1939, over 11410.00 frames. ], tot_loss[loss=0.3493, simple_loss=0.399, pruned_loss=0.1498, over 3079107.12 frames. ], batch size: 247, lr: 3.87e-02, grad_scale: 16.0 2023-04-27 14:49:52,616 INFO [optim.py:368] (2/8) 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 14:50:18,296 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 14:50:45,161 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:50:50,990 INFO [train.py:904] (2/8) Epoch 1, batch 6750, loss[loss=0.3148, simple_loss=0.3663, pruned_loss=0.1317, over 17239.00 frames. ], tot_loss[loss=0.3484, simple_loss=0.3979, pruned_loss=0.1494, over 3076749.09 frames. ], batch size: 44, lr: 3.86e-02, grad_scale: 16.0 2023-04-27 14:51:13,468 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7117, 1.6308, 1.9592, 3.0119, 3.1290, 2.6930, 2.0810, 2.5473], device='cuda:2'), covar=tensor([0.0068, 0.0657, 0.0338, 0.0084, 0.0071, 0.0167, 0.0223, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0037, 0.0086, 0.0065, 0.0044, 0.0040, 0.0039, 0.0052, 0.0038], device='cuda:2'), out_proj_covar=tensor([6.5748e-05, 1.5203e-04, 1.2066e-04, 7.9414e-05, 6.9769e-05, 7.0283e-05, 8.4741e-05, 6.7752e-05], device='cuda:2') 2023-04-27 14:51:50,791 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:05,969 INFO [train.py:904] (2/8) Epoch 1, batch 6800, loss[loss=0.3346, simple_loss=0.4023, pruned_loss=0.1335, over 16824.00 frames. ], tot_loss[loss=0.3467, simple_loss=0.3973, pruned_loss=0.1481, over 3095845.15 frames. ], batch size: 83, lr: 3.85e-02, grad_scale: 16.0 2023-04-27 14:52:15,792 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-04-27 14:52:24,954 INFO [optim.py:368] (2/8) 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 14:52:33,197 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:52:41,647 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:46,846 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:23,170 INFO [train.py:904] (2/8) Epoch 1, batch 6850, loss[loss=0.3503, simple_loss=0.412, pruned_loss=0.1443, over 16244.00 frames. ], tot_loss[loss=0.3481, simple_loss=0.3979, pruned_loss=0.1491, over 3077210.29 frames. ], batch size: 165, lr: 3.84e-02, grad_scale: 16.0 2023-04-27 14:53:35,323 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:46,427 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:57,789 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:03,548 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:54:37,392 INFO [train.py:904] (2/8) Epoch 1, batch 6900, loss[loss=0.3028, simple_loss=0.3653, pruned_loss=0.1201, over 16994.00 frames. ], tot_loss[loss=0.3467, simple_loss=0.3989, pruned_loss=0.1472, over 3086375.91 frames. ], batch size: 55, lr: 3.83e-02, grad_scale: 16.0 2023-04-27 14:54:47,053 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:49,774 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:55,476 INFO [optim.py:368] (2/8) 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 14:54:59,971 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:09,425 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:36,561 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:55:54,253 INFO [train.py:904] (2/8) Epoch 1, batch 6950, loss[loss=0.3313, simple_loss=0.3862, pruned_loss=0.1382, over 16945.00 frames. ], tot_loss[loss=0.3519, simple_loss=0.4024, pruned_loss=0.1507, over 3088839.76 frames. ], batch size: 109, lr: 3.82e-02, grad_scale: 16.0 2023-04-27 14:56:25,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:25,267 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:47,681 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:12,232 INFO [train.py:904] (2/8) Epoch 1, batch 7000, loss[loss=0.3729, simple_loss=0.4289, pruned_loss=0.1585, over 16689.00 frames. ], tot_loss[loss=0.3513, simple_loss=0.4028, pruned_loss=0.1499, over 3080304.01 frames. ], batch size: 134, lr: 3.81e-02, grad_scale: 16.0 2023-04-27 14:57:14,195 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 14:57:30,860 INFO [optim.py:368] (2/8) 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 14:58:27,029 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:58:31,408 INFO [train.py:904] (2/8) Epoch 1, batch 7050, loss[loss=0.3225, simple_loss=0.3887, pruned_loss=0.1282, over 16826.00 frames. ], tot_loss[loss=0.3501, simple_loss=0.4025, pruned_loss=0.1488, over 3085026.68 frames. ], batch size: 102, lr: 3.80e-02, grad_scale: 16.0 2023-04-27 14:58:42,453 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7057, 3.2790, 3.2479, 3.9980, 2.9355, 3.9127, 2.8675, 2.6875], device='cuda:2'), covar=tensor([0.0277, 0.0304, 0.0242, 0.0203, 0.1017, 0.0167, 0.0562, 0.0975], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0101, 0.0085, 0.0110, 0.0178, 0.0097, 0.0122, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 14:59:51,865 INFO [train.py:904] (2/8) Epoch 1, batch 7100, loss[loss=0.3017, simple_loss=0.3731, pruned_loss=0.1151, over 16866.00 frames. ], tot_loss[loss=0.346, simple_loss=0.3993, pruned_loss=0.1464, over 3113786.18 frames. ], batch size: 109, lr: 3.79e-02, grad_scale: 16.0 2023-04-27 15:00:05,739 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:00:11,238 INFO [optim.py:368] (2/8) 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 15:00:20,545 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:00:27,657 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:02,758 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0382, 3.8591, 4.2951, 4.3410, 4.4480, 3.9373, 4.1114, 4.2764], device='cuda:2'), covar=tensor([0.0274, 0.0307, 0.0355, 0.0329, 0.0326, 0.0272, 0.0658, 0.0243], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0118, 0.0136, 0.0133, 0.0146, 0.0120, 0.0169, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:01:11,086 INFO [train.py:904] (2/8) Epoch 1, batch 7150, loss[loss=0.4459, simple_loss=0.4442, pruned_loss=0.2238, over 11583.00 frames. ], tot_loss[loss=0.3439, simple_loss=0.3965, pruned_loss=0.1456, over 3106567.68 frames. ], batch size: 247, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:01:30,119 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:40,838 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6113, 2.6721, 2.3780, 2.1551, 2.5798, 2.4925, 2.6781, 1.8209], device='cuda:2'), covar=tensor([0.0997, 0.0107, 0.0138, 0.0337, 0.0102, 0.0123, 0.0096, 0.0559], device='cuda:2'), in_proj_covar=tensor([0.0106, 0.0041, 0.0042, 0.0070, 0.0038, 0.0039, 0.0043, 0.0071], device='cuda:2'), out_proj_covar=tensor([2.0457e-04, 8.7156e-05, 9.1147e-05, 1.4271e-04, 8.1462e-05, 8.9559e-05, 9.0097e-05, 1.4545e-04], device='cuda:2') 2023-04-27 15:01:41,724 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:02:23,529 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-27 15:02:27,356 INFO [train.py:904] (2/8) Epoch 1, batch 7200, loss[loss=0.3236, simple_loss=0.3847, pruned_loss=0.1312, over 15294.00 frames. ], tot_loss[loss=0.3426, simple_loss=0.3954, pruned_loss=0.1449, over 3073637.46 frames. ], batch size: 190, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:02:46,727 INFO [optim.py:368] (2/8) 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 15:03:02,467 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:03:19,571 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:03:46,512 INFO [train.py:904] (2/8) Epoch 1, batch 7250, loss[loss=0.3281, simple_loss=0.3729, pruned_loss=0.1417, over 16233.00 frames. ], tot_loss[loss=0.3392, simple_loss=0.3922, pruned_loss=0.1431, over 3055535.45 frames. ], batch size: 165, lr: 3.77e-02, grad_scale: 8.0 2023-04-27 15:04:05,153 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.19 vs. limit=5.0 2023-04-27 15:04:06,709 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:21,265 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:35,935 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:48,799 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9816, 4.5527, 4.3002, 1.7773, 4.5874, 4.5637, 3.6084, 3.9524], device='cuda:2'), covar=tensor([0.0369, 0.0081, 0.0124, 0.1924, 0.0098, 0.0095, 0.0199, 0.0154], device='cuda:2'), in_proj_covar=tensor([0.0084, 0.0066, 0.0064, 0.0148, 0.0066, 0.0057, 0.0070, 0.0082], device='cuda:2'), out_proj_covar=tensor([1.3290e-04, 1.0487e-04, 1.0660e-04, 2.2657e-04, 1.0985e-04, 9.6454e-05, 1.2655e-04, 1.3010e-04], device='cuda:2') 2023-04-27 15:04:58,722 INFO [train.py:904] (2/8) Epoch 1, batch 7300, loss[loss=0.3451, simple_loss=0.3968, pruned_loss=0.1467, over 17096.00 frames. ], tot_loss[loss=0.3382, simple_loss=0.3912, pruned_loss=0.1426, over 3068559.05 frames. ], batch size: 47, lr: 3.76e-02, grad_scale: 8.0 2023-04-27 15:05:19,382 INFO [optim.py:368] (2/8) 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 15:05:26,877 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3202, 3.2546, 3.1707, 2.9288, 3.2982, 2.4495, 3.0831, 3.1718], device='cuda:2'), covar=tensor([0.0065, 0.0047, 0.0075, 0.0237, 0.0055, 0.0515, 0.0069, 0.0089], device='cuda:2'), in_proj_covar=tensor([0.0047, 0.0037, 0.0051, 0.0075, 0.0040, 0.0075, 0.0051, 0.0051], device='cuda:2'), out_proj_covar=tensor([1.2820e-04, 9.7456e-05, 1.3646e-04, 1.7665e-04, 1.0001e-04, 1.7938e-04, 1.3275e-04, 1.5224e-04], device='cuda:2') 2023-04-27 15:05:49,056 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:53,636 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:15,329 INFO [train.py:904] (2/8) Epoch 1, batch 7350, loss[loss=0.3214, simple_loss=0.3762, pruned_loss=0.1333, over 16598.00 frames. ], tot_loss[loss=0.3349, simple_loss=0.3889, pruned_loss=0.1405, over 3077234.65 frames. ], batch size: 62, lr: 3.75e-02, grad_scale: 8.0 2023-04-27 15:06:45,451 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7982, 5.2957, 5.1946, 5.1952, 5.2234, 5.6920, 5.4878, 5.2130], device='cuda:2'), covar=tensor([0.0646, 0.1065, 0.0954, 0.1484, 0.1933, 0.0739, 0.0819, 0.2104], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0200, 0.0174, 0.0183, 0.0218, 0.0170, 0.0163, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:07:01,292 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8990, 3.0818, 3.3425, 3.3501, 3.4275, 3.0859, 3.1898, 3.3421], device='cuda:2'), covar=tensor([0.0347, 0.0335, 0.0378, 0.0390, 0.0346, 0.0337, 0.0603, 0.0317], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0113, 0.0129, 0.0126, 0.0139, 0.0114, 0.0163, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:07:21,563 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-27 15:07:30,952 INFO [train.py:904] (2/8) Epoch 1, batch 7400, loss[loss=0.3162, simple_loss=0.3735, pruned_loss=0.1294, over 16692.00 frames. ], tot_loss[loss=0.3379, simple_loss=0.3909, pruned_loss=0.1425, over 3050152.08 frames. ], batch size: 57, lr: 3.74e-02, grad_scale: 8.0 2023-04-27 15:07:36,101 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:07:40,545 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7407.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:07:50,855 INFO [optim.py:368] (2/8) 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 15:07:54,299 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2815, 3.1785, 3.0445, 2.7959, 3.2609, 2.3591, 3.0513, 3.0397], device='cuda:2'), covar=tensor([0.0088, 0.0067, 0.0103, 0.0297, 0.0066, 0.0580, 0.0086, 0.0111], device='cuda:2'), in_proj_covar=tensor([0.0047, 0.0038, 0.0053, 0.0076, 0.0040, 0.0077, 0.0052, 0.0053], device='cuda:2'), out_proj_covar=tensor([1.3063e-04, 9.9714e-05, 1.4492e-04, 1.8124e-04, 1.0137e-04, 1.8610e-04, 1.3767e-04, 1.5856e-04], device='cuda:2') 2023-04-27 15:08:52,556 INFO [train.py:904] (2/8) Epoch 1, batch 7450, loss[loss=0.425, simple_loss=0.4378, pruned_loss=0.2061, over 11391.00 frames. ], tot_loss[loss=0.3398, simple_loss=0.3924, pruned_loss=0.1436, over 3061114.16 frames. ], batch size: 246, lr: 3.73e-02, grad_scale: 8.0 2023-04-27 15:09:22,456 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:10:15,193 INFO [train.py:904] (2/8) Epoch 1, batch 7500, loss[loss=0.3586, simple_loss=0.4063, pruned_loss=0.1555, over 15236.00 frames. ], tot_loss[loss=0.3418, simple_loss=0.3945, pruned_loss=0.1446, over 3054182.44 frames. ], batch size: 190, lr: 3.72e-02, grad_scale: 8.0 2023-04-27 15:10:35,036 INFO [optim.py:368] (2/8) 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 15:10:43,191 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:11:06,755 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:11:31,890 INFO [train.py:904] (2/8) Epoch 1, batch 7550, loss[loss=0.3637, simple_loss=0.4098, pruned_loss=0.1588, over 16174.00 frames. ], tot_loss[loss=0.3418, simple_loss=0.3936, pruned_loss=0.145, over 3049297.55 frames. ], batch size: 165, lr: 3.72e-02, grad_scale: 4.0 2023-04-27 15:11:54,174 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:12:21,503 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:12:50,415 INFO [train.py:904] (2/8) Epoch 1, batch 7600, loss[loss=0.3212, simple_loss=0.3787, pruned_loss=0.1319, over 16905.00 frames. ], tot_loss[loss=0.3398, simple_loss=0.3917, pruned_loss=0.1439, over 3079197.52 frames. ], batch size: 109, lr: 3.71e-02, grad_scale: 8.0 2023-04-27 15:13:10,044 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:12,789 INFO [optim.py:368] (2/8) 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 15:13:29,661 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:40,334 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:14:13,719 INFO [train.py:904] (2/8) Epoch 1, batch 7650, loss[loss=0.4298, simple_loss=0.4381, pruned_loss=0.2107, over 11282.00 frames. ], tot_loss[loss=0.342, simple_loss=0.3929, pruned_loss=0.1456, over 3063503.19 frames. ], batch size: 248, lr: 3.70e-02, grad_scale: 8.0 2023-04-27 15:15:11,825 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 15:15:14,467 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:22,741 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 15:15:24,838 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7391, 3.7116, 4.1736, 4.1803, 4.2476, 3.8025, 3.8491, 4.0626], device='cuda:2'), covar=tensor([0.0334, 0.0321, 0.0371, 0.0374, 0.0365, 0.0288, 0.0602, 0.0302], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0120, 0.0140, 0.0136, 0.0152, 0.0126, 0.0176, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:15:39,928 INFO [train.py:904] (2/8) Epoch 1, batch 7700, loss[loss=0.3182, simple_loss=0.3826, pruned_loss=0.1269, over 16814.00 frames. ], tot_loss[loss=0.343, simple_loss=0.3934, pruned_loss=0.1463, over 3063467.19 frames. ], batch size: 102, lr: 3.69e-02, grad_scale: 8.0 2023-04-27 15:15:45,458 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:16:00,933 INFO [optim.py:368] (2/8) 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 15:16:04,270 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3081, 1.5352, 2.0029, 2.4291, 2.8690, 2.6241, 1.6724, 2.6243], device='cuda:2'), covar=tensor([0.0113, 0.0787, 0.0341, 0.0180, 0.0070, 0.0158, 0.0374, 0.0119], device='cuda:2'), in_proj_covar=tensor([0.0042, 0.0092, 0.0069, 0.0050, 0.0043, 0.0043, 0.0062, 0.0040], device='cuda:2'), out_proj_covar=tensor([7.6554e-05, 1.6657e-04, 1.3253e-04, 9.5344e-05, 7.5728e-05, 7.8647e-05, 1.0483e-04, 7.4212e-05], device='cuda:2') 2023-04-27 15:16:45,695 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:57,056 INFO [train.py:904] (2/8) Epoch 1, batch 7750, loss[loss=0.3253, simple_loss=0.3825, pruned_loss=0.1341, over 16594.00 frames. ], tot_loss[loss=0.3417, simple_loss=0.3928, pruned_loss=0.1453, over 3066716.33 frames. ], batch size: 35, lr: 3.68e-02, grad_scale: 8.0 2023-04-27 15:16:59,252 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:17:16,061 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:05,848 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4363, 3.3894, 3.7900, 3.8552, 3.9200, 3.4876, 3.5717, 3.7819], device='cuda:2'), covar=tensor([0.0336, 0.0404, 0.0533, 0.0519, 0.0388, 0.0416, 0.0774, 0.0343], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0116, 0.0136, 0.0134, 0.0147, 0.0123, 0.0173, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:18:14,288 INFO [train.py:904] (2/8) Epoch 1, batch 7800, loss[loss=0.316, simple_loss=0.3827, pruned_loss=0.1247, over 16387.00 frames. ], tot_loss[loss=0.344, simple_loss=0.3946, pruned_loss=0.1466, over 3073762.31 frames. ], batch size: 146, lr: 3.67e-02, grad_scale: 8.0 2023-04-27 15:18:16,721 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:16,982 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-04-27 15:18:19,395 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:36,277 INFO [optim.py:368] (2/8) 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 15:18:43,853 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:18:57,598 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 15:19:31,859 INFO [train.py:904] (2/8) Epoch 1, batch 7850, loss[loss=0.3468, simple_loss=0.4102, pruned_loss=0.1417, over 16735.00 frames. ], tot_loss[loss=0.3446, simple_loss=0.3963, pruned_loss=0.1465, over 3090146.73 frames. ], batch size: 89, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:19:51,317 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7863.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:19:56,834 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:20:49,083 INFO [train.py:904] (2/8) Epoch 1, batch 7900, loss[loss=0.3074, simple_loss=0.3702, pruned_loss=0.1223, over 16453.00 frames. ], tot_loss[loss=0.3419, simple_loss=0.3942, pruned_loss=0.1448, over 3086962.86 frames. ], batch size: 68, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:21:11,998 INFO [optim.py:368] (2/8) 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 15:21:39,234 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:21:42,160 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5609, 3.4063, 3.8907, 3.8526, 3.9466, 3.5421, 3.6387, 3.7856], device='cuda:2'), covar=tensor([0.0299, 0.0347, 0.0376, 0.0397, 0.0374, 0.0303, 0.0661, 0.0290], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0119, 0.0142, 0.0138, 0.0153, 0.0126, 0.0178, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:22:08,695 INFO [train.py:904] (2/8) Epoch 1, batch 7950, loss[loss=0.3202, simple_loss=0.3797, pruned_loss=0.1304, over 16896.00 frames. ], tot_loss[loss=0.3426, simple_loss=0.3946, pruned_loss=0.1453, over 3077571.25 frames. ], batch size: 116, lr: 3.65e-02, grad_scale: 8.0 2023-04-27 15:22:43,882 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3009, 4.2177, 4.3045, 4.7759, 4.6839, 4.3533, 4.7180, 4.6079], device='cuda:2'), covar=tensor([0.0492, 0.0395, 0.1122, 0.0258, 0.0437, 0.0384, 0.0326, 0.0312], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0173, 0.0262, 0.0181, 0.0156, 0.0161, 0.0146, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:22:51,833 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 15:22:54,878 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:56,121 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:23:30,885 INFO [train.py:904] (2/8) Epoch 1, batch 8000, loss[loss=0.3554, simple_loss=0.3974, pruned_loss=0.1567, over 15329.00 frames. ], tot_loss[loss=0.3415, simple_loss=0.3938, pruned_loss=0.1446, over 3085089.26 frames. ], batch size: 190, lr: 3.64e-02, grad_scale: 8.0 2023-04-27 15:23:47,613 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9707, 3.8794, 3.9062, 4.3534, 4.3341, 4.1288, 4.3648, 4.0951], device='cuda:2'), covar=tensor([0.0455, 0.0354, 0.0990, 0.0275, 0.0364, 0.0355, 0.0251, 0.0334], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0169, 0.0254, 0.0177, 0.0152, 0.0157, 0.0142, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:23:51,332 INFO [optim.py:368] (2/8) 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 15:24:45,916 INFO [train.py:904] (2/8) Epoch 1, batch 8050, loss[loss=0.4119, simple_loss=0.4299, pruned_loss=0.197, over 11298.00 frames. ], tot_loss[loss=0.3408, simple_loss=0.3933, pruned_loss=0.1441, over 3078697.60 frames. ], batch size: 248, lr: 3.63e-02, grad_scale: 8.0 2023-04-27 15:25:04,638 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:09,706 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7259, 3.8536, 2.1034, 4.6164, 4.7220, 4.3849, 2.6050, 3.6652], device='cuda:2'), covar=tensor([0.2721, 0.0330, 0.2178, 0.0124, 0.0104, 0.0233, 0.1141, 0.0460], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0090, 0.0157, 0.0058, 0.0061, 0.0078, 0.0130, 0.0111], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:25:57,834 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:00,725 INFO [train.py:904] (2/8) Epoch 1, batch 8100, loss[loss=0.3917, simple_loss=0.4194, pruned_loss=0.182, over 11688.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3924, pruned_loss=0.1428, over 3069603.81 frames. ], batch size: 247, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:26:15,158 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:22,779 INFO [optim.py:368] (2/8) 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 15:27:16,440 INFO [train.py:904] (2/8) Epoch 1, batch 8150, loss[loss=0.3061, simple_loss=0.3671, pruned_loss=0.1225, over 16799.00 frames. ], tot_loss[loss=0.3318, simple_loss=0.3868, pruned_loss=0.1384, over 3101709.56 frames. ], batch size: 102, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:27:27,357 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:28:33,138 INFO [train.py:904] (2/8) Epoch 1, batch 8200, loss[loss=0.3368, simple_loss=0.3937, pruned_loss=0.1399, over 15326.00 frames. ], tot_loss[loss=0.3301, simple_loss=0.384, pruned_loss=0.1381, over 3099725.27 frames. ], batch size: 190, lr: 3.61e-02, grad_scale: 4.0 2023-04-27 15:28:40,935 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9469, 2.9777, 3.3510, 3.3345, 3.4113, 2.9814, 3.1481, 3.2871], device='cuda:2'), covar=tensor([0.0413, 0.0456, 0.0473, 0.0446, 0.0455, 0.0465, 0.0751, 0.0356], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0125, 0.0145, 0.0142, 0.0157, 0.0131, 0.0188, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:28:56,620 INFO [optim.py:368] (2/8) 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 15:29:35,826 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6517, 2.5773, 2.5442, 2.1998, 2.5283, 2.3166, 2.5212, 1.7218], device='cuda:2'), covar=tensor([0.0969, 0.0130, 0.0112, 0.0328, 0.0122, 0.0182, 0.0094, 0.0728], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0042, 0.0047, 0.0077, 0.0044, 0.0046, 0.0047, 0.0086], device='cuda:2'), out_proj_covar=tensor([2.2808e-04, 9.6484e-05, 1.0891e-04, 1.6593e-04, 9.9407e-05, 1.1187e-04, 1.0331e-04, 1.8044e-04], device='cuda:2') 2023-04-27 15:29:53,197 INFO [train.py:904] (2/8) Epoch 1, batch 8250, loss[loss=0.3259, simple_loss=0.3874, pruned_loss=0.1322, over 15289.00 frames. ], tot_loss[loss=0.3285, simple_loss=0.3841, pruned_loss=0.1365, over 3076724.29 frames. ], batch size: 190, lr: 3.60e-02, grad_scale: 4.0 2023-04-27 15:30:15,869 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:41,221 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:14,606 INFO [train.py:904] (2/8) Epoch 1, batch 8300, loss[loss=0.2889, simple_loss=0.3604, pruned_loss=0.1087, over 15208.00 frames. ], tot_loss[loss=0.3186, simple_loss=0.3778, pruned_loss=0.1297, over 3079675.87 frames. ], batch size: 190, lr: 3.59e-02, grad_scale: 4.0 2023-04-27 15:31:40,147 INFO [optim.py:368] (2/8) 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 15:31:48,741 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2584, 3.2847, 1.5203, 3.2225, 2.0410, 3.3014, 1.6175, 2.5567], device='cuda:2'), covar=tensor([0.0064, 0.0100, 0.1613, 0.0064, 0.0912, 0.0191, 0.1448, 0.0657], device='cuda:2'), in_proj_covar=tensor([0.0061, 0.0070, 0.0149, 0.0065, 0.0125, 0.0075, 0.0153, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:31:56,704 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:59,949 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:20,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5289, 1.3632, 1.5780, 1.7474, 1.7792, 1.6524, 1.6040, 1.8473], device='cuda:2'), covar=tensor([0.0110, 0.0497, 0.0209, 0.0160, 0.0086, 0.0198, 0.0241, 0.0098], device='cuda:2'), in_proj_covar=tensor([0.0046, 0.0093, 0.0072, 0.0054, 0.0043, 0.0046, 0.0065, 0.0041], device='cuda:2'), out_proj_covar=tensor([8.6737e-05, 1.6869e-04, 1.3943e-04, 1.0202e-04, 7.5970e-05, 8.4593e-05, 1.1164e-04, 7.6573e-05], device='cuda:2') 2023-04-27 15:32:36,049 INFO [train.py:904] (2/8) Epoch 1, batch 8350, loss[loss=0.2717, simple_loss=0.3479, pruned_loss=0.09775, over 16534.00 frames. ], tot_loss[loss=0.3117, simple_loss=0.374, pruned_loss=0.1247, over 3074515.12 frames. ], batch size: 68, lr: 3.58e-02, grad_scale: 4.0 2023-04-27 15:33:55,168 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:57,931 INFO [train.py:904] (2/8) Epoch 1, batch 8400, loss[loss=0.2972, simple_loss=0.3613, pruned_loss=0.1166, over 15351.00 frames. ], tot_loss[loss=0.3055, simple_loss=0.3692, pruned_loss=0.1209, over 3058584.80 frames. ], batch size: 190, lr: 3.58e-02, grad_scale: 8.0 2023-04-27 15:34:04,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5779, 2.9734, 2.9024, 2.4513, 2.9268, 2.8737, 3.1059, 1.6527], device='cuda:2'), covar=tensor([0.1377, 0.0161, 0.0119, 0.0413, 0.0133, 0.0151, 0.0113, 0.0894], device='cuda:2'), in_proj_covar=tensor([0.0116, 0.0043, 0.0049, 0.0081, 0.0045, 0.0047, 0.0048, 0.0090], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 15:34:21,541 INFO [optim.py:368] (2/8) 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 15:34:41,308 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-27 15:34:45,130 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2706, 4.5177, 4.4361, 4.4538, 4.5363, 4.9332, 4.8094, 4.3237], device='cuda:2'), covar=tensor([0.0796, 0.0969, 0.0695, 0.1041, 0.1419, 0.0653, 0.0510, 0.1340], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0198, 0.0169, 0.0175, 0.0210, 0.0172, 0.0149, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-27 15:35:12,364 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:18,229 INFO [train.py:904] (2/8) Epoch 1, batch 8450, loss[loss=0.2768, simple_loss=0.3526, pruned_loss=0.1005, over 15489.00 frames. ], tot_loss[loss=0.3006, simple_loss=0.3664, pruned_loss=0.1175, over 3075258.97 frames. ], batch size: 190, lr: 3.57e-02, grad_scale: 8.0 2023-04-27 15:35:30,871 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:36:39,490 INFO [train.py:904] (2/8) Epoch 1, batch 8500, loss[loss=0.2442, simple_loss=0.3302, pruned_loss=0.07908, over 16590.00 frames. ], tot_loss[loss=0.2932, simple_loss=0.3605, pruned_loss=0.113, over 3065115.33 frames. ], batch size: 62, lr: 3.56e-02, grad_scale: 8.0 2023-04-27 15:36:40,917 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6979, 3.7797, 3.4260, 3.5902, 2.9399, 2.1209, 4.0075, 4.3711], device='cuda:2'), covar=tensor([0.1734, 0.0570, 0.0755, 0.0290, 0.1587, 0.1414, 0.0217, 0.0051], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0167, 0.0190, 0.0116, 0.0180, 0.0153, 0.0127, 0.0067], device='cuda:2'), out_proj_covar=tensor([2.4375e-04, 1.9683e-04, 2.0910e-04, 1.3650e-04, 2.1735e-04, 1.8432e-04, 1.5397e-04, 8.6740e-05], device='cuda:2') 2023-04-27 15:36:49,162 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:37:04,602 INFO [optim.py:368] (2/8) 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 15:38:03,991 INFO [train.py:904] (2/8) Epoch 1, batch 8550, loss[loss=0.2977, simple_loss=0.3518, pruned_loss=0.1219, over 12012.00 frames. ], tot_loss[loss=0.2889, simple_loss=0.3563, pruned_loss=0.1107, over 3037752.50 frames. ], batch size: 248, lr: 3.55e-02, grad_scale: 8.0 2023-04-27 15:39:44,565 INFO [train.py:904] (2/8) Epoch 1, batch 8600, loss[loss=0.2797, simple_loss=0.3403, pruned_loss=0.1095, over 12440.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.357, pruned_loss=0.1101, over 3032088.16 frames. ], batch size: 248, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:40:17,874 INFO [optim.py:368] (2/8) 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 15:40:27,930 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:40:50,833 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0575, 5.4381, 5.1583, 5.2976, 4.6894, 4.8703, 4.8560, 5.5082], device='cuda:2'), covar=tensor([0.0301, 0.0572, 0.0737, 0.0263, 0.0516, 0.0354, 0.0466, 0.0415], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0207, 0.0190, 0.0124, 0.0154, 0.0131, 0.0176, 0.0134], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-27 15:41:26,180 INFO [train.py:904] (2/8) Epoch 1, batch 8650, loss[loss=0.2408, simple_loss=0.327, pruned_loss=0.07725, over 16871.00 frames. ], tot_loss[loss=0.2846, simple_loss=0.354, pruned_loss=0.1076, over 3025882.75 frames. ], batch size: 116, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:41:33,753 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=6.86 vs. limit=5.0 2023-04-27 15:43:13,564 INFO [train.py:904] (2/8) Epoch 1, batch 8700, loss[loss=0.2723, simple_loss=0.3323, pruned_loss=0.1062, over 12056.00 frames. ], tot_loss[loss=0.2802, simple_loss=0.3501, pruned_loss=0.1051, over 3034586.95 frames. ], batch size: 247, lr: 3.53e-02, grad_scale: 8.0 2023-04-27 15:43:41,220 INFO [optim.py:368] (2/8) 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 15:44:49,919 INFO [train.py:904] (2/8) Epoch 1, batch 8750, loss[loss=0.3013, simple_loss=0.3724, pruned_loss=0.1151, over 16415.00 frames. ], tot_loss[loss=0.2774, simple_loss=0.3483, pruned_loss=0.1033, over 3038936.54 frames. ], batch size: 165, lr: 3.52e-02, grad_scale: 8.0 2023-04-27 15:45:14,094 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:46:06,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8348, 3.1416, 2.0943, 3.6402, 3.7681, 3.7026, 2.0122, 3.0421], device='cuda:2'), covar=tensor([0.2055, 0.0286, 0.1767, 0.0105, 0.0109, 0.0267, 0.1151, 0.0473], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0091, 0.0160, 0.0060, 0.0065, 0.0079, 0.0135, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:46:42,406 INFO [train.py:904] (2/8) Epoch 1, batch 8800, loss[loss=0.2756, simple_loss=0.3485, pruned_loss=0.1013, over 16847.00 frames. ], tot_loss[loss=0.2756, simple_loss=0.3468, pruned_loss=0.1021, over 3049829.59 frames. ], batch size: 116, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:47:13,493 INFO [optim.py:368] (2/8) 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 15:47:25,124 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:47:47,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6861, 3.2670, 1.9148, 4.1061, 4.1967, 4.0922, 2.3611, 3.2228], device='cuda:2'), covar=tensor([0.2204, 0.0319, 0.1899, 0.0075, 0.0092, 0.0235, 0.0969, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0091, 0.0158, 0.0059, 0.0065, 0.0080, 0.0135, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 15:48:15,476 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3467, 5.6151, 5.3473, 5.5715, 4.9171, 4.9025, 5.1257, 5.6676], device='cuda:2'), covar=tensor([0.0241, 0.0544, 0.0466, 0.0234, 0.0441, 0.0236, 0.0391, 0.0333], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0201, 0.0179, 0.0118, 0.0147, 0.0122, 0.0167, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-27 15:48:15,832 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 2023-04-27 15:48:19,469 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-27 15:48:27,336 INFO [train.py:904] (2/8) Epoch 1, batch 8850, loss[loss=0.2584, simple_loss=0.3403, pruned_loss=0.08823, over 17221.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3485, pruned_loss=0.1005, over 3048795.92 frames. ], batch size: 45, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:50:04,404 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-27 15:50:13,539 INFO [train.py:904] (2/8) Epoch 1, batch 8900, loss[loss=0.2595, simple_loss=0.3452, pruned_loss=0.08685, over 16861.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3478, pruned_loss=0.09886, over 3058920.67 frames. ], batch size: 96, lr: 3.50e-02, grad_scale: 8.0 2023-04-27 15:50:42,918 INFO [optim.py:368] (2/8) 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 15:50:56,857 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:13,373 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:52:13,511 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 15:52:19,950 INFO [train.py:904] (2/8) Epoch 1, batch 8950, loss[loss=0.313, simple_loss=0.3734, pruned_loss=0.1263, over 12861.00 frames. ], tot_loss[loss=0.2725, simple_loss=0.3474, pruned_loss=0.09881, over 3088549.62 frames. ], batch size: 248, lr: 3.49e-02, grad_scale: 8.0 2023-04-27 15:53:00,568 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:38,137 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:42,544 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:07,338 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6599, 3.6735, 3.7013, 2.8691, 3.5711, 3.4754, 3.6233, 1.9804], device='cuda:2'), covar=tensor([0.1526, 0.0095, 0.0058, 0.0370, 0.0078, 0.0097, 0.0085, 0.0924], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0045, 0.0050, 0.0085, 0.0046, 0.0049, 0.0052, 0.0095], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 15:54:08,281 INFO [train.py:904] (2/8) Epoch 1, batch 9000, loss[loss=0.24, simple_loss=0.3145, pruned_loss=0.08269, over 16916.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3432, pruned_loss=0.09652, over 3075878.83 frames. ], batch size: 116, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:54:08,282 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 15:54:19,187 INFO [train.py:938] (2/8) Epoch 1, validation: loss=0.2299, simple_loss=0.3267, pruned_loss=0.06658, over 944034.00 frames. 2023-04-27 15:54:19,188 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 15:54:46,212 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 15:54:56,843 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3230, 3.1477, 3.0556, 1.4730, 3.1856, 3.1964, 2.7403, 2.9367], device='cuda:2'), covar=tensor([0.0610, 0.0109, 0.0182, 0.2020, 0.0114, 0.0090, 0.0321, 0.0239], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0066, 0.0064, 0.0148, 0.0062, 0.0063, 0.0078, 0.0088], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-27 15:55:07,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6378, 3.8466, 3.8530, 2.7468, 3.7332, 3.6378, 3.8298, 2.1317], device='cuda:2'), covar=tensor([0.1526, 0.0070, 0.0048, 0.0449, 0.0066, 0.0082, 0.0062, 0.0820], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0044, 0.0049, 0.0084, 0.0045, 0.0048, 0.0050, 0.0094], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 15:55:56,680 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 1, batch 9050, loss[loss=0.2557, simple_loss=0.3258, pruned_loss=0.09277, over 16645.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3464, pruned_loss=0.09854, over 3082104.62 frames. ], batch size: 89, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:56:26,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8685, 4.5347, 4.2699, 4.1739, 4.6959, 2.7022, 4.2712, 4.5402], device='cuda:2'), covar=tensor([0.0076, 0.0067, 0.0074, 0.0180, 0.0042, 0.0735, 0.0069, 0.0099], device='cuda:2'), in_proj_covar=tensor([0.0044, 0.0039, 0.0054, 0.0064, 0.0039, 0.0088, 0.0052, 0.0050], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-27 15:56:48,431 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:24,619 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7397, 3.7610, 3.6940, 2.6447, 3.5160, 3.5731, 3.5969, 1.7546], device='cuda:2'), covar=tensor([0.1380, 0.0051, 0.0054, 0.0485, 0.0083, 0.0094, 0.0069, 0.0985], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0045, 0.0050, 0.0086, 0.0046, 0.0048, 0.0051, 0.0094], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 15:57:45,072 INFO [train.py:904] (2/8) Epoch 1, batch 9100, loss[loss=0.2864, simple_loss=0.3503, pruned_loss=0.1112, over 12488.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3466, pruned_loss=0.09952, over 3077747.30 frames. ], batch size: 248, lr: 3.47e-02, grad_scale: 8.0 2023-04-27 15:58:01,498 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3398, 1.4198, 1.8981, 2.2673, 2.4058, 2.3453, 1.6467, 2.5533], device='cuda:2'), covar=tensor([0.0069, 0.0595, 0.0259, 0.0147, 0.0094, 0.0163, 0.0325, 0.0068], device='cuda:2'), in_proj_covar=tensor([0.0049, 0.0093, 0.0072, 0.0054, 0.0042, 0.0044, 0.0070, 0.0040], device='cuda:2'), out_proj_covar=tensor([9.0344e-05, 1.6839e-04, 1.3881e-04, 1.0127e-04, 7.4758e-05, 8.2228e-05, 1.2145e-04, 7.4425e-05], device='cuda:2') 2023-04-27 15:58:14,292 INFO [optim.py:368] (2/8) 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 15:58:14,874 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:15,871 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 15:59:42,636 INFO [train.py:904] (2/8) Epoch 1, batch 9150, loss[loss=0.2811, simple_loss=0.3424, pruned_loss=0.1099, over 12005.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3465, pruned_loss=0.09944, over 3050662.51 frames. ], batch size: 248, lr: 3.46e-02, grad_scale: 8.0 2023-04-27 16:01:15,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6635, 3.6127, 3.6431, 3.9725, 3.9236, 3.6836, 3.8655, 3.8438], device='cuda:2'), covar=tensor([0.0327, 0.0285, 0.0716, 0.0259, 0.0302, 0.0587, 0.0323, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0160, 0.0238, 0.0168, 0.0141, 0.0148, 0.0133, 0.0142], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:01:27,878 INFO [train.py:904] (2/8) Epoch 1, batch 9200, loss[loss=0.2599, simple_loss=0.332, pruned_loss=0.09396, over 16288.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3416, pruned_loss=0.09785, over 3051260.67 frames. ], batch size: 165, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:01:54,928 INFO [optim.py:368] (2/8) 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 16:03:04,216 INFO [train.py:904] (2/8) Epoch 1, batch 9250, loss[loss=0.2419, simple_loss=0.3079, pruned_loss=0.08799, over 11936.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3409, pruned_loss=0.09817, over 3027686.55 frames. ], batch size: 248, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:04:16,929 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:04:18,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3910, 3.2848, 3.2110, 3.0220, 3.3714, 2.1861, 3.1231, 3.1492], device='cuda:2'), covar=tensor([0.0083, 0.0081, 0.0084, 0.0198, 0.0056, 0.0940, 0.0085, 0.0115], device='cuda:2'), in_proj_covar=tensor([0.0046, 0.0040, 0.0054, 0.0067, 0.0041, 0.0092, 0.0054, 0.0052], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:04:30,824 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 16:04:52,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1885, 3.2029, 2.3689, 3.3040, 3.2408, 3.3388, 3.2831, 3.2375], device='cuda:2'), covar=tensor([0.0725, 0.0659, 0.2471, 0.1100, 0.1154, 0.0957, 0.0980, 0.1093], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0161, 0.0240, 0.0168, 0.0141, 0.0145, 0.0132, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:04:58,120 INFO [train.py:904] (2/8) Epoch 1, batch 9300, loss[loss=0.2636, simple_loss=0.3217, pruned_loss=0.1028, over 12368.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3383, pruned_loss=0.09615, over 3030007.15 frames. ], batch size: 246, lr: 3.44e-02, grad_scale: 8.0 2023-04-27 16:05:33,415 INFO [optim.py:368] (2/8) 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 16:05:40,481 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-04-27 16:06:12,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3172, 3.8475, 3.1151, 4.7053, 2.9921, 4.6166, 3.3751, 2.8786], device='cuda:2'), covar=tensor([0.0216, 0.0239, 0.0278, 0.0168, 0.1140, 0.0118, 0.0441, 0.1092], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0126, 0.0103, 0.0144, 0.0202, 0.0120, 0.0144, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:06:27,250 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:06:39,848 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9605, 1.4702, 1.5299, 1.2556, 1.5742, 1.6697, 1.7840, 1.7580], device='cuda:2'), covar=tensor([0.0037, 0.0240, 0.0128, 0.0202, 0.0095, 0.0153, 0.0061, 0.0093], device='cuda:2'), in_proj_covar=tensor([0.0032, 0.0074, 0.0057, 0.0064, 0.0050, 0.0062, 0.0033, 0.0043], device='cuda:2'), out_proj_covar=tensor([4.5073e-05, 1.1362e-04, 8.6440e-05, 9.8948e-05, 7.7637e-05, 9.2613e-05, 5.1728e-05, 6.9704e-05], device='cuda:2') 2023-04-27 16:06:44,327 INFO [train.py:904] (2/8) Epoch 1, batch 9350, loss[loss=0.2913, simple_loss=0.358, pruned_loss=0.1123, over 16248.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3383, pruned_loss=0.09594, over 3035802.91 frames. ], batch size: 165, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:07:24,456 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:07:31,433 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 16:08:07,846 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 16:08:26,592 INFO [train.py:904] (2/8) Epoch 1, batch 9400, loss[loss=0.2773, simple_loss=0.3667, pruned_loss=0.09395, over 16853.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3374, pruned_loss=0.09497, over 3027334.84 frames. ], batch size: 96, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:08:54,858 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:56,912 INFO [optim.py:368] (2/8) 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 16:08:57,387 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:08,169 INFO [train.py:904] (2/8) Epoch 1, batch 9450, loss[loss=0.2522, simple_loss=0.333, pruned_loss=0.08567, over 16260.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3388, pruned_loss=0.09548, over 3017808.10 frames. ], batch size: 165, lr: 3.42e-02, grad_scale: 8.0 2023-04-27 16:10:33,967 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:58,537 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:36,276 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 16:11:50,036 INFO [train.py:904] (2/8) Epoch 1, batch 9500, loss[loss=0.2565, simple_loss=0.3329, pruned_loss=0.09002, over 16904.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3379, pruned_loss=0.09469, over 3031351.66 frames. ], batch size: 96, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:12:21,411 INFO [optim.py:368] (2/8) 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 16:13:37,007 INFO [train.py:904] (2/8) Epoch 1, batch 9550, loss[loss=0.3027, simple_loss=0.3774, pruned_loss=0.114, over 15654.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3382, pruned_loss=0.09514, over 3048589.95 frames. ], batch size: 194, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:13:58,301 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0998, 3.1277, 3.0306, 1.6189, 3.1283, 3.2025, 2.9278, 2.7978], device='cuda:2'), covar=tensor([0.0753, 0.0124, 0.0248, 0.1758, 0.0111, 0.0088, 0.0267, 0.0270], device='cuda:2'), in_proj_covar=tensor([0.0105, 0.0072, 0.0071, 0.0149, 0.0064, 0.0066, 0.0078, 0.0092], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:14:41,176 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5559, 3.3664, 3.5049, 3.8862, 3.8158, 3.6508, 3.9057, 3.7962], device='cuda:2'), covar=tensor([0.0391, 0.0379, 0.0829, 0.0322, 0.0373, 0.0522, 0.0267, 0.0285], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0162, 0.0237, 0.0165, 0.0138, 0.0146, 0.0133, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:14:46,577 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:14:48,660 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:15:18,632 INFO [train.py:904] (2/8) Epoch 1, batch 9600, loss[loss=0.3018, simple_loss=0.3716, pruned_loss=0.116, over 15265.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.3403, pruned_loss=0.09646, over 3049404.78 frames. ], batch size: 191, lr: 3.40e-02, grad_scale: 8.0 2023-04-27 16:15:33,015 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1055, 2.8468, 2.1805, 3.3000, 3.3475, 3.3536, 2.0974, 2.7825], device='cuda:2'), covar=tensor([0.1593, 0.0303, 0.1471, 0.0103, 0.0149, 0.0290, 0.0932, 0.0520], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0096, 0.0159, 0.0060, 0.0071, 0.0086, 0.0139, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:15:48,621 INFO [optim.py:368] (2/8) 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 16:16:16,188 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6741, 3.4644, 3.3132, 2.7070, 3.3612, 3.2186, 3.5229, 1.7981], device='cuda:2'), covar=tensor([0.1192, 0.0062, 0.0096, 0.0368, 0.0056, 0.0096, 0.0064, 0.0798], device='cuda:2'), in_proj_covar=tensor([0.0114, 0.0045, 0.0052, 0.0086, 0.0046, 0.0049, 0.0050, 0.0095], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 16:16:20,294 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:45,902 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:53,223 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:08,566 INFO [train.py:904] (2/8) Epoch 1, batch 9650, loss[loss=0.2265, simple_loss=0.3096, pruned_loss=0.07176, over 17024.00 frames. ], tot_loss[loss=0.2688, simple_loss=0.3427, pruned_loss=0.09747, over 3032559.29 frames. ], batch size: 109, lr: 3.39e-02, grad_scale: 8.0 2023-04-27 16:17:21,393 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0121, 2.7907, 2.1872, 3.2232, 3.3016, 3.3867, 2.0972, 2.8168], device='cuda:2'), covar=tensor([0.1559, 0.0292, 0.1324, 0.0118, 0.0134, 0.0255, 0.0911, 0.0499], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0096, 0.0159, 0.0060, 0.0071, 0.0084, 0.0138, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:17:26,829 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:34,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4756, 3.4603, 1.2281, 3.4720, 2.1893, 3.2987, 1.4458, 2.6264], device='cuda:2'), covar=tensor([0.0059, 0.0098, 0.2041, 0.0062, 0.0874, 0.0318, 0.1808, 0.0611], device='cuda:2'), in_proj_covar=tensor([0.0066, 0.0076, 0.0158, 0.0070, 0.0138, 0.0089, 0.0165, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 16:17:47,473 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 16:17:52,782 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:35,170 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:57,048 INFO [train.py:904] (2/8) Epoch 1, batch 9700, loss[loss=0.2731, simple_loss=0.3484, pruned_loss=0.09886, over 15438.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3403, pruned_loss=0.09579, over 3035285.14 frames. ], batch size: 191, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:18:59,616 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6860, 3.6998, 3.7845, 3.7945, 3.7732, 4.1830, 4.0948, 3.8407], device='cuda:2'), covar=tensor([0.1345, 0.1534, 0.0866, 0.1742, 0.2439, 0.0879, 0.0707, 0.2033], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0216, 0.0180, 0.0185, 0.0224, 0.0178, 0.0154, 0.0230], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-27 16:19:24,736 INFO [optim.py:368] (2/8) 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 16:19:28,115 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:33,509 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:20:27,187 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:20:40,562 INFO [train.py:904] (2/8) Epoch 1, batch 9750, loss[loss=0.2804, simple_loss=0.3597, pruned_loss=0.1005, over 15330.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3387, pruned_loss=0.09554, over 3049466.65 frames. ], batch size: 190, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:21:17,327 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:19,291 INFO [train.py:904] (2/8) Epoch 1, batch 9800, loss[loss=0.2426, simple_loss=0.3349, pruned_loss=0.07514, over 16421.00 frames. ], tot_loss[loss=0.2634, simple_loss=0.3384, pruned_loss=0.09421, over 3048705.50 frames. ], batch size: 68, lr: 3.37e-02, grad_scale: 8.0 2023-04-27 16:22:26,087 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:47,902 INFO [optim.py:368] (2/8) 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 16:22:57,778 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7445, 4.3153, 4.2991, 4.0294, 4.4194, 1.6655, 4.1474, 4.4794], device='cuda:2'), covar=tensor([0.0053, 0.0076, 0.0057, 0.0175, 0.0050, 0.1419, 0.0058, 0.0083], device='cuda:2'), in_proj_covar=tensor([0.0044, 0.0039, 0.0054, 0.0064, 0.0040, 0.0093, 0.0052, 0.0051], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-27 16:24:05,875 INFO [train.py:904] (2/8) Epoch 1, batch 9850, loss[loss=0.2594, simple_loss=0.3406, pruned_loss=0.08907, over 16983.00 frames. ], tot_loss[loss=0.264, simple_loss=0.34, pruned_loss=0.094, over 3059169.58 frames. ], batch size: 109, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:24:50,127 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4217, 3.5877, 3.8243, 3.8646, 3.8373, 3.6248, 3.3993, 3.7682], device='cuda:2'), covar=tensor([0.0384, 0.0437, 0.0530, 0.0554, 0.0560, 0.0369, 0.0905, 0.0386], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0109, 0.0122, 0.0127, 0.0134, 0.0115, 0.0160, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:25:18,138 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5486, 3.3591, 1.4449, 3.5701, 2.1007, 3.5155, 1.7666, 2.7922], device='cuda:2'), covar=tensor([0.0052, 0.0163, 0.2154, 0.0054, 0.1153, 0.0293, 0.1749, 0.0750], device='cuda:2'), in_proj_covar=tensor([0.0064, 0.0077, 0.0157, 0.0069, 0.0137, 0.0087, 0.0164, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 16:25:58,729 INFO [train.py:904] (2/8) Epoch 1, batch 9900, loss[loss=0.2412, simple_loss=0.3145, pruned_loss=0.08401, over 12123.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.3398, pruned_loss=0.09339, over 3043554.17 frames. ], batch size: 247, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:26:31,897 INFO [optim.py:368] (2/8) 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 16:27:32,224 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:27:35,224 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.86 vs. limit=5.0 2023-04-27 16:27:57,263 INFO [train.py:904] (2/8) Epoch 1, batch 9950, loss[loss=0.2531, simple_loss=0.3359, pruned_loss=0.08522, over 16116.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3422, pruned_loss=0.09407, over 3044543.54 frames. ], batch size: 165, lr: 3.35e-02, grad_scale: 8.0 2023-04-27 16:28:33,746 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5397, 6.0165, 5.7219, 5.9516, 5.2882, 5.2616, 5.4701, 6.1763], device='cuda:2'), covar=tensor([0.0321, 0.0471, 0.0567, 0.0240, 0.0390, 0.0284, 0.0355, 0.0335], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0217, 0.0192, 0.0132, 0.0156, 0.0129, 0.0175, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:30:02,393 INFO [train.py:904] (2/8) Epoch 1, batch 10000, loss[loss=0.2549, simple_loss=0.3455, pruned_loss=0.08212, over 16657.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3395, pruned_loss=0.09237, over 3068529.58 frames. ], batch size: 134, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:30:27,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6082, 3.4804, 2.0415, 3.9424, 3.8949, 3.9399, 1.9211, 3.2029], device='cuda:2'), covar=tensor([0.2006, 0.0267, 0.1689, 0.0071, 0.0115, 0.0184, 0.1158, 0.0475], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0098, 0.0158, 0.0059, 0.0070, 0.0083, 0.0139, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:30:30,128 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:30:32,721 INFO [optim.py:368] (2/8) 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 16:30:35,681 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-27 16:31:21,941 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:31:41,577 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-04-27 16:31:43,672 INFO [train.py:904] (2/8) Epoch 1, batch 10050, loss[loss=0.3044, simple_loss=0.3743, pruned_loss=0.1172, over 16368.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3384, pruned_loss=0.09146, over 3063956.41 frames. ], batch size: 146, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:32:06,029 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:32:21,680 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-04-27 16:32:23,093 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:32:31,189 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-04-27 16:33:16,685 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:18,974 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:33:19,583 INFO [train.py:904] (2/8) Epoch 1, batch 10100, loss[loss=0.2551, simple_loss=0.3211, pruned_loss=0.09452, over 16966.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.3392, pruned_loss=0.09213, over 3078595.83 frames. ], batch size: 109, lr: 3.33e-02, grad_scale: 16.0 2023-04-27 16:33:49,434 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:34:06,077 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:34:25,742 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6825, 3.7611, 3.3207, 3.3963, 2.6036, 2.2469, 3.9928, 4.4030], device='cuda:2'), covar=tensor([0.1624, 0.0598, 0.0859, 0.0311, 0.1514, 0.1291, 0.0232, 0.0062], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0197, 0.0213, 0.0134, 0.0174, 0.0166, 0.0151, 0.0084], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-27 16:35:05,143 INFO [train.py:904] (2/8) Epoch 2, batch 0, loss[loss=0.2618, simple_loss=0.3293, pruned_loss=0.09718, over 16993.00 frames. ], tot_loss[loss=0.2618, simple_loss=0.3293, pruned_loss=0.09718, over 16993.00 frames. ], batch size: 41, lr: 3.26e-02, grad_scale: 8.0 2023-04-27 16:35:05,143 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 16:35:12,735 INFO [train.py:938] (2/8) Epoch 2, validation: loss=0.2235, simple_loss=0.3221, pruned_loss=0.06242, over 944034.00 frames. 2023-04-27 16:35:12,736 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 16:35:57,332 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 16:36:22,853 INFO [train.py:904] (2/8) Epoch 2, batch 50, loss[loss=0.3541, simple_loss=0.3849, pruned_loss=0.1617, over 16408.00 frames. ], tot_loss[loss=0.3239, simple_loss=0.374, pruned_loss=0.1369, over 740885.18 frames. ], batch size: 145, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:36:45,679 INFO [optim.py:368] (2/8) 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 16:37:04,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1978, 4.1289, 3.8545, 1.8163, 3.0900, 2.2183, 3.6953, 4.0986], device='cuda:2'), covar=tensor([0.0224, 0.0346, 0.0271, 0.1799, 0.0704, 0.1120, 0.0547, 0.0227], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0085, 0.0125, 0.0155, 0.0143, 0.0139, 0.0138, 0.0084], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-27 16:37:16,966 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:37:30,888 INFO [train.py:904] (2/8) Epoch 2, batch 100, loss[loss=0.3957, simple_loss=0.4244, pruned_loss=0.1834, over 12223.00 frames. ], tot_loss[loss=0.3123, simple_loss=0.3647, pruned_loss=0.1299, over 1317049.95 frames. ], batch size: 246, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:37:33,024 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2327, 5.0002, 4.6998, 4.2817, 4.8736, 2.4992, 4.5469, 4.9763], device='cuda:2'), covar=tensor([0.0065, 0.0068, 0.0080, 0.0284, 0.0057, 0.1158, 0.0079, 0.0100], device='cuda:2'), in_proj_covar=tensor([0.0049, 0.0043, 0.0062, 0.0076, 0.0044, 0.0099, 0.0058, 0.0059], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:38:03,392 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1050, 4.0796, 3.7049, 3.5768, 3.8282, 1.8024, 3.6861, 4.0173], device='cuda:2'), covar=tensor([0.0095, 0.0070, 0.0117, 0.0274, 0.0083, 0.1271, 0.0103, 0.0125], device='cuda:2'), in_proj_covar=tensor([0.0049, 0.0043, 0.0063, 0.0077, 0.0044, 0.0099, 0.0059, 0.0059], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:38:22,578 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:38:38,636 INFO [train.py:904] (2/8) Epoch 2, batch 150, loss[loss=0.3422, simple_loss=0.3805, pruned_loss=0.1519, over 15456.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.3588, pruned_loss=0.1237, over 1765280.27 frames. ], batch size: 190, lr: 3.24e-02, grad_scale: 4.0 2023-04-27 16:38:56,092 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:01,491 INFO [optim.py:368] (2/8) 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 16:39:24,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6499, 4.3929, 4.5413, 4.9918, 5.0263, 4.4143, 5.0898, 4.8243], device='cuda:2'), covar=tensor([0.0382, 0.0382, 0.0915, 0.0297, 0.0309, 0.0372, 0.0232, 0.0285], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0208, 0.0314, 0.0215, 0.0176, 0.0175, 0.0167, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:39:47,448 INFO [train.py:904] (2/8) Epoch 2, batch 200, loss[loss=0.3382, simple_loss=0.3928, pruned_loss=0.1418, over 15479.00 frames. ], tot_loss[loss=0.3033, simple_loss=0.3593, pruned_loss=0.1237, over 2110271.51 frames. ], batch size: 190, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:40:03,409 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:50,468 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:40:55,037 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:58,054 INFO [train.py:904] (2/8) Epoch 2, batch 250, loss[loss=0.2302, simple_loss=0.2971, pruned_loss=0.08163, over 17009.00 frames. ], tot_loss[loss=0.3, simple_loss=0.3556, pruned_loss=0.1222, over 2370184.35 frames. ], batch size: 41, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:41:11,020 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:41:22,650 INFO [optim.py:368] (2/8) 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 16:41:22,933 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:42:03,850 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:09,609 INFO [train.py:904] (2/8) Epoch 2, batch 300, loss[loss=0.2799, simple_loss=0.3319, pruned_loss=0.114, over 16810.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.351, pruned_loss=0.1191, over 2581491.93 frames. ], batch size: 83, lr: 3.22e-02, grad_scale: 4.0 2023-04-27 16:42:14,783 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:21,298 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0760, 2.1106, 2.1605, 1.9748, 2.6454, 2.6898, 2.8368, 2.9032], device='cuda:2'), covar=tensor([0.0144, 0.0224, 0.0140, 0.0192, 0.0105, 0.0128, 0.0074, 0.0080], device='cuda:2'), in_proj_covar=tensor([0.0042, 0.0083, 0.0069, 0.0075, 0.0066, 0.0073, 0.0041, 0.0050], device='cuda:2'), out_proj_covar=tensor([6.1097e-05, 1.2787e-04, 1.0523e-04, 1.1566e-04, 1.0299e-04, 1.1438e-04, 6.5031e-05, 8.1521e-05], device='cuda:2') 2023-04-27 16:42:36,296 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:43:19,861 INFO [train.py:904] (2/8) Epoch 2, batch 350, loss[loss=0.2927, simple_loss=0.3374, pruned_loss=0.124, over 16717.00 frames. ], tot_loss[loss=0.2875, simple_loss=0.3452, pruned_loss=0.1149, over 2741592.07 frames. ], batch size: 134, lr: 3.21e-02, grad_scale: 4.0 2023-04-27 16:43:39,864 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:43:42,893 INFO [optim.py:368] (2/8) 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 16:44:27,071 INFO [train.py:904] (2/8) Epoch 2, batch 400, loss[loss=0.2644, simple_loss=0.3378, pruned_loss=0.09554, over 17122.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3435, pruned_loss=0.1137, over 2879335.06 frames. ], batch size: 49, lr: 3.21e-02, grad_scale: 8.0 2023-04-27 16:44:29,277 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8694, 4.6480, 4.3900, 3.6759, 4.6344, 2.1505, 4.3672, 4.6772], device='cuda:2'), covar=tensor([0.0072, 0.0087, 0.0090, 0.0425, 0.0069, 0.1451, 0.0091, 0.0113], device='cuda:2'), in_proj_covar=tensor([0.0056, 0.0047, 0.0068, 0.0089, 0.0049, 0.0101, 0.0064, 0.0065], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:45:31,822 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9670, 4.3053, 3.6551, 3.9438, 3.0107, 2.1515, 4.7818, 5.2249], device='cuda:2'), covar=tensor([0.1897, 0.0547, 0.0985, 0.0370, 0.2121, 0.1511, 0.0193, 0.0054], device='cuda:2'), in_proj_covar=tensor([0.0230, 0.0203, 0.0221, 0.0142, 0.0213, 0.0169, 0.0160, 0.0098], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-27 16:45:36,485 INFO [train.py:904] (2/8) Epoch 2, batch 450, loss[loss=0.2768, simple_loss=0.3259, pruned_loss=0.1139, over 16706.00 frames. ], tot_loss[loss=0.2812, simple_loss=0.3403, pruned_loss=0.1111, over 2983206.35 frames. ], batch size: 134, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:45:51,414 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 16:45:59,585 INFO [optim.py:368] (2/8) 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 16:46:44,100 INFO [train.py:904] (2/8) Epoch 2, batch 500, loss[loss=0.2507, simple_loss=0.3261, pruned_loss=0.08761, over 17197.00 frames. ], tot_loss[loss=0.2778, simple_loss=0.3377, pruned_loss=0.1089, over 3062740.99 frames. ], batch size: 45, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:47:44,952 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:47:52,718 INFO [train.py:904] (2/8) Epoch 2, batch 550, loss[loss=0.2832, simple_loss=0.3344, pruned_loss=0.116, over 16849.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3369, pruned_loss=0.1085, over 3124201.28 frames. ], batch size: 96, lr: 3.19e-02, grad_scale: 8.0 2023-04-27 16:48:02,224 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8555, 1.7276, 1.7501, 1.6721, 2.6236, 2.3556, 3.2286, 3.2674], device='cuda:2'), covar=tensor([0.0038, 0.0264, 0.0195, 0.0263, 0.0107, 0.0159, 0.0040, 0.0061], device='cuda:2'), in_proj_covar=tensor([0.0043, 0.0082, 0.0069, 0.0078, 0.0069, 0.0077, 0.0042, 0.0052], device='cuda:2'), out_proj_covar=tensor([6.3882e-05, 1.2638e-04, 1.0577e-04, 1.2145e-04, 1.0777e-04, 1.2081e-04, 6.6777e-05, 8.5709e-05], device='cuda:2') 2023-04-27 16:48:17,046 INFO [optim.py:368] (2/8) 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 16:48:17,339 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:48:17,827 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 16:48:43,724 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 16:48:51,486 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:52,707 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:02,286 INFO [train.py:904] (2/8) Epoch 2, batch 600, loss[loss=0.2673, simple_loss=0.325, pruned_loss=0.1048, over 15451.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3352, pruned_loss=0.1084, over 3164186.66 frames. ], batch size: 190, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:49:21,916 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:22,402 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-27 16:49:24,353 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:49:33,730 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:11,093 INFO [train.py:904] (2/8) Epoch 2, batch 650, loss[loss=0.255, simple_loss=0.3138, pruned_loss=0.09806, over 16596.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3339, pruned_loss=0.1079, over 3208457.05 frames. ], batch size: 75, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:50:16,432 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:19,697 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-04-27 16:50:23,981 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:50:30,055 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0542, 3.8321, 3.9752, 4.4172, 4.3424, 4.1236, 4.2811, 4.2209], device='cuda:2'), covar=tensor([0.0451, 0.0522, 0.1172, 0.0297, 0.0441, 0.0626, 0.0408, 0.0380], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0232, 0.0352, 0.0243, 0.0198, 0.0203, 0.0177, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:50:33,023 INFO [optim.py:368] (2/8) 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 16:50:56,206 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2023-04-27 16:50:57,062 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:51:15,335 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.17 vs. limit=5.0 2023-04-27 16:51:19,283 INFO [train.py:904] (2/8) Epoch 2, batch 700, loss[loss=0.2504, simple_loss=0.3269, pruned_loss=0.08694, over 17107.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3337, pruned_loss=0.1069, over 3239199.90 frames. ], batch size: 47, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:51:25,559 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9035, 5.5195, 5.3489, 5.3846, 5.4346, 5.9248, 5.8127, 5.4099], device='cuda:2'), covar=tensor([0.0557, 0.0965, 0.1070, 0.1352, 0.1837, 0.0607, 0.0617, 0.1623], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0267, 0.0228, 0.0224, 0.0288, 0.0225, 0.0191, 0.0296], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 16:52:06,312 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:52:26,663 INFO [train.py:904] (2/8) Epoch 2, batch 750, loss[loss=0.2452, simple_loss=0.3154, pruned_loss=0.08755, over 17216.00 frames. ], tot_loss[loss=0.2736, simple_loss=0.3342, pruned_loss=0.1064, over 3259232.50 frames. ], batch size: 44, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:52:50,207 INFO [optim.py:368] (2/8) 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 16:53:00,721 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 2023-04-27 16:53:28,035 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:53:33,387 INFO [train.py:904] (2/8) Epoch 2, batch 800, loss[loss=0.3023, simple_loss=0.3475, pruned_loss=0.1286, over 16266.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.3336, pruned_loss=0.1066, over 3265251.54 frames. ], batch size: 165, lr: 3.16e-02, grad_scale: 8.0 2023-04-27 16:54:21,215 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-27 16:54:42,354 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9364, 1.7254, 2.3871, 2.6801, 3.1979, 2.9288, 1.7654, 2.9556], device='cuda:2'), covar=tensor([0.0068, 0.0449, 0.0209, 0.0183, 0.0063, 0.0149, 0.0344, 0.0058], device='cuda:2'), in_proj_covar=tensor([0.0064, 0.0103, 0.0086, 0.0072, 0.0049, 0.0054, 0.0087, 0.0048], device='cuda:2'), out_proj_covar=tensor([1.1805e-04, 1.8511e-04, 1.6461e-04, 1.3612e-04, 8.8285e-05, 1.0553e-04, 1.5131e-04, 8.9609e-05], device='cuda:2') 2023-04-27 16:54:43,055 INFO [train.py:904] (2/8) Epoch 2, batch 850, loss[loss=0.2506, simple_loss=0.3228, pruned_loss=0.08922, over 16448.00 frames. ], tot_loss[loss=0.2732, simple_loss=0.3337, pruned_loss=0.1063, over 3285790.26 frames. ], batch size: 68, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:55:06,012 INFO [optim.py:368] (2/8) 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 16:55:49,289 INFO [train.py:904] (2/8) Epoch 2, batch 900, loss[loss=0.247, simple_loss=0.3194, pruned_loss=0.08727, over 16565.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3325, pruned_loss=0.1051, over 3282047.00 frames. ], batch size: 62, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:56:09,960 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:24,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1150, 4.3217, 1.7022, 4.3549, 2.7025, 4.3063, 2.1019, 2.9368], device='cuda:2'), covar=tensor([0.0049, 0.0111, 0.1640, 0.0040, 0.0777, 0.0201, 0.1295, 0.0585], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0088, 0.0164, 0.0073, 0.0146, 0.0108, 0.0165, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 16:56:56,836 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:58,324 INFO [train.py:904] (2/8) Epoch 2, batch 950, loss[loss=0.2978, simple_loss=0.34, pruned_loss=0.1278, over 16436.00 frames. ], tot_loss[loss=0.2703, simple_loss=0.3317, pruned_loss=0.1044, over 3297370.01 frames. ], batch size: 146, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:57:10,371 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:57:13,955 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:57:21,473 INFO [optim.py:368] (2/8) 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 16:57:30,665 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0898, 4.4030, 3.5651, 4.0623, 3.2129, 2.5798, 4.8429, 5.4199], device='cuda:2'), covar=tensor([0.1719, 0.0524, 0.0926, 0.0354, 0.2125, 0.1220, 0.0196, 0.0048], device='cuda:2'), in_proj_covar=tensor([0.0233, 0.0207, 0.0223, 0.0148, 0.0228, 0.0170, 0.0165, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-27 16:57:38,008 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:06,961 INFO [train.py:904] (2/8) Epoch 2, batch 1000, loss[loss=0.259, simple_loss=0.3356, pruned_loss=0.09123, over 17113.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.33, pruned_loss=0.1039, over 3310120.25 frames. ], batch size: 49, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:58:12,167 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1926, 4.3056, 2.4917, 5.0981, 5.0190, 4.4876, 2.4927, 4.2628], device='cuda:2'), covar=tensor([0.1473, 0.0270, 0.1484, 0.0077, 0.0105, 0.0272, 0.1046, 0.0314], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0104, 0.0158, 0.0063, 0.0084, 0.0092, 0.0140, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 16:58:16,902 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:37,468 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 2, batch 1050, loss[loss=0.2978, simple_loss=0.3378, pruned_loss=0.1289, over 16698.00 frames. ], tot_loss[loss=0.2692, simple_loss=0.3299, pruned_loss=0.1043, over 3315391.68 frames. ], batch size: 124, lr: 3.13e-02, grad_scale: 8.0 2023-04-27 16:59:36,187 INFO [optim.py:368] (2/8) 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 17:00:00,098 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:00,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2114, 3.8036, 2.8235, 4.4846, 2.5198, 4.2997, 3.0235, 2.6450], device='cuda:2'), covar=tensor([0.0276, 0.0277, 0.0319, 0.0191, 0.1442, 0.0177, 0.0591, 0.1526], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0146, 0.0121, 0.0169, 0.0226, 0.0139, 0.0159, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:00:08,425 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:00:21,895 INFO [train.py:904] (2/8) Epoch 2, batch 1100, loss[loss=0.243, simple_loss=0.3208, pruned_loss=0.0826, over 16756.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3297, pruned_loss=0.1029, over 3312314.20 frames. ], batch size: 62, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:01:28,364 INFO [train.py:904] (2/8) Epoch 2, batch 1150, loss[loss=0.2745, simple_loss=0.347, pruned_loss=0.101, over 17112.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.329, pruned_loss=0.1014, over 3322541.15 frames. ], batch size: 49, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:01:52,088 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8423, 4.1562, 3.2392, 3.3253, 3.0509, 2.1703, 4.2747, 5.1323], device='cuda:2'), covar=tensor([0.1822, 0.0544, 0.1017, 0.0543, 0.1940, 0.1452, 0.0285, 0.0063], device='cuda:2'), in_proj_covar=tensor([0.0231, 0.0210, 0.0223, 0.0147, 0.0230, 0.0170, 0.0167, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-27 17:01:52,662 INFO [optim.py:368] (2/8) 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 17:02:39,328 INFO [train.py:904] (2/8) Epoch 2, batch 1200, loss[loss=0.2257, simple_loss=0.3058, pruned_loss=0.07283, over 17124.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3267, pruned_loss=0.1003, over 3331488.39 frames. ], batch size: 48, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:03:46,800 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:47,486 INFO [train.py:904] (2/8) Epoch 2, batch 1250, loss[loss=0.2751, simple_loss=0.3349, pruned_loss=0.1077, over 15955.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3265, pruned_loss=0.1014, over 3325735.79 frames. ], batch size: 35, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:04:03,689 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2654, 3.1620, 3.1017, 3.5102, 3.4511, 3.3063, 3.4308, 3.3939], device='cuda:2'), covar=tensor([0.0387, 0.0434, 0.1103, 0.0349, 0.0411, 0.1017, 0.0404, 0.0403], device='cuda:2'), in_proj_covar=tensor([0.0210, 0.0234, 0.0351, 0.0244, 0.0205, 0.0201, 0.0183, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:04:10,333 INFO [optim.py:368] (2/8) 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 17:04:26,864 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:29,347 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2116, 2.2828, 2.5597, 2.3106, 2.9669, 2.7261, 3.5949, 3.5159], device='cuda:2'), covar=tensor([0.0029, 0.0236, 0.0136, 0.0192, 0.0100, 0.0150, 0.0067, 0.0053], device='cuda:2'), in_proj_covar=tensor([0.0040, 0.0082, 0.0072, 0.0079, 0.0071, 0.0080, 0.0045, 0.0054], device='cuda:2'), out_proj_covar=tensor([6.3026e-05, 1.2626e-04, 1.1194e-04, 1.2419e-04, 1.1284e-04, 1.2843e-04, 7.0223e-05, 9.0512e-05], device='cuda:2') 2023-04-27 17:04:49,846 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:50,052 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0354, 3.9837, 1.7791, 4.0953, 2.6479, 4.0622, 1.9369, 2.9888], device='cuda:2'), covar=tensor([0.0047, 0.0128, 0.1628, 0.0041, 0.0761, 0.0216, 0.1335, 0.0518], device='cuda:2'), in_proj_covar=tensor([0.0076, 0.0090, 0.0164, 0.0077, 0.0150, 0.0112, 0.0167, 0.0142], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 17:04:53,724 INFO [train.py:904] (2/8) Epoch 2, batch 1300, loss[loss=0.2388, simple_loss=0.3081, pruned_loss=0.08477, over 16799.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3267, pruned_loss=0.1022, over 3322154.47 frames. ], batch size: 39, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:05:30,716 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:01,549 INFO [train.py:904] (2/8) Epoch 2, batch 1350, loss[loss=0.2338, simple_loss=0.3052, pruned_loss=0.08118, over 17207.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3261, pruned_loss=0.101, over 3313086.06 frames. ], batch size: 46, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:06:24,754 INFO [optim.py:368] (2/8) 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 17:06:41,102 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:45,023 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:57,536 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:07:09,714 INFO [train.py:904] (2/8) Epoch 2, batch 1400, loss[loss=0.3036, simple_loss=0.3486, pruned_loss=0.1293, over 12334.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3263, pruned_loss=0.1016, over 3314444.68 frames. ], batch size: 246, lr: 3.09e-02, grad_scale: 8.0 2023-04-27 17:07:32,217 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-04-27 17:07:46,406 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-27 17:08:02,438 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 17:08:03,056 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:08:08,144 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:08:17,621 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 17:08:19,413 INFO [train.py:904] (2/8) Epoch 2, batch 1450, loss[loss=0.2463, simple_loss=0.3176, pruned_loss=0.08748, over 17204.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.325, pruned_loss=0.1003, over 3317700.42 frames. ], batch size: 46, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:08:25,056 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-04-27 17:08:43,789 INFO [optim.py:368] (2/8) 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 17:08:48,381 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9193, 4.7377, 4.8283, 5.3171, 5.3424, 4.7995, 5.2821, 5.1601], device='cuda:2'), covar=tensor([0.0484, 0.0475, 0.1233, 0.0289, 0.0338, 0.0414, 0.0293, 0.0298], device='cuda:2'), in_proj_covar=tensor([0.0222, 0.0245, 0.0368, 0.0265, 0.0212, 0.0210, 0.0194, 0.0214], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:09:06,493 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:09:25,389 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3699, 3.8009, 3.5994, 1.6854, 3.8904, 3.9344, 3.2851, 3.3375], device='cuda:2'), covar=tensor([0.0652, 0.0055, 0.0202, 0.1401, 0.0063, 0.0048, 0.0240, 0.0194], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0076, 0.0075, 0.0152, 0.0068, 0.0068, 0.0094, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:09:26,110 INFO [train.py:904] (2/8) Epoch 2, batch 1500, loss[loss=0.3021, simple_loss=0.3403, pruned_loss=0.132, over 16915.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3257, pruned_loss=0.1012, over 3321639.65 frames. ], batch size: 116, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:09:50,024 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9536, 4.6034, 4.7423, 4.8302, 4.2441, 4.7797, 4.7806, 4.3902], device='cuda:2'), covar=tensor([0.0261, 0.0168, 0.0183, 0.0116, 0.0851, 0.0166, 0.0236, 0.0241], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0088, 0.0164, 0.0132, 0.0202, 0.0129, 0.0113, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:10:14,048 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9065, 2.2781, 2.1215, 2.1020, 2.9386, 2.7528, 3.4346, 3.1542], device='cuda:2'), covar=tensor([0.0028, 0.0166, 0.0141, 0.0175, 0.0079, 0.0113, 0.0040, 0.0086], device='cuda:2'), in_proj_covar=tensor([0.0041, 0.0083, 0.0075, 0.0082, 0.0073, 0.0081, 0.0046, 0.0056], device='cuda:2'), out_proj_covar=tensor([6.5762e-05, 1.2776e-04, 1.1687e-04, 1.2918e-04, 1.1786e-04, 1.3088e-04, 7.2268e-05, 9.5710e-05], device='cuda:2') 2023-04-27 17:10:29,975 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:33,631 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:36,145 INFO [train.py:904] (2/8) Epoch 2, batch 1550, loss[loss=0.2757, simple_loss=0.3371, pruned_loss=0.1071, over 17192.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3272, pruned_loss=0.1025, over 3320715.59 frames. ], batch size: 44, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:10:58,964 INFO [optim.py:368] (2/8) 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 17:11:30,144 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-27 17:11:44,739 INFO [train.py:904] (2/8) Epoch 2, batch 1600, loss[loss=0.2867, simple_loss=0.3302, pruned_loss=0.1216, over 16895.00 frames. ], tot_loss[loss=0.2701, simple_loss=0.3311, pruned_loss=0.1046, over 3315931.63 frames. ], batch size: 116, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:11:57,713 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:12:53,456 INFO [train.py:904] (2/8) Epoch 2, batch 1650, loss[loss=0.2693, simple_loss=0.3256, pruned_loss=0.1065, over 16233.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.3334, pruned_loss=0.1055, over 3322161.44 frames. ], batch size: 165, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:13:03,329 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3745, 4.4348, 2.1080, 4.4571, 2.7106, 4.3337, 2.2224, 3.2635], device='cuda:2'), covar=tensor([0.0065, 0.0109, 0.1472, 0.0053, 0.0926, 0.0237, 0.1540, 0.0589], device='cuda:2'), in_proj_covar=tensor([0.0081, 0.0095, 0.0167, 0.0078, 0.0155, 0.0115, 0.0172, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 17:13:04,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9210, 1.8789, 1.9412, 1.7397, 2.7025, 2.4743, 3.2597, 3.0754], device='cuda:2'), covar=tensor([0.0022, 0.0182, 0.0152, 0.0195, 0.0092, 0.0160, 0.0073, 0.0056], device='cuda:2'), in_proj_covar=tensor([0.0040, 0.0082, 0.0074, 0.0082, 0.0072, 0.0081, 0.0047, 0.0055], device='cuda:2'), out_proj_covar=tensor([6.4536e-05, 1.2648e-04, 1.1501e-04, 1.2974e-04, 1.1488e-04, 1.3124e-04, 7.3873e-05, 9.4132e-05], device='cuda:2') 2023-04-27 17:13:15,755 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:17,808 INFO [optim.py:368] (2/8) 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 17:13:35,479 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:56,913 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:03,726 INFO [train.py:904] (2/8) Epoch 2, batch 1700, loss[loss=0.284, simple_loss=0.3371, pruned_loss=0.1154, over 16864.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3358, pruned_loss=0.1068, over 3317454.47 frames. ], batch size: 109, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:14:40,640 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:41,658 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:55,418 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:13,200 INFO [train.py:904] (2/8) Epoch 2, batch 1750, loss[loss=0.2374, simple_loss=0.3063, pruned_loss=0.0843, over 16830.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3366, pruned_loss=0.1064, over 3307967.97 frames. ], batch size: 42, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:15:21,867 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9925, 3.8603, 4.3814, 4.4467, 4.4956, 3.9651, 4.1694, 4.3345], device='cuda:2'), covar=tensor([0.0319, 0.0405, 0.0351, 0.0366, 0.0354, 0.0339, 0.0643, 0.0290], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0146, 0.0169, 0.0163, 0.0192, 0.0158, 0.0243, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-27 17:15:21,909 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:36,931 INFO [optim.py:368] (2/8) 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 17:15:39,710 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:17,815 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:21,481 INFO [train.py:904] (2/8) Epoch 2, batch 1800, loss[loss=0.2247, simple_loss=0.2991, pruned_loss=0.07519, over 16860.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3377, pruned_loss=0.1068, over 3304872.14 frames. ], batch size: 42, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:17:03,316 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:07,157 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:16,283 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:33,761 INFO [train.py:904] (2/8) Epoch 2, batch 1850, loss[loss=0.3176, simple_loss=0.3664, pruned_loss=0.1344, over 16877.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3395, pruned_loss=0.1071, over 3302823.32 frames. ], batch size: 116, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:17:45,614 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:57,887 INFO [optim.py:368] (2/8) 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 17:18:19,781 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8675, 3.1929, 3.5306, 2.9344, 3.6468, 3.7607, 3.9273, 1.9598], device='cuda:2'), covar=tensor([0.0771, 0.0179, 0.0090, 0.0304, 0.0063, 0.0067, 0.0044, 0.0623], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0048, 0.0056, 0.0095, 0.0048, 0.0052, 0.0056, 0.0100], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:2') 2023-04-27 17:18:35,130 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:18:43,197 INFO [train.py:904] (2/8) Epoch 2, batch 1900, loss[loss=0.2828, simple_loss=0.3563, pruned_loss=0.1047, over 16651.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3379, pruned_loss=0.1051, over 3306420.26 frames. ], batch size: 62, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:18:48,992 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:17,243 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:51,447 INFO [train.py:904] (2/8) Epoch 2, batch 1950, loss[loss=0.3089, simple_loss=0.3587, pruned_loss=0.1295, over 16432.00 frames. ], tot_loss[loss=0.2735, simple_loss=0.3375, pruned_loss=0.1048, over 3297015.37 frames. ], batch size: 146, lr: 3.03e-02, grad_scale: 8.0 2023-04-27 17:20:14,620 INFO [optim.py:368] (2/8) 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 17:20:40,774 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:20:59,921 INFO [train.py:904] (2/8) Epoch 2, batch 2000, loss[loss=0.3404, simple_loss=0.3656, pruned_loss=0.1575, over 16853.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3379, pruned_loss=0.1048, over 3293228.55 frames. ], batch size: 116, lr: 3.02e-02, grad_scale: 8.0 2023-04-27 17:21:06,921 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0456, 4.3834, 2.7325, 5.2945, 5.1800, 4.6875, 2.1338, 4.0008], device='cuda:2'), covar=tensor([0.1728, 0.0282, 0.1363, 0.0057, 0.0122, 0.0312, 0.1201, 0.0451], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0110, 0.0166, 0.0067, 0.0096, 0.0103, 0.0149, 0.0136], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 17:21:28,808 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:37,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5952, 5.7813, 5.1630, 5.9833, 5.3589, 4.7590, 5.6068, 5.8133], device='cuda:2'), covar=tensor([0.0729, 0.0904, 0.1780, 0.0313, 0.0718, 0.0649, 0.0506, 0.0809], device='cuda:2'), in_proj_covar=tensor([0.0210, 0.0295, 0.0266, 0.0178, 0.0202, 0.0172, 0.0234, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:21:51,848 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:57,196 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0190, 2.0086, 2.4378, 2.6906, 2.9257, 2.8732, 1.9125, 2.9279], device='cuda:2'), covar=tensor([0.0053, 0.0251, 0.0146, 0.0115, 0.0047, 0.0082, 0.0196, 0.0058], device='cuda:2'), in_proj_covar=tensor([0.0069, 0.0108, 0.0092, 0.0079, 0.0053, 0.0054, 0.0091, 0.0056], device='cuda:2'), out_proj_covar=tensor([1.2647e-04, 1.9528e-04, 1.7583e-04, 1.5131e-04, 9.4826e-05, 1.0482e-04, 1.5939e-04, 1.0478e-04], device='cuda:2') 2023-04-27 17:22:09,011 INFO [train.py:904] (2/8) Epoch 2, batch 2050, loss[loss=0.2236, simple_loss=0.2995, pruned_loss=0.07382, over 17207.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3373, pruned_loss=0.1052, over 3300087.35 frames. ], batch size: 44, lr: 3.02e-02, grad_scale: 16.0 2023-04-27 17:22:11,063 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:32,914 INFO [optim.py:368] (2/8) 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 17:22:58,999 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:20,403 INFO [train.py:904] (2/8) Epoch 2, batch 2100, loss[loss=0.2258, simple_loss=0.3062, pruned_loss=0.07268, over 17132.00 frames. ], tot_loss[loss=0.2732, simple_loss=0.3366, pruned_loss=0.1049, over 3305835.96 frames. ], batch size: 47, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:23:53,734 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:15,599 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:28,734 INFO [train.py:904] (2/8) Epoch 2, batch 2150, loss[loss=0.3056, simple_loss=0.3508, pruned_loss=0.1302, over 16464.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3379, pruned_loss=0.1064, over 3311475.68 frames. ], batch size: 146, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:24:33,513 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:51,834 INFO [optim.py:368] (2/8) 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 17:24:56,124 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0377, 3.3666, 3.2843, 1.2670, 3.5251, 3.4600, 2.9730, 2.9489], device='cuda:2'), covar=tensor([0.0929, 0.0130, 0.0205, 0.1925, 0.0108, 0.0090, 0.0390, 0.0339], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0075, 0.0076, 0.0147, 0.0069, 0.0066, 0.0093, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:25:22,226 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:23,337 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:38,424 INFO [train.py:904] (2/8) Epoch 2, batch 2200, loss[loss=0.3828, simple_loss=0.4107, pruned_loss=0.1774, over 12397.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3382, pruned_loss=0.1064, over 3309046.06 frames. ], batch size: 247, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:25:44,410 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:56,971 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-27 17:26:13,016 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3428, 4.0692, 4.1158, 4.2014, 3.7592, 4.1964, 3.9639, 3.9114], device='cuda:2'), covar=tensor([0.0239, 0.0170, 0.0167, 0.0135, 0.0662, 0.0181, 0.0337, 0.0206], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0089, 0.0160, 0.0129, 0.0196, 0.0130, 0.0114, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:26:21,486 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0218, 2.1084, 1.6376, 1.8579, 2.6461, 2.7430, 2.9071, 2.7056], device='cuda:2'), covar=tensor([0.0079, 0.0183, 0.0158, 0.0161, 0.0086, 0.0118, 0.0064, 0.0099], device='cuda:2'), in_proj_covar=tensor([0.0044, 0.0087, 0.0080, 0.0084, 0.0075, 0.0088, 0.0049, 0.0061], device='cuda:2'), out_proj_covar=tensor([7.2252e-05, 1.3404e-04, 1.2354e-04, 1.3216e-04, 1.2108e-04, 1.4251e-04, 7.8505e-05, 1.0373e-04], device='cuda:2') 2023-04-27 17:26:27,532 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-04-27 17:26:48,778 INFO [train.py:904] (2/8) Epoch 2, batch 2250, loss[loss=0.2928, simple_loss=0.3395, pruned_loss=0.1231, over 16885.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3385, pruned_loss=0.107, over 3298207.95 frames. ], batch size: 116, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:26:51,418 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:26:51,633 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8586, 2.1319, 1.7031, 1.6634, 2.6139, 2.5769, 2.7847, 2.6830], device='cuda:2'), covar=tensor([0.0049, 0.0128, 0.0120, 0.0155, 0.0062, 0.0092, 0.0045, 0.0064], device='cuda:2'), in_proj_covar=tensor([0.0043, 0.0085, 0.0078, 0.0083, 0.0074, 0.0086, 0.0049, 0.0059], device='cuda:2'), out_proj_covar=tensor([7.0788e-05, 1.3199e-04, 1.2116e-04, 1.3096e-04, 1.1927e-04, 1.3987e-04, 7.7564e-05, 1.0095e-04], device='cuda:2') 2023-04-27 17:27:12,081 INFO [optim.py:368] (2/8) 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 17:27:13,948 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 17:27:32,340 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:51,303 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:57,827 INFO [train.py:904] (2/8) Epoch 2, batch 2300, loss[loss=0.2445, simple_loss=0.3253, pruned_loss=0.08181, over 17083.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3393, pruned_loss=0.1072, over 3302191.69 frames. ], batch size: 47, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:28:06,374 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2533, 4.9979, 4.9359, 4.4852, 4.9935, 2.3934, 4.7000, 5.0278], device='cuda:2'), covar=tensor([0.0050, 0.0056, 0.0056, 0.0247, 0.0042, 0.0940, 0.0056, 0.0080], device='cuda:2'), in_proj_covar=tensor([0.0064, 0.0056, 0.0079, 0.0102, 0.0060, 0.0104, 0.0075, 0.0079], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:28:19,479 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-27 17:28:26,533 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:06,509 INFO [train.py:904] (2/8) Epoch 2, batch 2350, loss[loss=0.3012, simple_loss=0.3535, pruned_loss=0.1244, over 16729.00 frames. ], tot_loss[loss=0.2776, simple_loss=0.3397, pruned_loss=0.1078, over 3308524.98 frames. ], batch size: 89, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:29:07,832 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:15,023 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:31,565 INFO [optim.py:368] (2/8) 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 17:29:33,872 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:30:14,655 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:30:15,611 INFO [train.py:904] (2/8) Epoch 2, batch 2400, loss[loss=0.316, simple_loss=0.3716, pruned_loss=0.1302, over 15489.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3389, pruned_loss=0.1067, over 3315503.35 frames. ], batch size: 190, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:30:52,442 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:02,841 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 17:31:08,428 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 17:31:26,491 INFO [train.py:904] (2/8) Epoch 2, batch 2450, loss[loss=0.2594, simple_loss=0.3217, pruned_loss=0.09856, over 16778.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.3385, pruned_loss=0.1058, over 3312596.34 frames. ], batch size: 39, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:31:31,790 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:48,576 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9125, 4.1074, 3.3349, 3.6735, 3.0750, 2.3923, 4.3962, 4.9620], device='cuda:2'), covar=tensor([0.1539, 0.0475, 0.0830, 0.0361, 0.1555, 0.1080, 0.0211, 0.0089], device='cuda:2'), in_proj_covar=tensor([0.0245, 0.0226, 0.0229, 0.0160, 0.0256, 0.0177, 0.0174, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:31:51,020 INFO [optim.py:368] (2/8) 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 17:31:59,228 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:20,233 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:31,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9992, 2.8607, 2.2679, 3.3201, 3.2110, 3.2911, 1.8520, 2.7278], device='cuda:2'), covar=tensor([0.1244, 0.0292, 0.1113, 0.0105, 0.0189, 0.0284, 0.0975, 0.0516], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0107, 0.0161, 0.0067, 0.0095, 0.0104, 0.0145, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 17:32:35,566 INFO [train.py:904] (2/8) Epoch 2, batch 2500, loss[loss=0.2606, simple_loss=0.3366, pruned_loss=0.09232, over 17119.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.3391, pruned_loss=0.1059, over 3302457.12 frames. ], batch size: 48, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:32:36,844 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:04,217 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 17:33:27,071 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:43,145 INFO [train.py:904] (2/8) Epoch 2, batch 2550, loss[loss=0.2722, simple_loss=0.3419, pruned_loss=0.1013, over 16533.00 frames. ], tot_loss[loss=0.2754, simple_loss=0.339, pruned_loss=0.1059, over 3312485.67 frames. ], batch size: 68, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:34:02,569 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-04-27 17:34:08,155 INFO [optim.py:368] (2/8) 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 17:34:27,584 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:53,761 INFO [train.py:904] (2/8) Epoch 2, batch 2600, loss[loss=0.2771, simple_loss=0.328, pruned_loss=0.1131, over 16807.00 frames. ], tot_loss[loss=0.275, simple_loss=0.339, pruned_loss=0.1055, over 3314834.93 frames. ], batch size: 124, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:35:08,901 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:32,885 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:01,522 INFO [train.py:904] (2/8) Epoch 2, batch 2650, loss[loss=0.2997, simple_loss=0.3531, pruned_loss=0.1231, over 16341.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.3394, pruned_loss=0.1044, over 3315790.63 frames. ], batch size: 165, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:36:03,293 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:27,046 INFO [optim.py:368] (2/8) 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 17:36:33,936 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 17:36:54,733 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 17:36:57,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1938, 1.7441, 2.2478, 2.5750, 3.2896, 3.2508, 1.8952, 3.3203], device='cuda:2'), covar=tensor([0.0064, 0.0347, 0.0196, 0.0184, 0.0053, 0.0064, 0.0208, 0.0061], device='cuda:2'), in_proj_covar=tensor([0.0070, 0.0107, 0.0090, 0.0080, 0.0054, 0.0054, 0.0087, 0.0053], device='cuda:2'), out_proj_covar=tensor([1.2981e-04, 1.9344e-04, 1.7280e-04, 1.5362e-04, 9.8199e-05, 1.0411e-04, 1.5222e-04, 1.0033e-04], device='cuda:2') 2023-04-27 17:37:09,359 INFO [train.py:904] (2/8) Epoch 2, batch 2700, loss[loss=0.308, simple_loss=0.3521, pruned_loss=0.1319, over 16740.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3399, pruned_loss=0.1041, over 3314500.58 frames. ], batch size: 134, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:19,507 INFO [train.py:904] (2/8) Epoch 2, batch 2750, loss[loss=0.2808, simple_loss=0.3576, pruned_loss=0.102, over 17250.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3396, pruned_loss=0.1029, over 3318477.25 frames. ], batch size: 52, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:41,652 INFO [optim.py:368] (2/8) 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 17:39:02,066 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 17:39:26,263 INFO [train.py:904] (2/8) Epoch 2, batch 2800, loss[loss=0.2722, simple_loss=0.3439, pruned_loss=0.1003, over 16616.00 frames. ], tot_loss[loss=0.2752, simple_loss=0.3407, pruned_loss=0.1048, over 3315618.96 frames. ], batch size: 75, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:39:49,782 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-27 17:39:56,572 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-27 17:40:33,619 INFO [train.py:904] (2/8) Epoch 2, batch 2850, loss[loss=0.2704, simple_loss=0.3241, pruned_loss=0.1083, over 16812.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3387, pruned_loss=0.1038, over 3321384.47 frames. ], batch size: 102, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:40:57,333 INFO [optim.py:368] (2/8) 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 17:41:38,785 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-04-27 17:41:41,130 INFO [train.py:904] (2/8) Epoch 2, batch 2900, loss[loss=0.3441, simple_loss=0.3698, pruned_loss=0.1593, over 15596.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3382, pruned_loss=0.1055, over 3320032.84 frames. ], batch size: 190, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:41:57,813 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 17:42:30,747 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2134, 4.0182, 4.0905, 4.1833, 3.5358, 4.1442, 3.9088, 3.8465], device='cuda:2'), covar=tensor([0.0262, 0.0189, 0.0203, 0.0143, 0.0830, 0.0174, 0.0437, 0.0250], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0094, 0.0168, 0.0134, 0.0205, 0.0137, 0.0117, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:42:49,008 INFO [train.py:904] (2/8) Epoch 2, batch 2950, loss[loss=0.2911, simple_loss=0.3574, pruned_loss=0.1124, over 16670.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.3363, pruned_loss=0.1052, over 3313894.39 frames. ], batch size: 62, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:42:50,482 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:13,803 INFO [optim.py:368] (2/8) 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 17:43:14,147 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:43:53,665 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 2, batch 3000, loss[loss=0.2307, simple_loss=0.3016, pruned_loss=0.07994, over 16837.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.338, pruned_loss=0.1069, over 3302825.26 frames. ], batch size: 42, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:43:54,563 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 17:44:03,911 INFO [train.py:938] (2/8) Epoch 2, validation: loss=0.1858, simple_loss=0.2917, pruned_loss=0.04, over 944034.00 frames. 2023-04-27 17:44:03,912 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 17:45:09,406 INFO [train.py:904] (2/8) Epoch 2, batch 3050, loss[loss=0.2425, simple_loss=0.3148, pruned_loss=0.08514, over 17223.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3373, pruned_loss=0.106, over 3312474.67 frames. ], batch size: 44, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:45:33,139 INFO [optim.py:368] (2/8) 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 17:46:15,659 INFO [train.py:904] (2/8) Epoch 2, batch 3100, loss[loss=0.2552, simple_loss=0.3148, pruned_loss=0.09778, over 16762.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3372, pruned_loss=0.1063, over 3301439.46 frames. ], batch size: 83, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:46:44,782 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 17:47:05,209 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5715, 4.5435, 2.0133, 4.6253, 2.8385, 4.4027, 2.1716, 3.0579], device='cuda:2'), covar=tensor([0.0026, 0.0073, 0.1400, 0.0035, 0.0671, 0.0292, 0.1195, 0.0590], device='cuda:2'), in_proj_covar=tensor([0.0080, 0.0100, 0.0168, 0.0078, 0.0150, 0.0130, 0.0170, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 17:47:22,152 INFO [train.py:904] (2/8) Epoch 2, batch 3150, loss[loss=0.2446, simple_loss=0.3135, pruned_loss=0.08784, over 17180.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3349, pruned_loss=0.1052, over 3308337.58 frames. ], batch size: 46, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:47:26,091 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8913, 4.2744, 3.5091, 3.5905, 3.1925, 2.1948, 4.4122, 5.1444], device='cuda:2'), covar=tensor([0.1880, 0.0550, 0.1037, 0.0489, 0.1934, 0.1331, 0.0304, 0.0104], device='cuda:2'), in_proj_covar=tensor([0.0238, 0.0224, 0.0231, 0.0156, 0.0254, 0.0178, 0.0182, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:47:44,588 INFO [optim.py:368] (2/8) 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 17:48:28,284 INFO [train.py:904] (2/8) Epoch 2, batch 3200, loss[loss=0.2415, simple_loss=0.2994, pruned_loss=0.09174, over 16775.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3328, pruned_loss=0.1031, over 3317532.72 frames. ], batch size: 39, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:34,225 INFO [train.py:904] (2/8) Epoch 2, batch 3250, loss[loss=0.2899, simple_loss=0.349, pruned_loss=0.1154, over 16580.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3324, pruned_loss=0.1028, over 3317697.82 frames. ], batch size: 68, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:55,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6839, 2.8311, 2.5929, 3.9969, 2.1942, 3.7896, 2.2905, 2.5256], device='cuda:2'), covar=tensor([0.0287, 0.0349, 0.0272, 0.0174, 0.1271, 0.0174, 0.0649, 0.0909], device='cuda:2'), in_proj_covar=tensor([0.0187, 0.0159, 0.0134, 0.0189, 0.0240, 0.0150, 0.0167, 0.0220], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:49:58,154 INFO [optim.py:368] (2/8) 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 17:49:58,512 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:50:03,244 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 17:50:42,665 INFO [train.py:904] (2/8) Epoch 2, batch 3300, loss[loss=0.2739, simple_loss=0.3387, pruned_loss=0.1046, over 16741.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3324, pruned_loss=0.1023, over 3322384.89 frames. ], batch size: 102, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:51:02,652 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:51:48,084 INFO [train.py:904] (2/8) Epoch 2, batch 3350, loss[loss=0.2501, simple_loss=0.3286, pruned_loss=0.08577, over 17195.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3336, pruned_loss=0.103, over 3315504.81 frames. ], batch size: 46, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:52:13,284 INFO [optim.py:368] (2/8) 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 17:52:26,160 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-27 17:52:56,206 INFO [train.py:904] (2/8) Epoch 2, batch 3400, loss[loss=0.2111, simple_loss=0.284, pruned_loss=0.06913, over 16952.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3335, pruned_loss=0.1025, over 3314406.94 frames. ], batch size: 41, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:53:04,661 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6288, 4.3995, 4.0913, 1.8238, 4.4869, 4.6387, 3.5976, 3.4400], device='cuda:2'), covar=tensor([0.0741, 0.0071, 0.0222, 0.1549, 0.0067, 0.0037, 0.0252, 0.0297], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0084, 0.0080, 0.0150, 0.0076, 0.0072, 0.0099, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-27 17:54:05,265 INFO [train.py:904] (2/8) Epoch 2, batch 3450, loss[loss=0.3259, simple_loss=0.3596, pruned_loss=0.1461, over 16269.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.3312, pruned_loss=0.1018, over 3311521.68 frames. ], batch size: 145, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:29,794 INFO [optim.py:368] (2/8) 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 17:55:11,512 INFO [train.py:904] (2/8) Epoch 2, batch 3500, loss[loss=0.2867, simple_loss=0.3368, pruned_loss=0.1183, over 16755.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3287, pruned_loss=0.09981, over 3317694.46 frames. ], batch size: 134, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:55:40,473 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4797, 4.3853, 4.2387, 3.7480, 4.3255, 2.3750, 4.1838, 4.3580], device='cuda:2'), covar=tensor([0.0061, 0.0048, 0.0066, 0.0311, 0.0052, 0.0860, 0.0064, 0.0088], device='cuda:2'), in_proj_covar=tensor([0.0068, 0.0058, 0.0084, 0.0109, 0.0065, 0.0106, 0.0080, 0.0086], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:56:21,450 INFO [train.py:904] (2/8) Epoch 2, batch 3550, loss[loss=0.2881, simple_loss=0.338, pruned_loss=0.119, over 16468.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.328, pruned_loss=0.09935, over 3309877.39 frames. ], batch size: 146, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:56:45,024 INFO [optim.py:368] (2/8) 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 17:56:53,882 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7830, 4.6741, 4.4326, 1.7193, 3.5965, 2.6106, 4.2519, 4.6076], device='cuda:2'), covar=tensor([0.0222, 0.0324, 0.0305, 0.2017, 0.0602, 0.1121, 0.0554, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0111, 0.0145, 0.0155, 0.0149, 0.0140, 0.0149, 0.0101], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 17:57:11,483 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:57:29,798 INFO [train.py:904] (2/8) Epoch 2, batch 3600, loss[loss=0.2484, simple_loss=0.3077, pruned_loss=0.09451, over 16924.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3261, pruned_loss=0.09807, over 3313559.71 frames. ], batch size: 41, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:57:52,772 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7321, 1.7256, 2.2378, 2.3712, 2.7445, 2.5280, 1.6793, 2.6041], device='cuda:2'), covar=tensor([0.0038, 0.0236, 0.0114, 0.0110, 0.0043, 0.0084, 0.0200, 0.0039], device='cuda:2'), in_proj_covar=tensor([0.0071, 0.0111, 0.0095, 0.0085, 0.0058, 0.0059, 0.0093, 0.0055], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-27 17:58:04,145 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5384, 3.9680, 4.2196, 3.6994, 4.1918, 4.2869, 4.4463, 2.2539], device='cuda:2'), covar=tensor([0.0562, 0.0062, 0.0050, 0.0213, 0.0052, 0.0055, 0.0039, 0.0528], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0048, 0.0057, 0.0101, 0.0052, 0.0055, 0.0059, 0.0101], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 17:58:37,347 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:58:40,990 INFO [train.py:904] (2/8) Epoch 2, batch 3650, loss[loss=0.2364, simple_loss=0.298, pruned_loss=0.08739, over 16740.00 frames. ], tot_loss[loss=0.2615, simple_loss=0.3254, pruned_loss=0.09886, over 3308985.60 frames. ], batch size: 102, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 17:59:08,134 INFO [optim.py:368] (2/8) 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 17:59:54,876 INFO [train.py:904] (2/8) Epoch 2, batch 3700, loss[loss=0.2669, simple_loss=0.3177, pruned_loss=0.1081, over 16474.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3233, pruned_loss=0.09956, over 3292471.08 frames. ], batch size: 146, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 18:00:24,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6797, 3.5430, 3.6773, 3.6176, 3.7170, 4.0207, 3.9584, 3.6466], device='cuda:2'), covar=tensor([0.1242, 0.1414, 0.1024, 0.1605, 0.1920, 0.0965, 0.0756, 0.1696], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0267, 0.0243, 0.0228, 0.0287, 0.0248, 0.0199, 0.0297], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:00:30,912 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9100, 3.0257, 2.4564, 4.2723, 2.2687, 3.9695, 2.4649, 2.4186], device='cuda:2'), covar=tensor([0.0243, 0.0366, 0.0333, 0.0172, 0.1272, 0.0161, 0.0658, 0.1071], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0165, 0.0138, 0.0197, 0.0244, 0.0153, 0.0172, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:00:55,916 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0435, 4.0727, 4.1108, 4.0643, 4.0521, 4.6093, 4.4560, 4.1072], device='cuda:2'), covar=tensor([0.1332, 0.1208, 0.1101, 0.1745, 0.2273, 0.0836, 0.0735, 0.1928], device='cuda:2'), in_proj_covar=tensor([0.0187, 0.0275, 0.0248, 0.0234, 0.0294, 0.0253, 0.0203, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:01:07,709 INFO [train.py:904] (2/8) Epoch 2, batch 3750, loss[loss=0.2959, simple_loss=0.3524, pruned_loss=0.1198, over 15668.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.3242, pruned_loss=0.1014, over 3285175.91 frames. ], batch size: 191, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:01:18,060 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 18:01:33,975 INFO [optim.py:368] (2/8) 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 18:01:34,509 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7483, 3.7610, 1.9035, 3.7623, 2.6236, 3.7461, 1.9847, 2.8157], device='cuda:2'), covar=tensor([0.0059, 0.0178, 0.1559, 0.0064, 0.0655, 0.0370, 0.1491, 0.0606], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0101, 0.0168, 0.0079, 0.0153, 0.0128, 0.0174, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 18:02:24,613 INFO [train.py:904] (2/8) Epoch 2, batch 3800, loss[loss=0.2645, simple_loss=0.3195, pruned_loss=0.1048, over 16746.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3258, pruned_loss=0.1027, over 3289012.84 frames. ], batch size: 83, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:03:00,461 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-27 18:03:40,207 INFO [train.py:904] (2/8) Epoch 2, batch 3850, loss[loss=0.2352, simple_loss=0.3035, pruned_loss=0.08347, over 16613.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3252, pruned_loss=0.1027, over 3287976.02 frames. ], batch size: 57, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:04:07,044 INFO [optim.py:368] (2/8) 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 18:04:07,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2419, 3.8316, 3.7722, 1.7340, 3.1767, 2.1320, 3.7642, 4.0398], device='cuda:2'), covar=tensor([0.0234, 0.0470, 0.0305, 0.1794, 0.0612, 0.1028, 0.0504, 0.0324], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0109, 0.0148, 0.0157, 0.0148, 0.0139, 0.0151, 0.0099], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-27 18:04:23,928 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8645, 4.3857, 4.5754, 1.8793, 3.4061, 2.6028, 4.2001, 4.7497], device='cuda:2'), covar=tensor([0.0192, 0.0329, 0.0237, 0.1725, 0.0647, 0.0953, 0.0593, 0.0319], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0107, 0.0147, 0.0155, 0.0147, 0.0139, 0.0149, 0.0098], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-27 18:04:53,515 INFO [train.py:904] (2/8) Epoch 2, batch 3900, loss[loss=0.2743, simple_loss=0.3355, pruned_loss=0.1065, over 15735.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3236, pruned_loss=0.102, over 3281993.51 frames. ], batch size: 190, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:05:25,206 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:05:55,289 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:06:05,397 INFO [train.py:904] (2/8) Epoch 2, batch 3950, loss[loss=0.2695, simple_loss=0.3212, pruned_loss=0.1089, over 16413.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.3231, pruned_loss=0.1024, over 3269991.69 frames. ], batch size: 146, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:06:08,863 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 18:06:34,023 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:06:44,872 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-27 18:06:52,450 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:07:17,633 INFO [train.py:904] (2/8) Epoch 2, batch 4000, loss[loss=0.284, simple_loss=0.3299, pruned_loss=0.119, over 16777.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3221, pruned_loss=0.1016, over 3279567.14 frames. ], batch size: 134, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:07:37,710 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:07:42,237 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0050, 4.8719, 4.6104, 3.6170, 4.7699, 2.0760, 4.5913, 4.6562], device='cuda:2'), covar=tensor([0.0069, 0.0056, 0.0090, 0.0460, 0.0064, 0.1464, 0.0071, 0.0135], device='cuda:2'), in_proj_covar=tensor([0.0064, 0.0056, 0.0079, 0.0101, 0.0063, 0.0105, 0.0075, 0.0080], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:08:02,161 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:08:31,014 INFO [train.py:904] (2/8) Epoch 2, batch 4050, loss[loss=0.2871, simple_loss=0.3428, pruned_loss=0.1157, over 12060.00 frames. ], tot_loss[loss=0.2597, simple_loss=0.321, pruned_loss=0.09923, over 3257773.31 frames. ], batch size: 247, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:08:47,340 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 18:08:56,314 INFO [optim.py:368] (2/8) 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 18:08:57,981 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8343, 1.7337, 1.5418, 1.6240, 2.5109, 2.3226, 2.6822, 2.6819], device='cuda:2'), covar=tensor([0.0016, 0.0164, 0.0161, 0.0192, 0.0067, 0.0116, 0.0028, 0.0043], device='cuda:2'), in_proj_covar=tensor([0.0046, 0.0091, 0.0089, 0.0093, 0.0083, 0.0090, 0.0054, 0.0066], device='cuda:2'), out_proj_covar=tensor([7.9443e-05, 1.4306e-04, 1.3829e-04, 1.5063e-04, 1.3750e-04, 1.4659e-04, 8.9314e-05, 1.1218e-04], device='cuda:2') 2023-04-27 18:09:39,961 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 18:09:43,226 INFO [train.py:904] (2/8) Epoch 2, batch 4100, loss[loss=0.2872, simple_loss=0.3556, pruned_loss=0.1094, over 16443.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3208, pruned_loss=0.09721, over 3261444.21 frames. ], batch size: 146, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:10:19,744 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4939, 4.0388, 3.8659, 1.7333, 3.0787, 2.5987, 3.7459, 4.2091], device='cuda:2'), covar=tensor([0.0190, 0.0391, 0.0349, 0.1792, 0.0673, 0.0882, 0.0642, 0.0401], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0105, 0.0148, 0.0156, 0.0147, 0.0138, 0.0151, 0.0098], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-27 18:10:32,370 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1702, 3.6084, 3.6490, 1.6457, 3.9778, 3.8892, 3.2782, 3.1113], device='cuda:2'), covar=tensor([0.0874, 0.0116, 0.0173, 0.1532, 0.0061, 0.0045, 0.0244, 0.0301], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0080, 0.0076, 0.0151, 0.0073, 0.0071, 0.0102, 0.0114], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:10:59,418 INFO [train.py:904] (2/8) Epoch 2, batch 4150, loss[loss=0.2852, simple_loss=0.3571, pruned_loss=0.1067, over 16958.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3297, pruned_loss=0.1017, over 3247105.50 frames. ], batch size: 96, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:10:59,827 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6649, 4.8663, 4.4993, 4.7326, 4.2750, 4.2809, 4.4675, 4.9073], device='cuda:2'), covar=tensor([0.0445, 0.0578, 0.0959, 0.0331, 0.0502, 0.0495, 0.0400, 0.0501], device='cuda:2'), in_proj_covar=tensor([0.0222, 0.0293, 0.0262, 0.0184, 0.0204, 0.0179, 0.0237, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:11:25,252 INFO [optim.py:368] (2/8) 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 18:11:30,393 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1533, 2.9869, 2.8374, 1.8111, 2.6155, 2.0124, 2.7876, 3.0578], device='cuda:2'), covar=tensor([0.0303, 0.0477, 0.0367, 0.1577, 0.0600, 0.0998, 0.0648, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0103, 0.0148, 0.0156, 0.0147, 0.0138, 0.0150, 0.0097], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-27 18:12:14,210 INFO [train.py:904] (2/8) Epoch 2, batch 4200, loss[loss=0.2945, simple_loss=0.3639, pruned_loss=0.1125, over 16256.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3391, pruned_loss=0.1054, over 3216485.69 frames. ], batch size: 165, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:15,500 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:26,255 INFO [train.py:904] (2/8) Epoch 2, batch 4250, loss[loss=0.2321, simple_loss=0.3166, pruned_loss=0.07381, over 16669.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3421, pruned_loss=0.1058, over 3196149.56 frames. ], batch size: 134, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:44,211 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:52,836 INFO [optim.py:368] (2/8) 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 18:14:05,688 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:24,081 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:38,545 INFO [train.py:904] (2/8) Epoch 2, batch 4300, loss[loss=0.2809, simple_loss=0.3588, pruned_loss=0.1015, over 16615.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3428, pruned_loss=0.1044, over 3187461.27 frames. ], batch size: 68, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:14:49,275 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:10,462 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:12,686 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:15:29,912 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:34,032 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 18:15:49,381 INFO [train.py:904] (2/8) Epoch 2, batch 4350, loss[loss=0.2854, simple_loss=0.358, pruned_loss=0.1064, over 16484.00 frames. ], tot_loss[loss=0.2797, simple_loss=0.3469, pruned_loss=0.1063, over 3195617.18 frames. ], batch size: 146, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:16:13,768 INFO [optim.py:368] (2/8) 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 18:16:56,174 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:16:59,398 INFO [train.py:904] (2/8) Epoch 2, batch 4400, loss[loss=0.3339, simple_loss=0.3874, pruned_loss=0.1402, over 12090.00 frames. ], tot_loss[loss=0.2828, simple_loss=0.3493, pruned_loss=0.1081, over 3173394.44 frames. ], batch size: 248, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:18:09,487 INFO [train.py:904] (2/8) Epoch 2, batch 4450, loss[loss=0.2902, simple_loss=0.3687, pruned_loss=0.1059, over 16887.00 frames. ], tot_loss[loss=0.2825, simple_loss=0.3512, pruned_loss=0.1069, over 3187955.63 frames. ], batch size: 109, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:18:34,420 INFO [optim.py:368] (2/8) 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 18:19:17,771 INFO [train.py:904] (2/8) Epoch 2, batch 4500, loss[loss=0.2786, simple_loss=0.3563, pruned_loss=0.1005, over 16860.00 frames. ], tot_loss[loss=0.2803, simple_loss=0.3497, pruned_loss=0.1055, over 3198915.72 frames. ], batch size: 116, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:20:28,458 INFO [train.py:904] (2/8) Epoch 2, batch 4550, loss[loss=0.2815, simple_loss=0.3579, pruned_loss=0.1025, over 17138.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3506, pruned_loss=0.1065, over 3189603.11 frames. ], batch size: 47, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:20:54,773 INFO [optim.py:368] (2/8) 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 18:21:08,325 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:25,018 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:43,176 INFO [train.py:904] (2/8) Epoch 2, batch 4600, loss[loss=0.2491, simple_loss=0.3365, pruned_loss=0.0809, over 16848.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3496, pruned_loss=0.1047, over 3198017.26 frames. ], batch size: 102, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:21:47,478 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 18:21:53,307 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:09,370 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8001, 4.0247, 2.3973, 5.2085, 4.8285, 3.8996, 2.5301, 3.3022], device='cuda:2'), covar=tensor([0.1878, 0.0378, 0.1657, 0.0056, 0.0116, 0.0254, 0.1206, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0114, 0.0164, 0.0063, 0.0096, 0.0105, 0.0148, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 18:22:10,141 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,543 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,625 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:22:54,487 INFO [train.py:904] (2/8) Epoch 2, batch 4650, loss[loss=0.2614, simple_loss=0.3379, pruned_loss=0.09247, over 16718.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3475, pruned_loss=0.1029, over 3219860.07 frames. ], batch size: 57, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:22:55,049 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:03,204 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:20,765 INFO [optim.py:368] (2/8) 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 18:23:27,861 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:23:49,061 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0612, 2.5656, 2.1869, 3.4810, 2.0622, 3.3766, 2.2436, 2.2855], device='cuda:2'), covar=tensor([0.0325, 0.0385, 0.0332, 0.0213, 0.1268, 0.0158, 0.0622, 0.1091], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0172, 0.0144, 0.0201, 0.0259, 0.0157, 0.0178, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:23:54,902 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:24:06,744 INFO [train.py:904] (2/8) Epoch 2, batch 4700, loss[loss=0.2523, simple_loss=0.3365, pruned_loss=0.08402, over 16772.00 frames. ], tot_loss[loss=0.2746, simple_loss=0.3454, pruned_loss=0.1019, over 3216710.38 frames. ], batch size: 116, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:20,207 INFO [train.py:904] (2/8) Epoch 2, batch 4750, loss[loss=0.2472, simple_loss=0.3232, pruned_loss=0.08554, over 16694.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3399, pruned_loss=0.09895, over 3233862.34 frames. ], batch size: 89, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:29,597 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5934, 2.7813, 2.4234, 4.0039, 2.2190, 3.9745, 2.5151, 2.3917], device='cuda:2'), covar=tensor([0.0302, 0.0410, 0.0326, 0.0179, 0.1320, 0.0131, 0.0602, 0.1247], device='cuda:2'), in_proj_covar=tensor([0.0201, 0.0171, 0.0143, 0.0198, 0.0254, 0.0157, 0.0177, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:25:45,828 INFO [optim.py:368] (2/8) 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 18:26:30,650 INFO [train.py:904] (2/8) Epoch 2, batch 4800, loss[loss=0.2994, simple_loss=0.3661, pruned_loss=0.1163, over 15237.00 frames. ], tot_loss[loss=0.266, simple_loss=0.3372, pruned_loss=0.09739, over 3217085.11 frames. ], batch size: 190, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:26:48,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6812, 3.6089, 3.9527, 4.0611, 4.0412, 3.7715, 3.7177, 3.8740], device='cuda:2'), covar=tensor([0.0226, 0.0275, 0.0384, 0.0343, 0.0357, 0.0204, 0.0663, 0.0282], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0140, 0.0162, 0.0155, 0.0185, 0.0150, 0.0231, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 18:27:43,242 INFO [train.py:904] (2/8) Epoch 2, batch 4850, loss[loss=0.3134, simple_loss=0.3642, pruned_loss=0.1313, over 12259.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3393, pruned_loss=0.09818, over 3204918.76 frames. ], batch size: 247, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:27:43,998 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 18:27:54,003 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1885, 3.1327, 3.2636, 3.4794, 3.4628, 3.2389, 3.3714, 3.4640], device='cuda:2'), covar=tensor([0.0391, 0.0414, 0.0727, 0.0303, 0.0337, 0.1211, 0.0590, 0.0260], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0256, 0.0355, 0.0263, 0.0208, 0.0195, 0.0197, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:28:11,112 INFO [optim.py:368] (2/8) 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 18:28:58,606 INFO [train.py:904] (2/8) Epoch 2, batch 4900, loss[loss=0.2628, simple_loss=0.3334, pruned_loss=0.09606, over 16482.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3391, pruned_loss=0.09727, over 3192867.46 frames. ], batch size: 68, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:29:24,594 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:04,914 INFO [zipformer.py:625] (2/8) 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] (2/8) attn_weights_entropy = tensor([3.8077, 3.8879, 4.2084, 4.1819, 4.2247, 3.8632, 3.8844, 3.9760], device='cuda:2'), covar=tensor([0.0220, 0.0225, 0.0268, 0.0323, 0.0279, 0.0214, 0.0566, 0.0281], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0141, 0.0164, 0.0158, 0.0184, 0.0151, 0.0229, 0.0146], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 18:30:10,700 INFO [train.py:904] (2/8) Epoch 2, batch 4950, loss[loss=0.2769, simple_loss=0.3496, pruned_loss=0.1021, over 16678.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3405, pruned_loss=0.09781, over 3187620.13 frames. ], batch size: 134, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:30:34,538 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:37,411 INFO [optim.py:368] (2/8) 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 18:31:12,911 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:31:22,555 INFO [train.py:904] (2/8) Epoch 2, batch 5000, loss[loss=0.2532, simple_loss=0.3247, pruned_loss=0.09083, over 16767.00 frames. ], tot_loss[loss=0.2684, simple_loss=0.341, pruned_loss=0.09796, over 3190941.02 frames. ], batch size: 39, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:32:09,620 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 18:32:21,457 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:26,443 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:34,517 INFO [train.py:904] (2/8) Epoch 2, batch 5050, loss[loss=0.2749, simple_loss=0.3482, pruned_loss=0.1008, over 16431.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.3406, pruned_loss=0.09712, over 3207796.42 frames. ], batch size: 146, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:00,585 INFO [optim.py:368] (2/8) 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 18:33:45,520 INFO [train.py:904] (2/8) Epoch 2, batch 5100, loss[loss=0.2964, simple_loss=0.3501, pruned_loss=0.1214, over 12156.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3387, pruned_loss=0.09592, over 3202771.06 frames. ], batch size: 247, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:53,359 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:34:49,197 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6271, 3.5063, 3.5729, 3.9047, 3.9236, 3.6141, 3.8764, 3.8790], device='cuda:2'), covar=tensor([0.0479, 0.0504, 0.1100, 0.0366, 0.0360, 0.0797, 0.0377, 0.0307], device='cuda:2'), in_proj_covar=tensor([0.0230, 0.0264, 0.0366, 0.0270, 0.0211, 0.0194, 0.0201, 0.0214], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:34:56,468 INFO [train.py:904] (2/8) Epoch 2, batch 5150, loss[loss=0.2531, simple_loss=0.334, pruned_loss=0.08612, over 17032.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3393, pruned_loss=0.09602, over 3180481.61 frames. ], batch size: 55, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:35:22,257 INFO [optim.py:368] (2/8) 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 18:36:10,019 INFO [train.py:904] (2/8) Epoch 2, batch 5200, loss[loss=0.2625, simple_loss=0.3302, pruned_loss=0.09738, over 16704.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3382, pruned_loss=0.09569, over 3188022.21 frames. ], batch size: 134, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:37:10,556 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-27 18:37:16,322 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 2, batch 5250, loss[loss=0.2438, simple_loss=0.3153, pruned_loss=0.08611, over 16494.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3351, pruned_loss=0.09512, over 3202812.22 frames. ], batch size: 68, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:37:48,866 INFO [optim.py:368] (2/8) 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 18:38:09,474 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-04-27 18:38:23,203 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:38:32,783 INFO [train.py:904] (2/8) Epoch 2, batch 5300, loss[loss=0.2187, simple_loss=0.2898, pruned_loss=0.07376, over 16704.00 frames. ], tot_loss[loss=0.26, simple_loss=0.3321, pruned_loss=0.0939, over 3195042.28 frames. ], batch size: 124, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:38:52,879 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3347, 4.1331, 4.0896, 4.2538, 3.3773, 4.1397, 4.1397, 3.8085], device='cuda:2'), covar=tensor([0.0296, 0.0266, 0.0247, 0.0159, 0.0969, 0.0214, 0.0293, 0.0300], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0094, 0.0160, 0.0123, 0.0190, 0.0131, 0.0110, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:39:10,719 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1414, 3.8245, 3.7146, 4.3251, 4.3391, 4.0248, 4.4095, 4.1926], device='cuda:2'), covar=tensor([0.0463, 0.0559, 0.1307, 0.0490, 0.0501, 0.0492, 0.0315, 0.0505], device='cuda:2'), in_proj_covar=tensor([0.0244, 0.0279, 0.0382, 0.0285, 0.0222, 0.0200, 0.0212, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:39:25,848 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8395, 3.7916, 1.5850, 3.7403, 2.4027, 3.8310, 1.9458, 2.8213], device='cuda:2'), covar=tensor([0.0031, 0.0097, 0.1675, 0.0042, 0.0743, 0.0188, 0.1319, 0.0610], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0097, 0.0164, 0.0076, 0.0153, 0.0122, 0.0170, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 18:39:43,334 INFO [train.py:904] (2/8) Epoch 2, batch 5350, loss[loss=0.2508, simple_loss=0.3247, pruned_loss=0.08846, over 16807.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3297, pruned_loss=0.09256, over 3194436.32 frames. ], batch size: 124, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:40:09,981 INFO [optim.py:368] (2/8) 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 18:40:15,938 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4521, 2.8015, 2.3894, 3.7617, 1.8530, 3.6176, 2.2532, 2.2592], device='cuda:2'), covar=tensor([0.0343, 0.0469, 0.0375, 0.0244, 0.1683, 0.0230, 0.0736, 0.1245], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0170, 0.0144, 0.0198, 0.0248, 0.0159, 0.0177, 0.0230], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:40:44,861 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8498, 4.5489, 4.6085, 4.6652, 4.0965, 4.6039, 4.6220, 4.3194], device='cuda:2'), covar=tensor([0.0223, 0.0187, 0.0161, 0.0103, 0.0757, 0.0165, 0.0141, 0.0208], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0095, 0.0159, 0.0124, 0.0189, 0.0131, 0.0109, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:40:56,836 INFO [train.py:904] (2/8) Epoch 2, batch 5400, loss[loss=0.3359, simple_loss=0.3888, pruned_loss=0.1415, over 12012.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3334, pruned_loss=0.09429, over 3193732.01 frames. ], batch size: 246, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:40:57,190 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:42:11,924 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-27 18:42:14,663 INFO [train.py:904] (2/8) Epoch 2, batch 5450, loss[loss=0.315, simple_loss=0.3692, pruned_loss=0.1305, over 16771.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3369, pruned_loss=0.09622, over 3199005.79 frames. ], batch size: 76, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:42:43,047 INFO [optim.py:368] (2/8) 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 18:42:57,087 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0736, 1.4178, 2.1700, 2.7626, 2.9172, 3.1395, 1.8093, 2.9257], device='cuda:2'), covar=tensor([0.0042, 0.0292, 0.0132, 0.0095, 0.0028, 0.0042, 0.0153, 0.0036], device='cuda:2'), in_proj_covar=tensor([0.0073, 0.0110, 0.0092, 0.0083, 0.0058, 0.0058, 0.0091, 0.0055], device='cuda:2'), out_proj_covar=tensor([1.2889e-04, 1.9554e-04, 1.7096e-04, 1.5686e-04, 9.8628e-05, 1.0542e-04, 1.5715e-04, 9.8362e-05], device='cuda:2') 2023-04-27 18:43:21,906 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7726, 4.7448, 4.5539, 4.0175, 4.6896, 1.9371, 4.4384, 4.5983], device='cuda:2'), covar=tensor([0.0054, 0.0038, 0.0049, 0.0271, 0.0040, 0.1227, 0.0056, 0.0069], device='cuda:2'), in_proj_covar=tensor([0.0062, 0.0051, 0.0076, 0.0099, 0.0060, 0.0109, 0.0070, 0.0076], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:43:34,026 INFO [train.py:904] (2/8) Epoch 2, batch 5500, loss[loss=0.3187, simple_loss=0.3821, pruned_loss=0.1277, over 16886.00 frames. ], tot_loss[loss=0.2792, simple_loss=0.3479, pruned_loss=0.1053, over 3169512.56 frames. ], batch size: 96, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:43:34,904 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1574, 2.4941, 1.9285, 2.2671, 3.1053, 2.8802, 3.9522, 3.4117], device='cuda:2'), covar=tensor([0.0006, 0.0117, 0.0152, 0.0147, 0.0060, 0.0140, 0.0020, 0.0050], device='cuda:2'), in_proj_covar=tensor([0.0041, 0.0096, 0.0095, 0.0100, 0.0085, 0.0098, 0.0054, 0.0069], device='cuda:2'), out_proj_covar=tensor([6.1251e-05, 1.5171e-04, 1.4595e-04, 1.6014e-04, 1.4042e-04, 1.6166e-04, 8.7282e-05, 1.1579e-04], device='cuda:2') 2023-04-27 18:44:17,503 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5694, 3.0138, 2.6159, 2.4384, 2.2003, 2.0603, 2.9999, 3.2278], device='cuda:2'), covar=tensor([0.1041, 0.0479, 0.0752, 0.0460, 0.1762, 0.0972, 0.0297, 0.0163], device='cuda:2'), in_proj_covar=tensor([0.0252, 0.0227, 0.0242, 0.0176, 0.0280, 0.0182, 0.0197, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:44:41,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6780, 3.2755, 3.2359, 2.3019, 3.1573, 3.2300, 3.2572, 1.5329], device='cuda:2'), covar=tensor([0.0625, 0.0035, 0.0053, 0.0333, 0.0057, 0.0075, 0.0043, 0.0527], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0051, 0.0057, 0.0106, 0.0051, 0.0055, 0.0057, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:44:50,702 INFO [train.py:904] (2/8) Epoch 2, batch 5550, loss[loss=0.2738, simple_loss=0.3532, pruned_loss=0.09719, over 16714.00 frames. ], tot_loss[loss=0.293, simple_loss=0.3582, pruned_loss=0.1139, over 3149360.72 frames. ], batch size: 83, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:45:19,404 INFO [optim.py:368] (2/8) 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 18:45:55,387 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=7.05 vs. limit=5.0 2023-04-27 18:46:11,044 INFO [train.py:904] (2/8) Epoch 2, batch 5600, loss[loss=0.4363, simple_loss=0.444, pruned_loss=0.2142, over 10917.00 frames. ], tot_loss[loss=0.304, simple_loss=0.3654, pruned_loss=0.1213, over 3108723.39 frames. ], batch size: 247, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:47:34,290 INFO [train.py:904] (2/8) Epoch 2, batch 5650, loss[loss=0.4321, simple_loss=0.4483, pruned_loss=0.2079, over 11092.00 frames. ], tot_loss[loss=0.3154, simple_loss=0.3735, pruned_loss=0.1286, over 3070048.30 frames. ], batch size: 246, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:01,991 INFO [optim.py:368] (2/8) 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 18:48:07,014 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0317, 3.0831, 3.1583, 1.6663, 3.3257, 3.2904, 2.6746, 2.6201], device='cuda:2'), covar=tensor([0.1004, 0.0143, 0.0179, 0.1500, 0.0094, 0.0074, 0.0359, 0.0442], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0082, 0.0077, 0.0154, 0.0074, 0.0072, 0.0106, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 18:48:53,334 INFO [train.py:904] (2/8) Epoch 2, batch 5700, loss[loss=0.3163, simple_loss=0.3815, pruned_loss=0.1256, over 16644.00 frames. ], tot_loss[loss=0.3178, simple_loss=0.3753, pruned_loss=0.1302, over 3067521.78 frames. ], batch size: 134, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:53,704 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:49:28,396 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7247, 2.6225, 1.6027, 2.6224, 2.1679, 2.6477, 1.9559, 2.3977], device='cuda:2'), covar=tensor([0.0076, 0.0232, 0.1373, 0.0072, 0.0666, 0.0521, 0.1091, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0072, 0.0101, 0.0169, 0.0075, 0.0160, 0.0127, 0.0174, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 18:49:44,687 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:12,115 INFO [train.py:904] (2/8) Epoch 2, batch 5750, loss[loss=0.3408, simple_loss=0.3759, pruned_loss=0.1529, over 11370.00 frames. ], tot_loss[loss=0.3214, simple_loss=0.3788, pruned_loss=0.132, over 3063249.77 frames. ], batch size: 247, lr: 2.69e-02, grad_scale: 8.0 2023-04-27 18:50:42,007 INFO [optim.py:368] (2/8) 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 18:51:22,881 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:51:33,504 INFO [train.py:904] (2/8) Epoch 2, batch 5800, loss[loss=0.2644, simple_loss=0.3419, pruned_loss=0.09345, over 16695.00 frames. ], tot_loss[loss=0.3174, simple_loss=0.3768, pruned_loss=0.1291, over 3060237.80 frames. ], batch size: 89, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:56,020 INFO [train.py:904] (2/8) Epoch 2, batch 5850, loss[loss=0.3364, simple_loss=0.3889, pruned_loss=0.142, over 15447.00 frames. ], tot_loss[loss=0.3129, simple_loss=0.3734, pruned_loss=0.1262, over 3076608.15 frames. ], batch size: 191, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:57,355 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3940, 4.0938, 4.1409, 4.2190, 3.6798, 4.1778, 4.1592, 3.9176], device='cuda:2'), covar=tensor([0.0267, 0.0208, 0.0171, 0.0123, 0.0793, 0.0191, 0.0249, 0.0266], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0088, 0.0146, 0.0113, 0.0177, 0.0121, 0.0103, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 18:52:57,405 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 18:53:55,060 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-27 18:54:18,254 INFO [train.py:904] (2/8) Epoch 2, batch 5900, loss[loss=0.2718, simple_loss=0.3534, pruned_loss=0.09506, over 16235.00 frames. ], tot_loss[loss=0.3102, simple_loss=0.3717, pruned_loss=0.1244, over 3095852.14 frames. ], batch size: 35, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:54:39,928 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:17,943 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:42,173 INFO [train.py:904] (2/8) Epoch 2, batch 5950, loss[loss=0.294, simple_loss=0.3621, pruned_loss=0.113, over 16769.00 frames. ], tot_loss[loss=0.3084, simple_loss=0.3722, pruned_loss=0.1223, over 3104445.22 frames. ], batch size: 124, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:56:13,606 INFO [optim.py:368] (2/8) 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 18:56:56,132 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:01,489 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:04,567 INFO [train.py:904] (2/8) Epoch 2, batch 6000, loss[loss=0.2911, simple_loss=0.3522, pruned_loss=0.115, over 16955.00 frames. ], tot_loss[loss=0.3091, simple_loss=0.3721, pruned_loss=0.123, over 3081367.52 frames. ], batch size: 109, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:57:04,567 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 18:57:15,919 INFO [train.py:938] (2/8) Epoch 2, validation: loss=0.2302, simple_loss=0.3372, pruned_loss=0.06166, over 944034.00 frames. 2023-04-27 18:57:15,919 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 18:57:21,104 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 18:58:34,798 INFO [train.py:904] (2/8) Epoch 2, batch 6050, loss[loss=0.27, simple_loss=0.3478, pruned_loss=0.09606, over 16353.00 frames. ], tot_loss[loss=0.307, simple_loss=0.3699, pruned_loss=0.122, over 3072786.31 frames. ], batch size: 146, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 18:58:48,637 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16210.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:04,137 INFO [optim.py:368] (2/8) 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 18:59:23,100 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6141, 1.8649, 1.4415, 1.5577, 2.3318, 2.1930, 2.5299, 2.5598], device='cuda:2'), covar=tensor([0.0016, 0.0151, 0.0183, 0.0193, 0.0081, 0.0129, 0.0036, 0.0064], device='cuda:2'), in_proj_covar=tensor([0.0040, 0.0095, 0.0097, 0.0100, 0.0085, 0.0096, 0.0053, 0.0070], device='cuda:2'), out_proj_covar=tensor([5.8607e-05, 1.4828e-04, 1.4833e-04, 1.5846e-04, 1.4043e-04, 1.5626e-04, 8.5857e-05, 1.1629e-04], device='cuda:2') 2023-04-27 18:59:33,561 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:52,054 INFO [train.py:904] (2/8) Epoch 2, batch 6100, loss[loss=0.3261, simple_loss=0.3835, pruned_loss=0.1343, over 16274.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3679, pruned_loss=0.1195, over 3098573.81 frames. ], batch size: 165, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:01:17,145 INFO [train.py:904] (2/8) Epoch 2, batch 6150, loss[loss=0.2512, simple_loss=0.3215, pruned_loss=0.09047, over 17264.00 frames. ], tot_loss[loss=0.302, simple_loss=0.3662, pruned_loss=0.1189, over 3086366.39 frames. ], batch size: 52, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:01:22,673 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-27 19:01:45,849 INFO [optim.py:368] (2/8) 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 19:02:34,532 INFO [train.py:904] (2/8) Epoch 2, batch 6200, loss[loss=0.2473, simple_loss=0.3288, pruned_loss=0.08285, over 16817.00 frames. ], tot_loss[loss=0.3003, simple_loss=0.3641, pruned_loss=0.1182, over 3097927.69 frames. ], batch size: 102, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:02:45,547 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:06,831 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:17,716 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6729, 2.9762, 2.2411, 4.1002, 2.0895, 4.0605, 2.4653, 2.3098], device='cuda:2'), covar=tensor([0.0287, 0.0364, 0.0354, 0.0171, 0.1311, 0.0140, 0.0611, 0.1049], device='cuda:2'), in_proj_covar=tensor([0.0207, 0.0177, 0.0148, 0.0206, 0.0258, 0.0161, 0.0180, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:03:24,684 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7761, 2.6621, 1.6610, 2.7126, 2.1984, 2.6209, 1.8972, 2.4125], device='cuda:2'), covar=tensor([0.0072, 0.0205, 0.1240, 0.0061, 0.0613, 0.0472, 0.1002, 0.0487], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0105, 0.0173, 0.0076, 0.0161, 0.0134, 0.0178, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 19:03:29,524 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3169, 4.3301, 4.9130, 4.8849, 4.8030, 4.3186, 4.4423, 4.3751], device='cuda:2'), covar=tensor([0.0232, 0.0255, 0.0272, 0.0286, 0.0342, 0.0231, 0.0628, 0.0298], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0149, 0.0170, 0.0163, 0.0196, 0.0162, 0.0247, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 19:03:51,369 INFO [train.py:904] (2/8) Epoch 2, batch 6250, loss[loss=0.3399, simple_loss=0.3941, pruned_loss=0.1429, over 15553.00 frames. ], tot_loss[loss=0.2995, simple_loss=0.3637, pruned_loss=0.1177, over 3120714.74 frames. ], batch size: 191, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:04:18,705 INFO [optim.py:368] (2/8) 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 19:04:22,869 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-27 19:04:37,352 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:51,246 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:57,154 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6211, 4.4593, 4.3487, 3.6891, 4.3372, 1.7307, 4.1426, 4.4170], device='cuda:2'), covar=tensor([0.0052, 0.0049, 0.0057, 0.0278, 0.0056, 0.1350, 0.0058, 0.0084], device='cuda:2'), in_proj_covar=tensor([0.0061, 0.0049, 0.0074, 0.0095, 0.0057, 0.0105, 0.0068, 0.0075], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:05:05,085 INFO [train.py:904] (2/8) Epoch 2, batch 6300, loss[loss=0.3472, simple_loss=0.4023, pruned_loss=0.146, over 16882.00 frames. ], tot_loss[loss=0.2979, simple_loss=0.363, pruned_loss=0.1163, over 3127167.19 frames. ], batch size: 90, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:22,078 INFO [train.py:904] (2/8) Epoch 2, batch 6350, loss[loss=0.2659, simple_loss=0.3341, pruned_loss=0.09882, over 16247.00 frames. ], tot_loss[loss=0.301, simple_loss=0.365, pruned_loss=0.1185, over 3123371.50 frames. ], batch size: 35, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:29,788 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:06:48,149 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7542, 2.6237, 1.7272, 2.7394, 2.1898, 2.7615, 1.8792, 2.2819], device='cuda:2'), covar=tensor([0.0077, 0.0221, 0.1041, 0.0063, 0.0572, 0.0384, 0.1056, 0.0531], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0103, 0.0166, 0.0074, 0.0157, 0.0132, 0.0175, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 19:06:52,065 INFO [optim.py:368] (2/8) 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 19:07:13,301 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9231, 3.8140, 4.3726, 4.3477, 4.3511, 3.9502, 4.0930, 4.0281], device='cuda:2'), covar=tensor([0.0257, 0.0296, 0.0300, 0.0339, 0.0348, 0.0229, 0.0582, 0.0290], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0149, 0.0168, 0.0165, 0.0196, 0.0161, 0.0246, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-27 19:07:21,377 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:07:39,358 INFO [train.py:904] (2/8) Epoch 2, batch 6400, loss[loss=0.2987, simple_loss=0.355, pruned_loss=0.1212, over 16267.00 frames. ], tot_loss[loss=0.304, simple_loss=0.3663, pruned_loss=0.1208, over 3101953.02 frames. ], batch size: 165, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:08:34,558 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:08:55,789 INFO [train.py:904] (2/8) Epoch 2, batch 6450, loss[loss=0.2757, simple_loss=0.3522, pruned_loss=0.09966, over 16884.00 frames. ], tot_loss[loss=0.3011, simple_loss=0.3647, pruned_loss=0.1187, over 3100558.94 frames. ], batch size: 96, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:09:25,708 INFO [optim.py:368] (2/8) 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 19:10:13,782 INFO [train.py:904] (2/8) Epoch 2, batch 6500, loss[loss=0.2973, simple_loss=0.3599, pruned_loss=0.1173, over 16675.00 frames. ], tot_loss[loss=0.2984, simple_loss=0.3618, pruned_loss=0.1175, over 3088603.58 frames. ], batch size: 134, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:10:24,341 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:06,985 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3724, 2.5614, 2.3152, 3.8082, 1.9423, 3.5108, 2.2222, 2.2886], device='cuda:2'), covar=tensor([0.0281, 0.0455, 0.0355, 0.0165, 0.1448, 0.0209, 0.0683, 0.1030], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0180, 0.0150, 0.0208, 0.0259, 0.0164, 0.0182, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:11:13,055 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6852, 3.8714, 3.0600, 2.6585, 2.9349, 2.0882, 3.9079, 4.3474], device='cuda:2'), covar=tensor([0.1825, 0.0560, 0.1031, 0.0790, 0.1766, 0.1249, 0.0328, 0.0170], device='cuda:2'), in_proj_covar=tensor([0.0257, 0.0225, 0.0237, 0.0178, 0.0273, 0.0181, 0.0199, 0.0139], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:11:16,434 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4400, 1.5332, 1.9046, 2.4135, 2.4904, 2.3299, 1.4564, 2.4625], device='cuda:2'), covar=tensor([0.0034, 0.0196, 0.0113, 0.0082, 0.0037, 0.0073, 0.0167, 0.0033], device='cuda:2'), in_proj_covar=tensor([0.0069, 0.0109, 0.0095, 0.0082, 0.0061, 0.0059, 0.0095, 0.0053], device='cuda:2'), out_proj_covar=tensor([1.2089e-04, 1.9172e-04, 1.7309e-04, 1.5082e-04, 1.0473e-04, 1.0547e-04, 1.6319e-04, 9.2472e-05], device='cuda:2') 2023-04-27 19:11:32,119 INFO [train.py:904] (2/8) Epoch 2, batch 6550, loss[loss=0.3365, simple_loss=0.3827, pruned_loss=0.1451, over 11805.00 frames. ], tot_loss[loss=0.3011, simple_loss=0.3653, pruned_loss=0.1185, over 3096384.46 frames. ], batch size: 247, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:11:37,689 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:48,569 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-27 19:11:59,595 INFO [optim.py:368] (2/8) 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 19:12:11,158 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:32,953 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:47,616 INFO [train.py:904] (2/8) Epoch 2, batch 6600, loss[loss=0.365, simple_loss=0.4006, pruned_loss=0.1647, over 11548.00 frames. ], tot_loss[loss=0.3036, simple_loss=0.3681, pruned_loss=0.1195, over 3106895.93 frames. ], batch size: 246, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:13:21,306 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9948, 3.1233, 3.4363, 3.4225, 3.4357, 3.1628, 3.2576, 3.3227], device='cuda:2'), covar=tensor([0.0247, 0.0299, 0.0310, 0.0351, 0.0306, 0.0275, 0.0537, 0.0249], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0150, 0.0166, 0.0168, 0.0196, 0.0165, 0.0253, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-27 19:13:36,987 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 19:13:46,766 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:06,684 INFO [train.py:904] (2/8) Epoch 2, batch 6650, loss[loss=0.281, simple_loss=0.3531, pruned_loss=0.1044, over 16685.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3688, pruned_loss=0.1212, over 3092995.53 frames. ], batch size: 134, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:14:13,187 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:26,123 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1636, 2.9990, 2.8676, 1.8461, 2.5601, 2.1015, 2.7753, 3.1064], device='cuda:2'), covar=tensor([0.0292, 0.0343, 0.0344, 0.1321, 0.0567, 0.0818, 0.0478, 0.0267], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0106, 0.0152, 0.0154, 0.0150, 0.0138, 0.0149, 0.0101], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 19:14:35,748 INFO [optim.py:368] (2/8) 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 19:15:04,743 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-27 19:15:21,370 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 19:15:23,513 INFO [train.py:904] (2/8) Epoch 2, batch 6700, loss[loss=0.3638, simple_loss=0.3961, pruned_loss=0.1657, over 11911.00 frames. ], tot_loss[loss=0.3046, simple_loss=0.3674, pruned_loss=0.1209, over 3097338.93 frames. ], batch size: 248, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:15:26,028 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 19:15:26,818 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:15:50,006 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:15:58,571 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2480, 5.1270, 5.2468, 5.6203, 5.6187, 4.9683, 5.6041, 5.4988], device='cuda:2'), covar=tensor([0.0473, 0.0367, 0.0723, 0.0239, 0.0348, 0.0345, 0.0323, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0244, 0.0274, 0.0382, 0.0287, 0.0224, 0.0197, 0.0221, 0.0224], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:15:58,998 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 19:16:23,537 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3882, 3.3794, 3.3357, 2.7129, 3.3352, 1.8961, 3.0927, 3.1352], device='cuda:2'), covar=tensor([0.0094, 0.0071, 0.0074, 0.0290, 0.0069, 0.1331, 0.0093, 0.0147], device='cuda:2'), in_proj_covar=tensor([0.0063, 0.0052, 0.0077, 0.0097, 0.0058, 0.0111, 0.0070, 0.0077], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:16:38,378 INFO [train.py:904] (2/8) Epoch 2, batch 6750, loss[loss=0.2533, simple_loss=0.3316, pruned_loss=0.08744, over 16459.00 frames. ], tot_loss[loss=0.3062, simple_loss=0.3679, pruned_loss=0.1223, over 3092987.44 frames. ], batch size: 146, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:16:47,224 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:07,555 INFO [optim.py:368] (2/8) 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 19:17:22,255 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:17:53,311 INFO [train.py:904] (2/8) Epoch 2, batch 6800, loss[loss=0.2835, simple_loss=0.3568, pruned_loss=0.1051, over 16647.00 frames. ], tot_loss[loss=0.303, simple_loss=0.3657, pruned_loss=0.1201, over 3103097.25 frames. ], batch size: 134, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:18:18,723 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:19:03,149 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-27 19:19:10,216 INFO [train.py:904] (2/8) Epoch 2, batch 6850, loss[loss=0.2759, simple_loss=0.3627, pruned_loss=0.09461, over 17146.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.3671, pruned_loss=0.1209, over 3098756.54 frames. ], batch size: 46, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:19:38,485 INFO [optim.py:368] (2/8) 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 19:19:49,777 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:20:24,230 INFO [train.py:904] (2/8) Epoch 2, batch 6900, loss[loss=0.3009, simple_loss=0.3761, pruned_loss=0.1128, over 16729.00 frames. ], tot_loss[loss=0.3031, simple_loss=0.3685, pruned_loss=0.1189, over 3124410.14 frames. ], batch size: 89, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:21:01,218 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:40,735 INFO [train.py:904] (2/8) Epoch 2, batch 6950, loss[loss=0.27, simple_loss=0.3438, pruned_loss=0.09814, over 16460.00 frames. ], tot_loss[loss=0.3066, simple_loss=0.371, pruned_loss=0.1211, over 3140039.96 frames. ], batch size: 75, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:22:09,949 INFO [optim.py:368] (2/8) 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 19:22:54,958 INFO [train.py:904] (2/8) Epoch 2, batch 7000, loss[loss=0.3067, simple_loss=0.3787, pruned_loss=0.1173, over 16474.00 frames. ], tot_loss[loss=0.3064, simple_loss=0.3714, pruned_loss=0.1207, over 3122912.46 frames. ], batch size: 146, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:23:19,289 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 19:23:53,087 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:10,869 INFO [train.py:904] (2/8) Epoch 2, batch 7050, loss[loss=0.2824, simple_loss=0.3558, pruned_loss=0.1045, over 16580.00 frames. ], tot_loss[loss=0.3069, simple_loss=0.372, pruned_loss=0.121, over 3114869.32 frames. ], batch size: 76, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:24:19,682 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-27 19:24:37,000 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5436, 3.4240, 3.6732, 3.9155, 3.8970, 3.5204, 3.8815, 3.8734], device='cuda:2'), covar=tensor([0.0576, 0.0503, 0.0943, 0.0381, 0.0393, 0.0836, 0.0366, 0.0345], device='cuda:2'), in_proj_covar=tensor([0.0237, 0.0272, 0.0373, 0.0276, 0.0223, 0.0198, 0.0215, 0.0222], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:24:37,722 INFO [optim.py:368] (2/8) 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 19:24:43,958 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:24:49,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3527, 3.3028, 1.4403, 3.4306, 2.2295, 3.3843, 1.8988, 2.5158], device='cuda:2'), covar=tensor([0.0055, 0.0252, 0.1731, 0.0043, 0.0822, 0.0356, 0.1288, 0.0661], device='cuda:2'), in_proj_covar=tensor([0.0073, 0.0106, 0.0170, 0.0072, 0.0162, 0.0137, 0.0176, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 19:24:56,694 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5587, 2.7638, 2.3636, 4.0584, 1.9164, 3.8528, 2.3851, 2.3218], device='cuda:2'), covar=tensor([0.0288, 0.0445, 0.0349, 0.0169, 0.1545, 0.0171, 0.0662, 0.1111], device='cuda:2'), in_proj_covar=tensor([0.0212, 0.0183, 0.0156, 0.0213, 0.0263, 0.0170, 0.0185, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:25:22,999 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:23,663 INFO [train.py:904] (2/8) Epoch 2, batch 7100, loss[loss=0.2734, simple_loss=0.3474, pruned_loss=0.09974, over 16768.00 frames. ], tot_loss[loss=0.3058, simple_loss=0.3701, pruned_loss=0.1208, over 3098570.93 frames. ], batch size: 89, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:25:40,648 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:25:48,338 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 19:26:38,667 INFO [train.py:904] (2/8) Epoch 2, batch 7150, loss[loss=0.3405, simple_loss=0.3749, pruned_loss=0.153, over 11417.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3674, pruned_loss=0.1198, over 3082312.88 frames. ], batch size: 247, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:27:07,390 INFO [optim.py:368] (2/8) 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 19:27:52,977 INFO [train.py:904] (2/8) Epoch 2, batch 7200, loss[loss=0.2529, simple_loss=0.3334, pruned_loss=0.08619, over 16731.00 frames. ], tot_loss[loss=0.2982, simple_loss=0.3636, pruned_loss=0.1163, over 3099019.46 frames. ], batch size: 57, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:28:10,346 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0923, 3.9145, 4.1400, 4.4490, 4.4822, 3.9803, 4.5099, 4.3667], device='cuda:2'), covar=tensor([0.0576, 0.0567, 0.0961, 0.0327, 0.0344, 0.0634, 0.0270, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0241, 0.0275, 0.0380, 0.0283, 0.0224, 0.0205, 0.0217, 0.0228], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:28:10,395 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:28:44,455 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:29:13,592 INFO [train.py:904] (2/8) Epoch 2, batch 7250, loss[loss=0.2366, simple_loss=0.3189, pruned_loss=0.07718, over 16839.00 frames. ], tot_loss[loss=0.2949, simple_loss=0.3605, pruned_loss=0.1147, over 3092354.60 frames. ], batch size: 102, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:29:42,538 INFO [optim.py:368] (2/8) 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 19:29:47,288 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:30:19,160 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:30:29,420 INFO [train.py:904] (2/8) Epoch 2, batch 7300, loss[loss=0.2681, simple_loss=0.3419, pruned_loss=0.0972, over 17022.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3605, pruned_loss=0.1153, over 3078650.95 frames. ], batch size: 53, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:30:59,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0051, 4.7728, 4.7527, 4.8440, 4.3462, 4.8651, 4.7018, 4.5839], device='cuda:2'), covar=tensor([0.0154, 0.0092, 0.0120, 0.0078, 0.0440, 0.0098, 0.0122, 0.0141], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0091, 0.0144, 0.0115, 0.0167, 0.0122, 0.0102, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:31:42,675 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 19:31:46,484 INFO [train.py:904] (2/8) Epoch 2, batch 7350, loss[loss=0.3492, simple_loss=0.3832, pruned_loss=0.1576, over 11325.00 frames. ], tot_loss[loss=0.2964, simple_loss=0.3609, pruned_loss=0.116, over 3063521.23 frames. ], batch size: 248, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:31:58,154 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 19:32:16,421 INFO [optim.py:368] (2/8) 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 19:32:22,414 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:32:31,181 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 19:32:56,901 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:33:06,149 INFO [train.py:904] (2/8) Epoch 2, batch 7400, loss[loss=0.3605, simple_loss=0.3932, pruned_loss=0.1639, over 11116.00 frames. ], tot_loss[loss=0.298, simple_loss=0.3624, pruned_loss=0.1168, over 3067557.91 frames. ], batch size: 248, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:33:25,350 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:33:40,918 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:33:52,562 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 19:34:17,271 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8570, 4.0284, 3.7566, 3.9098, 3.4503, 3.7156, 3.7332, 3.9899], device='cuda:2'), covar=tensor([0.0483, 0.0599, 0.0749, 0.0344, 0.0528, 0.0635, 0.0475, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0219, 0.0303, 0.0279, 0.0192, 0.0203, 0.0190, 0.0243, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:34:27,189 INFO [train.py:904] (2/8) Epoch 2, batch 7450, loss[loss=0.2746, simple_loss=0.3553, pruned_loss=0.09692, over 16754.00 frames. ], tot_loss[loss=0.3002, simple_loss=0.3636, pruned_loss=0.1183, over 3066699.18 frames. ], batch size: 83, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:34:43,753 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 19:35:43,387 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6191, 4.4008, 4.3561, 1.9917, 4.5788, 4.6481, 3.4482, 3.6510], device='cuda:2'), covar=tensor([0.0700, 0.0077, 0.0133, 0.1398, 0.0042, 0.0032, 0.0241, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0084, 0.0082, 0.0150, 0.0075, 0.0070, 0.0105, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:35:49,110 INFO [train.py:904] (2/8) Epoch 2, batch 7500, loss[loss=0.3545, simple_loss=0.3933, pruned_loss=0.1578, over 11767.00 frames. ], tot_loss[loss=0.3011, simple_loss=0.3647, pruned_loss=0.1188, over 3047172.12 frames. ], batch size: 248, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:36:03,730 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 19:37:05,019 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:05,674 INFO [train.py:904] (2/8) Epoch 2, batch 7550, loss[loss=0.3081, simple_loss=0.3782, pruned_loss=0.119, over 16742.00 frames. ], tot_loss[loss=0.3014, simple_loss=0.3644, pruned_loss=0.1192, over 3046250.03 frames. ], batch size: 124, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:37:32,585 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:34,648 INFO [optim.py:368] (2/8) 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 19:37:50,573 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:56,789 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3331, 3.3578, 2.7872, 2.6511, 2.7013, 2.0755, 3.5014, 4.0168], device='cuda:2'), covar=tensor([0.2018, 0.0734, 0.1076, 0.0665, 0.1681, 0.1213, 0.0388, 0.0194], device='cuda:2'), in_proj_covar=tensor([0.0257, 0.0227, 0.0240, 0.0183, 0.0278, 0.0182, 0.0202, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:38:04,540 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:38:22,208 INFO [train.py:904] (2/8) Epoch 2, batch 7600, loss[loss=0.328, simple_loss=0.3901, pruned_loss=0.1329, over 15324.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3643, pruned_loss=0.1202, over 3024451.19 frames. ], batch size: 190, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:38:37,835 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:38:58,561 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 19:39:24,535 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:39:27,838 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-27 19:39:39,705 INFO [train.py:904] (2/8) Epoch 2, batch 7650, loss[loss=0.2981, simple_loss=0.3599, pruned_loss=0.1182, over 16740.00 frames. ], tot_loss[loss=0.3037, simple_loss=0.3653, pruned_loss=0.1211, over 3040411.06 frames. ], batch size: 83, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:40:10,726 INFO [optim.py:368] (2/8) 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 19:40:13,817 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8524, 3.2633, 2.4103, 4.1849, 4.0854, 3.9135, 1.6609, 3.0520], device='cuda:2'), covar=tensor([0.1301, 0.0330, 0.1145, 0.0057, 0.0128, 0.0223, 0.1196, 0.0540], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0122, 0.0164, 0.0067, 0.0108, 0.0117, 0.0152, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 19:40:49,701 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:40:56,260 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3622, 5.0811, 5.0739, 5.2375, 4.5268, 5.0979, 4.9577, 4.7545], device='cuda:2'), covar=tensor([0.0210, 0.0167, 0.0141, 0.0073, 0.0662, 0.0226, 0.0141, 0.0198], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0096, 0.0148, 0.0119, 0.0177, 0.0128, 0.0108, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 19:40:58,874 INFO [train.py:904] (2/8) Epoch 2, batch 7700, loss[loss=0.2906, simple_loss=0.3584, pruned_loss=0.1114, over 16247.00 frames. ], tot_loss[loss=0.3054, simple_loss=0.3661, pruned_loss=0.1223, over 3037693.33 frames. ], batch size: 165, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:42:04,290 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:16,828 INFO [train.py:904] (2/8) Epoch 2, batch 7750, loss[loss=0.3564, simple_loss=0.3922, pruned_loss=0.1603, over 11760.00 frames. ], tot_loss[loss=0.3047, simple_loss=0.366, pruned_loss=0.1216, over 3047302.06 frames. ], batch size: 247, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:42:40,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8003, 4.0462, 3.6943, 3.8996, 3.5323, 3.6673, 3.7684, 3.9045], device='cuda:2'), covar=tensor([0.0517, 0.0669, 0.0844, 0.0371, 0.0527, 0.0736, 0.0483, 0.0806], device='cuda:2'), in_proj_covar=tensor([0.0223, 0.0309, 0.0283, 0.0195, 0.0206, 0.0199, 0.0250, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:42:46,602 INFO [optim.py:368] (2/8) 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 19:43:18,006 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2309, 2.6184, 2.0955, 3.5289, 1.9890, 3.3070, 2.3380, 2.1570], device='cuda:2'), covar=tensor([0.0286, 0.0389, 0.0344, 0.0198, 0.1285, 0.0182, 0.0614, 0.1025], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0187, 0.0157, 0.0217, 0.0264, 0.0170, 0.0189, 0.0246], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:43:31,902 INFO [train.py:904] (2/8) Epoch 2, batch 7800, loss[loss=0.301, simple_loss=0.376, pruned_loss=0.113, over 16919.00 frames. ], tot_loss[loss=0.3065, simple_loss=0.3672, pruned_loss=0.1229, over 3032885.55 frames. ], batch size: 109, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:43:47,037 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:44:52,420 INFO [train.py:904] (2/8) Epoch 2, batch 7850, loss[loss=0.2877, simple_loss=0.3629, pruned_loss=0.1062, over 16703.00 frames. ], tot_loss[loss=0.306, simple_loss=0.3682, pruned_loss=0.1219, over 3051237.54 frames. ], batch size: 89, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:45:18,548 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18018.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:45:21,252 INFO [optim.py:368] (2/8) 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 19:45:22,942 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:45:48,790 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:46:05,621 INFO [train.py:904] (2/8) Epoch 2, batch 7900, loss[loss=0.3448, simple_loss=0.3903, pruned_loss=0.1496, over 15336.00 frames. ], tot_loss[loss=0.3026, simple_loss=0.3657, pruned_loss=0.1197, over 3073487.97 frames. ], batch size: 190, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:46:13,984 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:46:29,400 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:03,307 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:05,057 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:47:25,404 INFO [train.py:904] (2/8) Epoch 2, batch 7950, loss[loss=0.3621, simple_loss=0.3997, pruned_loss=0.1623, over 11259.00 frames. ], tot_loss[loss=0.3033, simple_loss=0.3658, pruned_loss=0.1204, over 3061668.55 frames. ], batch size: 246, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:47:56,298 INFO [optim.py:368] (2/8) 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 19:48:00,524 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 19:48:42,376 INFO [train.py:904] (2/8) Epoch 2, batch 8000, loss[loss=0.3343, simple_loss=0.3924, pruned_loss=0.1381, over 16742.00 frames. ], tot_loss[loss=0.3032, simple_loss=0.3653, pruned_loss=0.1206, over 3051022.29 frames. ], batch size: 124, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:49:02,497 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9655, 3.2215, 2.5022, 4.3382, 2.0826, 4.1859, 2.8627, 2.4903], device='cuda:2'), covar=tensor([0.0245, 0.0403, 0.0369, 0.0151, 0.1392, 0.0147, 0.0561, 0.1164], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0186, 0.0159, 0.0217, 0.0262, 0.0170, 0.0189, 0.0246], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:49:16,801 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:18,070 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:56,233 INFO [train.py:904] (2/8) Epoch 2, batch 8050, loss[loss=0.2507, simple_loss=0.3186, pruned_loss=0.09136, over 16348.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3644, pruned_loss=0.1201, over 3044580.03 frames. ], batch size: 68, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:50:24,925 INFO [optim.py:368] (2/8) 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 19:50:46,727 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:48,690 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:57,363 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8565, 1.5001, 2.0960, 2.5167, 2.3498, 2.8446, 1.3103, 2.7563], device='cuda:2'), covar=tensor([0.0033, 0.0239, 0.0118, 0.0084, 0.0044, 0.0043, 0.0211, 0.0032], device='cuda:2'), in_proj_covar=tensor([0.0070, 0.0112, 0.0096, 0.0082, 0.0063, 0.0056, 0.0099, 0.0056], device='cuda:2'), out_proj_covar=tensor([1.1941e-04, 1.9323e-04, 1.7120e-04, 1.4787e-04, 1.0685e-04, 9.7543e-05, 1.6948e-04, 9.3599e-05], device='cuda:2') 2023-04-27 19:51:11,132 INFO [train.py:904] (2/8) Epoch 2, batch 8100, loss[loss=0.3307, simple_loss=0.3819, pruned_loss=0.1398, over 15542.00 frames. ], tot_loss[loss=0.301, simple_loss=0.3637, pruned_loss=0.1192, over 3053273.92 frames. ], batch size: 191, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:51:20,288 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 19:51:38,303 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2691, 3.8329, 3.9220, 1.5719, 4.0450, 4.0799, 3.2507, 2.9385], device='cuda:2'), covar=tensor([0.0784, 0.0098, 0.0135, 0.1513, 0.0066, 0.0057, 0.0228, 0.0401], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0088, 0.0084, 0.0150, 0.0076, 0.0069, 0.0109, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 19:51:52,861 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7827, 2.9463, 2.3208, 3.8109, 3.6653, 3.6732, 1.7897, 2.6607], device='cuda:2'), covar=tensor([0.1402, 0.0359, 0.1134, 0.0061, 0.0160, 0.0237, 0.1096, 0.0679], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0122, 0.0164, 0.0069, 0.0112, 0.0120, 0.0155, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 19:52:20,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4003, 3.3035, 3.3042, 2.9276, 3.2902, 2.2331, 3.1159, 3.0714], device='cuda:2'), covar=tensor([0.0052, 0.0042, 0.0064, 0.0183, 0.0044, 0.0820, 0.0055, 0.0076], device='cuda:2'), in_proj_covar=tensor([0.0063, 0.0053, 0.0080, 0.0101, 0.0060, 0.0112, 0.0071, 0.0078], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 19:52:29,017 INFO [train.py:904] (2/8) Epoch 2, batch 8150, loss[loss=0.2609, simple_loss=0.3322, pruned_loss=0.09477, over 16725.00 frames. ], tot_loss[loss=0.298, simple_loss=0.3612, pruned_loss=0.1174, over 3067240.45 frames. ], batch size: 83, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:52:52,884 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:53:00,707 INFO [optim.py:368] (2/8) 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 19:53:48,016 INFO [train.py:904] (2/8) Epoch 2, batch 8200, loss[loss=0.2841, simple_loss=0.3548, pruned_loss=0.1067, over 16914.00 frames. ], tot_loss[loss=0.2961, simple_loss=0.3589, pruned_loss=0.1167, over 3068346.50 frames. ], batch size: 109, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:53:57,302 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:46,289 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:10,271 INFO [train.py:904] (2/8) Epoch 2, batch 8250, loss[loss=0.2301, simple_loss=0.3202, pruned_loss=0.07001, over 16863.00 frames. ], tot_loss[loss=0.2927, simple_loss=0.3572, pruned_loss=0.1141, over 3045904.74 frames. ], batch size: 96, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:55:15,137 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:44,640 INFO [optim.py:368] (2/8) 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 19:55:59,221 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 19:56:01,339 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 19:56:05,977 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:26,196 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 19:56:33,571 INFO [train.py:904] (2/8) Epoch 2, batch 8300, loss[loss=0.2452, simple_loss=0.3082, pruned_loss=0.09108, over 12206.00 frames. ], tot_loss[loss=0.2842, simple_loss=0.3516, pruned_loss=0.1084, over 3042825.18 frames. ], batch size: 246, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:57:37,624 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:47,612 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 19:57:54,234 INFO [train.py:904] (2/8) Epoch 2, batch 8350, loss[loss=0.2498, simple_loss=0.3325, pruned_loss=0.08357, over 16388.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3485, pruned_loss=0.1041, over 3049947.25 frames. ], batch size: 146, lr: 2.50e-02, grad_scale: 4.0 2023-04-27 19:58:06,718 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:58:28,816 INFO [optim.py:368] (2/8) 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 19:58:37,980 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4437, 3.3710, 3.2445, 3.3861, 2.9891, 3.3920, 3.1244, 3.1353], device='cuda:2'), covar=tensor([0.0308, 0.0206, 0.0251, 0.0164, 0.0551, 0.0218, 0.1120, 0.0327], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0098, 0.0151, 0.0120, 0.0174, 0.0128, 0.0110, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 19:58:42,168 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:58:44,074 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:15,561 INFO [train.py:904] (2/8) Epoch 2, batch 8400, loss[loss=0.2609, simple_loss=0.3319, pruned_loss=0.09498, over 16375.00 frames. ], tot_loss[loss=0.2724, simple_loss=0.3439, pruned_loss=0.1005, over 3040565.84 frames. ], batch size: 68, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 19:59:16,076 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:45,489 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:00:35,047 INFO [train.py:904] (2/8) Epoch 2, batch 8450, loss[loss=0.2482, simple_loss=0.3302, pruned_loss=0.08314, over 16104.00 frames. ], tot_loss[loss=0.2692, simple_loss=0.3419, pruned_loss=0.09822, over 3057347.31 frames. ], batch size: 35, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 20:00:41,210 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6709, 3.3360, 3.2137, 2.4993, 3.1885, 3.1217, 3.3096, 1.5097], device='cuda:2'), covar=tensor([0.0501, 0.0025, 0.0043, 0.0230, 0.0038, 0.0043, 0.0026, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0053, 0.0057, 0.0105, 0.0052, 0.0058, 0.0058, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 20:00:44,303 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8799, 4.5834, 4.6604, 4.0540, 4.6288, 1.8864, 4.3165, 4.5810], device='cuda:2'), covar=tensor([0.0052, 0.0047, 0.0053, 0.0207, 0.0044, 0.1371, 0.0059, 0.0081], device='cuda:2'), in_proj_covar=tensor([0.0062, 0.0051, 0.0079, 0.0091, 0.0059, 0.0111, 0.0071, 0.0076], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:00:50,029 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4930, 5.7880, 5.5398, 5.6551, 5.2229, 4.7679, 5.4312, 5.8752], device='cuda:2'), covar=tensor([0.0507, 0.0527, 0.0710, 0.0276, 0.0397, 0.0460, 0.0334, 0.0469], device='cuda:2'), in_proj_covar=tensor([0.0217, 0.0302, 0.0269, 0.0192, 0.0202, 0.0194, 0.0248, 0.0215], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:01:00,268 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:01:09,358 INFO [optim.py:368] (2/8) 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 20:01:11,245 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5293, 3.5289, 1.4175, 3.5274, 2.3587, 3.6058, 1.6579, 2.6075], device='cuda:2'), covar=tensor([0.0050, 0.0145, 0.1831, 0.0040, 0.0816, 0.0231, 0.1625, 0.0666], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0109, 0.0172, 0.0073, 0.0153, 0.0135, 0.0177, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 20:01:55,557 INFO [train.py:904] (2/8) Epoch 2, batch 8500, loss[loss=0.2633, simple_loss=0.3339, pruned_loss=0.09632, over 15233.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3363, pruned_loss=0.09392, over 3064075.15 frames. ], batch size: 190, lr: 2.49e-02, grad_scale: 8.0 2023-04-27 20:02:17,756 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:33,940 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:03:19,437 INFO [train.py:904] (2/8) Epoch 2, batch 8550, loss[loss=0.2528, simple_loss=0.316, pruned_loss=0.09482, over 11843.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3338, pruned_loss=0.09269, over 3039860.27 frames. ], batch size: 247, lr: 2.49e-02, grad_scale: 4.0 2023-04-27 20:03:39,443 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-27 20:04:03,004 INFO [optim.py:368] (2/8) 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 20:04:21,648 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3707, 4.2903, 4.0999, 4.1982, 3.7984, 4.1871, 4.0928, 3.9202], device='cuda:2'), covar=tensor([0.0204, 0.0112, 0.0158, 0.0090, 0.0439, 0.0138, 0.0233, 0.0190], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0093, 0.0144, 0.0115, 0.0164, 0.0123, 0.0104, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:04:26,428 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:04:56,378 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-27 20:04:58,635 INFO [train.py:904] (2/8) Epoch 2, batch 8600, loss[loss=0.3165, simple_loss=0.3582, pruned_loss=0.1374, over 12657.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3341, pruned_loss=0.09178, over 3038862.50 frames. ], batch size: 248, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:05:32,720 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8991, 2.8203, 2.7605, 1.6702, 2.9429, 2.9158, 2.5113, 2.4705], device='cuda:2'), covar=tensor([0.0728, 0.0122, 0.0165, 0.1208, 0.0099, 0.0086, 0.0325, 0.0398], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0082, 0.0082, 0.0150, 0.0073, 0.0069, 0.0107, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 20:05:42,825 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4341, 3.3225, 3.3905, 3.0708, 3.3143, 2.0596, 3.1610, 3.1457], device='cuda:2'), covar=tensor([0.0065, 0.0065, 0.0070, 0.0159, 0.0061, 0.1040, 0.0076, 0.0090], device='cuda:2'), in_proj_covar=tensor([0.0062, 0.0052, 0.0078, 0.0090, 0.0059, 0.0112, 0.0072, 0.0076], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:06:09,476 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2620, 2.9924, 2.8697, 1.9450, 2.6268, 2.0754, 2.7782, 3.0355], device='cuda:2'), covar=tensor([0.0285, 0.0488, 0.0437, 0.1428, 0.0661, 0.0903, 0.0659, 0.0416], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0103, 0.0154, 0.0152, 0.0142, 0.0137, 0.0148, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 20:06:37,153 INFO [train.py:904] (2/8) Epoch 2, batch 8650, loss[loss=0.2427, simple_loss=0.3285, pruned_loss=0.07846, over 16306.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3316, pruned_loss=0.08992, over 3023380.38 frames. ], batch size: 146, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:07:29,096 INFO [optim.py:368] (2/8) 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 20:07:44,386 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:46,590 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:14,547 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:14,757 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9598, 1.4981, 1.8946, 2.7615, 2.7299, 2.5131, 1.8424, 2.8076], device='cuda:2'), covar=tensor([0.0033, 0.0230, 0.0154, 0.0086, 0.0039, 0.0081, 0.0171, 0.0038], device='cuda:2'), in_proj_covar=tensor([0.0072, 0.0111, 0.0097, 0.0085, 0.0063, 0.0059, 0.0098, 0.0056], device='cuda:2'), out_proj_covar=tensor([1.2115e-04, 1.9113e-04, 1.7272e-04, 1.5101e-04, 1.0514e-04, 1.0079e-04, 1.6557e-04, 9.2167e-05], device='cuda:2') 2023-04-27 20:08:23,064 INFO [train.py:904] (2/8) Epoch 2, batch 8700, loss[loss=0.2374, simple_loss=0.3238, pruned_loss=0.07551, over 16155.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3274, pruned_loss=0.08695, over 3023008.97 frames. ], batch size: 165, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:08:49,481 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:09:14,853 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:09:17,053 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:10:01,022 INFO [train.py:904] (2/8) Epoch 2, batch 8750, loss[loss=0.2687, simple_loss=0.3472, pruned_loss=0.09515, over 16717.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3271, pruned_loss=0.08591, over 3037191.27 frames. ], batch size: 134, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:10:57,518 INFO [optim.py:368] (2/8) 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 20:11:10,329 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:11:53,567 INFO [train.py:904] (2/8) Epoch 2, batch 8800, loss[loss=0.2531, simple_loss=0.3308, pruned_loss=0.0877, over 16993.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3246, pruned_loss=0.08416, over 3044821.71 frames. ], batch size: 109, lr: 2.48e-02, grad_scale: 4.0 2023-04-27 20:12:49,706 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0421, 2.4533, 2.3357, 3.3159, 3.1335, 3.3160, 1.7334, 2.7904], device='cuda:2'), covar=tensor([0.1151, 0.0389, 0.0997, 0.0078, 0.0155, 0.0280, 0.1129, 0.0524], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0123, 0.0168, 0.0070, 0.0111, 0.0121, 0.0155, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 20:13:01,153 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8619, 3.4091, 3.4683, 2.3519, 3.2750, 3.3374, 3.5362, 1.8217], device='cuda:2'), covar=tensor([0.0450, 0.0025, 0.0029, 0.0253, 0.0033, 0.0042, 0.0018, 0.0385], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0053, 0.0054, 0.0101, 0.0050, 0.0057, 0.0055, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0001, 0.0003], device='cuda:2') 2023-04-27 20:13:18,173 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:13:38,729 INFO [train.py:904] (2/8) Epoch 2, batch 8850, loss[loss=0.2128, simple_loss=0.2949, pruned_loss=0.06539, over 12483.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3257, pruned_loss=0.08291, over 3019147.95 frames. ], batch size: 247, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:14:14,989 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9127, 4.6622, 4.8902, 5.1827, 5.2069, 4.5605, 5.2996, 5.1365], device='cuda:2'), covar=tensor([0.0363, 0.0448, 0.0803, 0.0314, 0.0316, 0.0268, 0.0213, 0.0291], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0267, 0.0349, 0.0262, 0.0203, 0.0186, 0.0209, 0.0215], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:14:28,278 INFO [optim.py:368] (2/8) 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 20:14:43,433 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:15:27,232 INFO [train.py:904] (2/8) Epoch 2, batch 8900, loss[loss=0.2141, simple_loss=0.2968, pruned_loss=0.0657, over 12382.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3257, pruned_loss=0.08168, over 3030923.81 frames. ], batch size: 246, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:15:59,019 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 20:17:32,225 INFO [train.py:904] (2/8) Epoch 2, batch 8950, loss[loss=0.2189, simple_loss=0.3037, pruned_loss=0.06705, over 16973.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3257, pruned_loss=0.08208, over 3043085.34 frames. ], batch size: 109, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:18:20,856 INFO [optim.py:368] (2/8) 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 20:18:32,647 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:18:58,723 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-27 20:19:11,868 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:21,318 INFO [train.py:904] (2/8) Epoch 2, batch 9000, loss[loss=0.2716, simple_loss=0.3423, pruned_loss=0.1004, over 11882.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3226, pruned_loss=0.08045, over 3058309.12 frames. ], batch size: 248, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:19:21,318 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 20:19:31,133 INFO [train.py:938] (2/8) Epoch 2, validation: loss=0.2044, simple_loss=0.3047, pruned_loss=0.05205, over 944034.00 frames. 2023-04-27 20:19:31,134 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 20:20:00,854 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:20:52,909 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:20:58,075 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9144, 1.6809, 2.0105, 2.6777, 2.6610, 2.6712, 1.5117, 2.7395], device='cuda:2'), covar=tensor([0.0034, 0.0210, 0.0133, 0.0092, 0.0039, 0.0064, 0.0202, 0.0037], device='cuda:2'), in_proj_covar=tensor([0.0073, 0.0111, 0.0097, 0.0084, 0.0065, 0.0059, 0.0100, 0.0055], device='cuda:2'), out_proj_covar=tensor([1.2460e-04, 1.8971e-04, 1.7128e-04, 1.4886e-04, 1.0812e-04, 9.7859e-05, 1.6812e-04, 9.0878e-05], device='cuda:2') 2023-04-27 20:21:02,065 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:21:10,978 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9641, 3.8639, 2.8266, 1.3574, 2.4470, 2.0381, 3.1939, 3.6546], device='cuda:2'), covar=tensor([0.0324, 0.0384, 0.0784, 0.2245, 0.1066, 0.1282, 0.0973, 0.0714], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0102, 0.0155, 0.0150, 0.0143, 0.0134, 0.0145, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 20:21:12,719 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1099, 2.6818, 2.3453, 3.4491, 3.3588, 3.3608, 1.9412, 2.8383], device='cuda:2'), covar=tensor([0.1021, 0.0306, 0.0910, 0.0059, 0.0163, 0.0285, 0.0942, 0.0483], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0125, 0.0168, 0.0069, 0.0112, 0.0121, 0.0156, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 20:21:13,872 INFO [train.py:904] (2/8) Epoch 2, batch 9050, loss[loss=0.2625, simple_loss=0.3282, pruned_loss=0.0984, over 12621.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3242, pruned_loss=0.08155, over 3052926.17 frames. ], batch size: 247, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:21:31,678 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7142, 3.6669, 3.7417, 3.7173, 3.7228, 4.1241, 4.0644, 3.6557], device='cuda:2'), covar=tensor([0.1522, 0.1190, 0.0938, 0.1910, 0.2113, 0.1114, 0.0705, 0.2022], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0259, 0.0244, 0.0240, 0.0288, 0.0270, 0.0205, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:21:37,537 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:22:01,198 INFO [optim.py:368] (2/8) 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 20:22:58,686 INFO [train.py:904] (2/8) Epoch 2, batch 9100, loss[loss=0.2575, simple_loss=0.3415, pruned_loss=0.08681, over 15438.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3236, pruned_loss=0.08182, over 3058846.71 frames. ], batch size: 191, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:23:00,169 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 20:24:22,925 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:24:55,899 INFO [train.py:904] (2/8) Epoch 2, batch 9150, loss[loss=0.258, simple_loss=0.3253, pruned_loss=0.09534, over 11897.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3231, pruned_loss=0.08074, over 3053568.62 frames. ], batch size: 248, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:25:49,332 INFO [optim.py:368] (2/8) 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 20:25:59,070 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:26:40,136 INFO [train.py:904] (2/8) Epoch 2, batch 9200, loss[loss=0.2326, simple_loss=0.3177, pruned_loss=0.07372, over 16126.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3177, pruned_loss=0.07879, over 3062436.73 frames. ], batch size: 165, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:27:30,640 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:03,128 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:16,034 INFO [train.py:904] (2/8) Epoch 2, batch 9250, loss[loss=0.201, simple_loss=0.2935, pruned_loss=0.05423, over 16443.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3174, pruned_loss=0.07946, over 3035082.91 frames. ], batch size: 68, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:29:05,881 INFO [optim.py:368] (2/8) 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 20:29:08,224 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-27 20:29:36,847 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-27 20:30:06,013 INFO [train.py:904] (2/8) Epoch 2, batch 9300, loss[loss=0.2252, simple_loss=0.3054, pruned_loss=0.07246, over 15448.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3155, pruned_loss=0.07862, over 3019408.62 frames. ], batch size: 191, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:30:14,342 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:31:20,780 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:31:49,020 INFO [train.py:904] (2/8) Epoch 2, batch 9350, loss[loss=0.2249, simple_loss=0.3015, pruned_loss=0.07411, over 12035.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3157, pruned_loss=0.07849, over 3040024.88 frames. ], batch size: 249, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:32:37,319 INFO [optim.py:368] (2/8) 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 20:33:28,460 INFO [train.py:904] (2/8) Epoch 2, batch 9400, loss[loss=0.23, simple_loss=0.2943, pruned_loss=0.08279, over 12509.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3163, pruned_loss=0.07795, over 3044535.20 frames. ], batch size: 248, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:34:39,210 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:08,265 INFO [train.py:904] (2/8) Epoch 2, batch 9450, loss[loss=0.2178, simple_loss=0.3088, pruned_loss=0.06337, over 16697.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.319, pruned_loss=0.07864, over 3051037.18 frames. ], batch size: 83, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:35:42,282 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:51,741 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8162, 2.6986, 2.6210, 2.0151, 2.4684, 2.6172, 2.7214, 1.8336], device='cuda:2'), covar=tensor([0.0364, 0.0039, 0.0048, 0.0214, 0.0055, 0.0069, 0.0037, 0.0322], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0055, 0.0057, 0.0105, 0.0053, 0.0057, 0.0059, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 20:35:56,416 INFO [optim.py:368] (2/8) 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 20:36:16,142 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:36:25,709 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5281, 4.1950, 4.4778, 4.7927, 4.7923, 4.2880, 4.8931, 4.7294], device='cuda:2'), covar=tensor([0.0516, 0.0543, 0.0907, 0.0361, 0.0410, 0.0388, 0.0301, 0.0283], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0271, 0.0357, 0.0263, 0.0209, 0.0185, 0.0214, 0.0216], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:36:48,218 INFO [train.py:904] (2/8) Epoch 2, batch 9500, loss[loss=0.2148, simple_loss=0.2932, pruned_loss=0.06816, over 12763.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3169, pruned_loss=0.0771, over 3060240.51 frames. ], batch size: 246, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:37:02,510 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7752, 5.3076, 5.3932, 5.3825, 5.3786, 5.8109, 5.6013, 5.3189], device='cuda:2'), covar=tensor([0.0537, 0.0988, 0.0631, 0.1080, 0.1262, 0.0592, 0.0597, 0.1286], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0256, 0.0239, 0.0229, 0.0279, 0.0260, 0.0195, 0.0294], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:37:46,277 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:38:25,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4955, 3.2607, 2.8736, 2.3919, 2.3514, 2.0532, 3.2016, 3.6506], device='cuda:2'), covar=tensor([0.1535, 0.0536, 0.0817, 0.0775, 0.1386, 0.1182, 0.0340, 0.0217], device='cuda:2'), in_proj_covar=tensor([0.0248, 0.0224, 0.0242, 0.0182, 0.0210, 0.0182, 0.0198, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 20:38:34,553 INFO [train.py:904] (2/8) Epoch 2, batch 9550, loss[loss=0.2421, simple_loss=0.3272, pruned_loss=0.07844, over 16916.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3165, pruned_loss=0.07709, over 3075028.69 frames. ], batch size: 109, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:39:23,641 INFO [optim.py:368] (2/8) 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 20:40:12,588 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19750.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:40:13,477 INFO [train.py:904] (2/8) Epoch 2, batch 9600, loss[loss=0.2234, simple_loss=0.3055, pruned_loss=0.07066, over 16621.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3184, pruned_loss=0.07835, over 3063075.18 frames. ], batch size: 57, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:40:16,275 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7646, 1.1494, 1.5094, 1.7548, 1.7726, 1.7596, 1.3851, 1.7226], device='cuda:2'), covar=tensor([0.0059, 0.0183, 0.0094, 0.0108, 0.0051, 0.0066, 0.0149, 0.0056], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0113, 0.0098, 0.0084, 0.0064, 0.0060, 0.0100, 0.0055], device='cuda:2'), out_proj_covar=tensor([1.2541e-04, 1.9173e-04, 1.7125e-04, 1.4625e-04, 1.0465e-04, 9.9075e-05, 1.6534e-04, 8.9096e-05], device='cuda:2') 2023-04-27 20:40:23,559 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7707, 2.8451, 2.3218, 3.9712, 3.7984, 3.7980, 1.5569, 2.6328], device='cuda:2'), covar=tensor([0.1456, 0.0525, 0.1198, 0.0066, 0.0149, 0.0274, 0.1310, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0124, 0.0163, 0.0068, 0.0110, 0.0121, 0.0151, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 20:41:22,094 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:42:00,141 INFO [train.py:904] (2/8) Epoch 2, batch 9650, loss[loss=0.25, simple_loss=0.3298, pruned_loss=0.08509, over 16166.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3207, pruned_loss=0.07909, over 3057708.98 frames. ], batch size: 165, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:42:54,373 INFO [optim.py:368] (2/8) 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 20:43:09,972 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:43:48,656 INFO [train.py:904] (2/8) Epoch 2, batch 9700, loss[loss=0.2199, simple_loss=0.3098, pruned_loss=0.065, over 15431.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3189, pruned_loss=0.07884, over 3041549.02 frames. ], batch size: 190, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:44:08,834 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7632, 2.6774, 2.5801, 1.8950, 2.4061, 2.5762, 2.7463, 1.7923], device='cuda:2'), covar=tensor([0.0359, 0.0030, 0.0048, 0.0228, 0.0047, 0.0052, 0.0026, 0.0324], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0053, 0.0057, 0.0103, 0.0051, 0.0056, 0.0056, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0001, 0.0003], device='cuda:2') 2023-04-27 20:45:33,563 INFO [train.py:904] (2/8) Epoch 2, batch 9750, loss[loss=0.25, simple_loss=0.3327, pruned_loss=0.08365, over 16476.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3178, pruned_loss=0.07902, over 3047843.32 frames. ], batch size: 147, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:46:21,255 INFO [optim.py:368] (2/8) 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 20:47:15,376 INFO [train.py:904] (2/8) Epoch 2, batch 9800, loss[loss=0.2402, simple_loss=0.3321, pruned_loss=0.07416, over 15371.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3179, pruned_loss=0.07791, over 3042635.78 frames. ], batch size: 191, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:48:00,420 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:49:05,346 INFO [train.py:904] (2/8) Epoch 2, batch 9850, loss[loss=0.22, simple_loss=0.2951, pruned_loss=0.07243, over 12524.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3193, pruned_loss=0.07784, over 3058187.39 frames. ], batch size: 248, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:49:46,060 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:49:55,600 INFO [optim.py:368] (2/8) 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 20:50:56,299 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8139, 1.1730, 1.4381, 1.8134, 1.8512, 1.5907, 1.4521, 1.8121], device='cuda:2'), covar=tensor([0.0060, 0.0166, 0.0092, 0.0091, 0.0044, 0.0072, 0.0140, 0.0047], device='cuda:2'), in_proj_covar=tensor([0.0077, 0.0111, 0.0098, 0.0084, 0.0065, 0.0060, 0.0100, 0.0054], device='cuda:2'), out_proj_covar=tensor([1.3029e-04, 1.8639e-04, 1.7126e-04, 1.4735e-04, 1.0530e-04, 9.9291e-05, 1.6499e-04, 8.8211e-05], device='cuda:2') 2023-04-27 20:50:58,119 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:50:58,949 INFO [train.py:904] (2/8) Epoch 2, batch 9900, loss[loss=0.2401, simple_loss=0.3294, pruned_loss=0.07537, over 16450.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3196, pruned_loss=0.07733, over 3062197.80 frames. ], batch size: 146, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:52:12,615 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:31,246 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:51,004 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:56,912 INFO [train.py:904] (2/8) Epoch 2, batch 9950, loss[loss=0.245, simple_loss=0.3222, pruned_loss=0.08392, over 12616.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3222, pruned_loss=0.07784, over 3075924.54 frames. ], batch size: 248, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:53:23,641 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:53:55,570 INFO [optim.py:368] (2/8) 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 20:54:55,881 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:54:56,740 INFO [train.py:904] (2/8) Epoch 2, batch 10000, loss[loss=0.2199, simple_loss=0.3107, pruned_loss=0.06452, over 16896.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3198, pruned_loss=0.07672, over 3081093.89 frames. ], batch size: 125, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:55:41,185 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:56:37,810 INFO [train.py:904] (2/8) Epoch 2, batch 10050, loss[loss=0.2372, simple_loss=0.3187, pruned_loss=0.07789, over 12261.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3191, pruned_loss=0.07607, over 3079000.02 frames. ], batch size: 248, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:56:44,607 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:57:24,815 INFO [optim.py:368] (2/8) 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 20:58:10,190 INFO [train.py:904] (2/8) Epoch 2, batch 10100, loss[loss=0.2172, simple_loss=0.289, pruned_loss=0.07272, over 12734.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3196, pruned_loss=0.07668, over 3072581.84 frames. ], batch size: 250, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:58:38,010 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:58:58,260 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:59:54,874 INFO [train.py:904] (2/8) Epoch 3, batch 0, loss[loss=0.431, simple_loss=0.4538, pruned_loss=0.2041, over 15382.00 frames. ], tot_loss[loss=0.431, simple_loss=0.4538, pruned_loss=0.2041, over 15382.00 frames. ], batch size: 190, lr: 2.28e-02, grad_scale: 8.0 2023-04-27 20:59:54,875 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 21:00:02,292 INFO [train.py:938] (2/8) Epoch 3, validation: loss=0.2012, simple_loss=0.3019, pruned_loss=0.05024, over 944034.00 frames. 2023-04-27 21:00:02,293 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 21:00:13,636 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4874, 4.2407, 4.4044, 4.6923, 4.7472, 4.2733, 4.8548, 4.6731], device='cuda:2'), covar=tensor([0.0529, 0.0594, 0.1091, 0.0493, 0.0447, 0.0399, 0.0356, 0.0344], device='cuda:2'), in_proj_covar=tensor([0.0241, 0.0280, 0.0362, 0.0276, 0.0210, 0.0193, 0.0216, 0.0219], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:00:31,871 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:38,799 INFO [optim.py:368] (2/8) 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 21:00:54,002 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0097, 5.5574, 5.4507, 5.4889, 5.4619, 5.9835, 5.7017, 5.4924], device='cuda:2'), covar=tensor([0.0530, 0.1080, 0.1197, 0.1419, 0.2274, 0.0666, 0.0855, 0.1715], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0286, 0.0261, 0.0254, 0.0317, 0.0279, 0.0214, 0.0329], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:01:12,891 INFO [train.py:904] (2/8) Epoch 3, batch 50, loss[loss=0.3086, simple_loss=0.3611, pruned_loss=0.1281, over 16299.00 frames. ], tot_loss[loss=0.2951, simple_loss=0.3495, pruned_loss=0.1203, over 745279.96 frames. ], batch size: 165, lr: 2.28e-02, grad_scale: 2.0 2023-04-27 21:01:45,735 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:02,780 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:19,621 INFO [train.py:904] (2/8) Epoch 3, batch 100, loss[loss=0.2972, simple_loss=0.3501, pruned_loss=0.1221, over 15493.00 frames. ], tot_loss[loss=0.2813, simple_loss=0.3405, pruned_loss=0.1111, over 1320991.83 frames. ], batch size: 190, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:02:43,360 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:54,677 INFO [optim.py:368] (2/8) 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 21:03:18,754 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:23,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0031, 3.8719, 3.2061, 1.6992, 2.6713, 1.9440, 3.4970, 3.7348], device='cuda:2'), covar=tensor([0.0186, 0.0334, 0.0432, 0.1512, 0.0709, 0.1046, 0.0484, 0.0463], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0104, 0.0154, 0.0148, 0.0139, 0.0135, 0.0142, 0.0109], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 21:03:24,369 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:27,023 INFO [train.py:904] (2/8) Epoch 3, batch 150, loss[loss=0.2446, simple_loss=0.3233, pruned_loss=0.08292, over 17279.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3354, pruned_loss=0.1052, over 1759771.65 frames. ], batch size: 52, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:03:49,093 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:05,880 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:09,675 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2023-04-27 21:04:35,430 INFO [train.py:904] (2/8) Epoch 3, batch 200, loss[loss=0.2837, simple_loss=0.3297, pruned_loss=0.1189, over 16842.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3328, pruned_loss=0.1029, over 2109338.48 frames. ], batch size: 90, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:09,761 INFO [optim.py:368] (2/8) 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 21:05:43,634 INFO [train.py:904] (2/8) Epoch 3, batch 250, loss[loss=0.2446, simple_loss=0.3287, pruned_loss=0.08019, over 17048.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.329, pruned_loss=0.1012, over 2375132.17 frames. ], batch size: 50, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:58,361 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:19,869 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5410, 4.3591, 1.9871, 4.5913, 2.7080, 4.6654, 2.2424, 3.2362], device='cuda:2'), covar=tensor([0.0033, 0.0158, 0.1639, 0.0029, 0.0931, 0.0228, 0.1389, 0.0615], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0123, 0.0172, 0.0078, 0.0159, 0.0148, 0.0179, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-27 21:06:35,566 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:54,934 INFO [train.py:904] (2/8) Epoch 3, batch 300, loss[loss=0.2495, simple_loss=0.3134, pruned_loss=0.09281, over 16471.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3259, pruned_loss=0.09851, over 2592084.07 frames. ], batch size: 146, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:07:29,012 INFO [optim.py:368] (2/8) 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 21:07:31,877 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9696, 2.2197, 1.7108, 1.9131, 2.8623, 2.7762, 2.8151, 2.9663], device='cuda:2'), covar=tensor([0.0076, 0.0138, 0.0156, 0.0174, 0.0072, 0.0090, 0.0056, 0.0059], device='cuda:2'), in_proj_covar=tensor([0.0057, 0.0113, 0.0109, 0.0113, 0.0102, 0.0110, 0.0064, 0.0082], device='cuda:2'), out_proj_covar=tensor([7.8763e-05, 1.7069e-04, 1.5911e-04, 1.6916e-04, 1.5768e-04, 1.6921e-04, 9.3752e-05, 1.2796e-04], device='cuda:2') 2023-04-27 21:07:42,779 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9659, 4.7780, 4.6622, 4.8098, 4.1770, 4.7168, 4.8689, 4.3997], device='cuda:2'), covar=tensor([0.0300, 0.0175, 0.0174, 0.0117, 0.0885, 0.0184, 0.0223, 0.0243], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0111, 0.0167, 0.0139, 0.0207, 0.0150, 0.0121, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:07:59,206 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:08:01,102 INFO [train.py:904] (2/8) Epoch 3, batch 350, loss[loss=0.2248, simple_loss=0.3071, pruned_loss=0.07122, over 17257.00 frames. ], tot_loss[loss=0.2582, simple_loss=0.3226, pruned_loss=0.09686, over 2756740.50 frames. ], batch size: 52, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:08:36,912 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 3, batch 400, loss[loss=0.2635, simple_loss=0.3181, pruned_loss=0.1044, over 16387.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.3214, pruned_loss=0.09633, over 2882168.45 frames. ], batch size: 146, lr: 2.26e-02, grad_scale: 4.0 2023-04-27 21:09:24,829 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 21:09:41,230 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:45,246 INFO [optim.py:368] (2/8) 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 21:10:08,711 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:10,665 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:17,850 INFO [train.py:904] (2/8) Epoch 3, batch 450, loss[loss=0.2556, simple_loss=0.3208, pruned_loss=0.0952, over 16478.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3203, pruned_loss=0.09558, over 2973679.10 frames. ], batch size: 75, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:10:29,527 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-27 21:10:41,862 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:49,476 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:14,833 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:18,627 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-27 21:11:24,055 INFO [train.py:904] (2/8) Epoch 3, batch 500, loss[loss=0.2357, simple_loss=0.3174, pruned_loss=0.07701, over 17126.00 frames. ], tot_loss[loss=0.2517, simple_loss=0.3179, pruned_loss=0.09277, over 3052263.18 frames. ], batch size: 49, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:11:46,147 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:00,584 INFO [optim.py:368] (2/8) 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 21:12:34,379 INFO [train.py:904] (2/8) Epoch 3, batch 550, loss[loss=0.2694, simple_loss=0.3235, pruned_loss=0.1076, over 16522.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3162, pruned_loss=0.09096, over 3107509.79 frames. ], batch size: 146, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:12:45,444 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:21,801 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9852, 4.9533, 4.7910, 4.2209, 4.7862, 1.9830, 4.4761, 4.8803], device='cuda:2'), covar=tensor([0.0058, 0.0045, 0.0067, 0.0247, 0.0045, 0.1275, 0.0068, 0.0072], device='cuda:2'), in_proj_covar=tensor([0.0071, 0.0058, 0.0091, 0.0103, 0.0066, 0.0116, 0.0079, 0.0089], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:13:40,833 INFO [train.py:904] (2/8) Epoch 3, batch 600, loss[loss=0.2744, simple_loss=0.3346, pruned_loss=0.1071, over 15400.00 frames. ], tot_loss[loss=0.2479, simple_loss=0.3152, pruned_loss=0.09035, over 3155229.76 frames. ], batch size: 190, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:13:50,522 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:15,688 INFO [optim.py:368] (2/8) 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 21:14:38,329 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:48,030 INFO [train.py:904] (2/8) Epoch 3, batch 650, loss[loss=0.2071, simple_loss=0.2798, pruned_loss=0.0672, over 16742.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3125, pruned_loss=0.08955, over 3172659.36 frames. ], batch size: 39, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:15:39,329 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-27 21:15:57,310 INFO [train.py:904] (2/8) Epoch 3, batch 700, loss[loss=0.2284, simple_loss=0.3023, pruned_loss=0.0773, over 17218.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.312, pruned_loss=0.0889, over 3207264.03 frames. ], batch size: 44, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:16:30,866 INFO [optim.py:368] (2/8) 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 21:16:55,472 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:16:58,237 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 21:17:03,823 INFO [train.py:904] (2/8) Epoch 3, batch 750, loss[loss=0.2137, simple_loss=0.2799, pruned_loss=0.07373, over 16531.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3142, pruned_loss=0.0909, over 3227244.14 frames. ], batch size: 75, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:17:06,022 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:35,034 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:57,837 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:10,815 INFO [train.py:904] (2/8) Epoch 3, batch 800, loss[loss=0.2439, simple_loss=0.3122, pruned_loss=0.08786, over 16848.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3136, pruned_loss=0.09034, over 3252017.36 frames. ], batch size: 83, lr: 2.24e-02, grad_scale: 8.0 2023-04-27 21:18:27,343 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:34,945 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:40,095 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:47,524 INFO [optim.py:368] (2/8) 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 21:19:11,073 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 21:19:20,166 INFO [train.py:904] (2/8) Epoch 3, batch 850, loss[loss=0.1983, simple_loss=0.2833, pruned_loss=0.05666, over 17191.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3123, pruned_loss=0.08893, over 3263812.96 frames. ], batch size: 46, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:19:58,189 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:20:27,600 INFO [train.py:904] (2/8) Epoch 3, batch 900, loss[loss=0.2288, simple_loss=0.2987, pruned_loss=0.07945, over 16518.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3112, pruned_loss=0.08823, over 3259588.85 frames. ], batch size: 68, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:21:03,013 INFO [optim.py:368] (2/8) 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 21:21:26,752 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:28,502 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9511, 4.7772, 4.9385, 5.2585, 5.3752, 4.6870, 5.3758, 5.2403], device='cuda:2'), covar=tensor([0.0462, 0.0521, 0.0976, 0.0340, 0.0322, 0.0403, 0.0278, 0.0259], device='cuda:2'), in_proj_covar=tensor([0.0291, 0.0340, 0.0466, 0.0349, 0.0266, 0.0242, 0.0259, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:21:35,850 INFO [train.py:904] (2/8) Epoch 3, batch 950, loss[loss=0.2148, simple_loss=0.2851, pruned_loss=0.07221, over 16273.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3104, pruned_loss=0.08764, over 3274065.21 frames. ], batch size: 36, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:22:34,056 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:22:44,925 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1471, 1.4615, 2.2712, 2.9250, 2.6455, 3.5544, 1.7112, 3.0465], device='cuda:2'), covar=tensor([0.0045, 0.0222, 0.0111, 0.0099, 0.0064, 0.0041, 0.0164, 0.0036], device='cuda:2'), in_proj_covar=tensor([0.0083, 0.0118, 0.0105, 0.0099, 0.0077, 0.0066, 0.0106, 0.0060], device='cuda:2'), out_proj_covar=tensor([1.3829e-04, 1.9764e-04, 1.7997e-04, 1.6991e-04, 1.2372e-04, 1.0963e-04, 1.7266e-04, 9.8717e-05], device='cuda:2') 2023-04-27 21:22:45,620 INFO [train.py:904] (2/8) Epoch 3, batch 1000, loss[loss=0.2434, simple_loss=0.3205, pruned_loss=0.08315, over 17015.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3095, pruned_loss=0.08772, over 3277450.50 frames. ], batch size: 50, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:23:01,340 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 21:23:23,975 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:55,372 INFO [train.py:904] (2/8) Epoch 3, batch 1050, loss[loss=0.2126, simple_loss=0.302, pruned_loss=0.06161, over 17107.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3092, pruned_loss=0.08782, over 3294672.54 frames. ], batch size: 47, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:24:25,396 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:24:46,820 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:02,813 INFO [train.py:904] (2/8) Epoch 3, batch 1100, loss[loss=0.2117, simple_loss=0.2926, pruned_loss=0.06539, over 17202.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3084, pruned_loss=0.08695, over 3297930.55 frames. ], batch size: 44, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:25:12,307 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:38,414 INFO [optim.py:368] (2/8) 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 21:26:10,402 INFO [train.py:904] (2/8) Epoch 3, batch 1150, loss[loss=0.2567, simple_loss=0.3107, pruned_loss=0.1014, over 16900.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3072, pruned_loss=0.08589, over 3302630.66 frames. ], batch size: 116, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:26:42,688 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:27:02,125 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7259, 2.6837, 2.7353, 4.3237, 2.0465, 3.9468, 2.3916, 2.5515], device='cuda:2'), covar=tensor([0.0315, 0.0561, 0.0355, 0.0152, 0.1508, 0.0202, 0.0829, 0.1076], device='cuda:2'), in_proj_covar=tensor([0.0245, 0.0217, 0.0181, 0.0245, 0.0286, 0.0187, 0.0209, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:27:19,793 INFO [train.py:904] (2/8) Epoch 3, batch 1200, loss[loss=0.2568, simple_loss=0.325, pruned_loss=0.09428, over 16654.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.306, pruned_loss=0.08465, over 3294970.14 frames. ], batch size: 62, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:27:56,780 INFO [optim.py:368] (2/8) 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 21:28:27,750 INFO [train.py:904] (2/8) Epoch 3, batch 1250, loss[loss=0.2374, simple_loss=0.2981, pruned_loss=0.08835, over 16493.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3063, pruned_loss=0.08448, over 3294456.10 frames. ], batch size: 75, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:28:37,552 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0939, 3.1231, 3.8112, 2.6086, 3.7148, 3.8809, 3.8270, 1.9307], device='cuda:2'), covar=tensor([0.0402, 0.0202, 0.0039, 0.0227, 0.0054, 0.0050, 0.0041, 0.0378], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0059, 0.0062, 0.0107, 0.0055, 0.0062, 0.0061, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:28:56,350 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2850, 5.7261, 5.4080, 5.5805, 4.8966, 4.7392, 5.2420, 5.8623], device='cuda:2'), covar=tensor([0.0546, 0.0602, 0.0922, 0.0365, 0.0603, 0.0589, 0.0530, 0.0604], device='cuda:2'), in_proj_covar=tensor([0.0264, 0.0381, 0.0329, 0.0231, 0.0253, 0.0229, 0.0296, 0.0263], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:29:30,664 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:29:39,771 INFO [train.py:904] (2/8) Epoch 3, batch 1300, loss[loss=0.2332, simple_loss=0.3121, pruned_loss=0.07716, over 17206.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3054, pruned_loss=0.08387, over 3303152.83 frames. ], batch size: 45, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:56,026 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:17,378 INFO [optim.py:368] (2/8) 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 21:30:17,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5861, 4.8626, 5.2317, 5.2181, 5.2585, 4.9374, 4.5666, 4.6651], device='cuda:2'), covar=tensor([0.0369, 0.0309, 0.0389, 0.0553, 0.0510, 0.0351, 0.1111, 0.0414], device='cuda:2'), in_proj_covar=tensor([0.0193, 0.0181, 0.0205, 0.0199, 0.0234, 0.0204, 0.0305, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 21:30:49,954 INFO [train.py:904] (2/8) Epoch 3, batch 1350, loss[loss=0.2413, simple_loss=0.3087, pruned_loss=0.08696, over 15505.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3055, pruned_loss=0.08308, over 3306582.27 frames. ], batch size: 191, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:30:56,219 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:59,537 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 21:31:09,542 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2251, 2.2218, 1.9483, 2.0749, 2.9421, 2.7817, 3.5941, 3.1357], device='cuda:2'), covar=tensor([0.0023, 0.0152, 0.0162, 0.0162, 0.0080, 0.0119, 0.0066, 0.0065], device='cuda:2'), in_proj_covar=tensor([0.0056, 0.0111, 0.0109, 0.0110, 0.0100, 0.0110, 0.0068, 0.0084], device='cuda:2'), out_proj_covar=tensor([7.8653e-05, 1.6655e-04, 1.5661e-04, 1.6257e-04, 1.5311e-04, 1.6663e-04, 1.0080e-04, 1.2965e-04], device='cuda:2') 2023-04-27 21:31:13,082 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:20,888 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:35,221 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:58,965 INFO [train.py:904] (2/8) Epoch 3, batch 1400, loss[loss=0.2401, simple_loss=0.3057, pruned_loss=0.08723, over 16805.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3067, pruned_loss=0.08445, over 3307225.66 frames. ], batch size: 102, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:32:08,267 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4621, 4.3347, 3.9481, 1.8990, 3.0914, 2.1747, 3.9348, 4.2475], device='cuda:2'), covar=tensor([0.0338, 0.0480, 0.0443, 0.1543, 0.0666, 0.1011, 0.0661, 0.0670], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0120, 0.0154, 0.0146, 0.0138, 0.0132, 0.0148, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 21:32:09,165 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:32:11,851 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 21:32:17,934 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2434, 5.1837, 4.9224, 4.3647, 5.0268, 2.2028, 4.7569, 5.0539], device='cuda:2'), covar=tensor([0.0043, 0.0044, 0.0061, 0.0286, 0.0047, 0.1103, 0.0060, 0.0093], device='cuda:2'), in_proj_covar=tensor([0.0076, 0.0062, 0.0096, 0.0113, 0.0071, 0.0117, 0.0086, 0.0099], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:32:35,074 INFO [optim.py:368] (2/8) 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 21:33:07,172 INFO [train.py:904] (2/8) Epoch 3, batch 1450, loss[loss=0.229, simple_loss=0.2922, pruned_loss=0.08292, over 16256.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3065, pruned_loss=0.08446, over 3300825.17 frames. ], batch size: 165, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:33:13,877 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:23,950 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:31,477 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3615, 4.6908, 4.4139, 4.4436, 4.0436, 4.1829, 4.1895, 4.7038], device='cuda:2'), covar=tensor([0.0497, 0.0697, 0.0896, 0.0400, 0.0611, 0.0704, 0.0563, 0.0701], device='cuda:2'), in_proj_covar=tensor([0.0268, 0.0381, 0.0328, 0.0230, 0.0249, 0.0229, 0.0299, 0.0262], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:33:39,412 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:42,842 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0833, 5.4303, 5.1033, 5.2286, 4.7406, 4.6617, 4.9332, 5.4917], device='cuda:2'), covar=tensor([0.0472, 0.0610, 0.0872, 0.0312, 0.0532, 0.0609, 0.0458, 0.0616], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0378, 0.0325, 0.0228, 0.0247, 0.0227, 0.0297, 0.0259], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:33:45,726 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-27 21:34:10,050 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-27 21:34:14,645 INFO [train.py:904] (2/8) Epoch 3, batch 1500, loss[loss=0.2496, simple_loss=0.3267, pruned_loss=0.08629, over 17081.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3062, pruned_loss=0.08459, over 3309891.47 frames. ], batch size: 53, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:34:43,974 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:47,230 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:52,726 INFO [optim.py:368] (2/8) 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 21:35:23,698 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:35:24,398 INFO [train.py:904] (2/8) Epoch 3, batch 1550, loss[loss=0.3137, simple_loss=0.3373, pruned_loss=0.1451, over 16870.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3075, pruned_loss=0.08614, over 3318104.08 frames. ], batch size: 109, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:09,158 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-27 21:36:31,485 INFO [train.py:904] (2/8) Epoch 3, batch 1600, loss[loss=0.2445, simple_loss=0.3135, pruned_loss=0.08768, over 16772.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3099, pruned_loss=0.08738, over 3324459.86 frames. ], batch size: 89, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:46,720 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:00,729 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 21:37:07,920 INFO [optim.py:368] (2/8) 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 21:37:16,160 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5146, 5.8744, 5.5657, 5.7998, 5.1223, 4.8645, 5.4415, 5.9670], device='cuda:2'), covar=tensor([0.0448, 0.0545, 0.0799, 0.0290, 0.0493, 0.0477, 0.0407, 0.0487], device='cuda:2'), in_proj_covar=tensor([0.0275, 0.0384, 0.0335, 0.0234, 0.0249, 0.0231, 0.0302, 0.0262], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:37:30,359 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.88 vs. limit=5.0 2023-04-27 21:37:38,900 INFO [train.py:904] (2/8) Epoch 3, batch 1650, loss[loss=0.2069, simple_loss=0.2793, pruned_loss=0.06721, over 16805.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3119, pruned_loss=0.08808, over 3322469.93 frames. ], batch size: 39, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:37:39,891 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:01,255 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:03,297 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:04,725 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:24,247 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:39,955 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5359, 4.1858, 4.4405, 3.1978, 4.1164, 4.5611, 4.3479, 2.3049], device='cuda:2'), covar=tensor([0.0378, 0.0038, 0.0025, 0.0219, 0.0028, 0.0028, 0.0019, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0058, 0.0062, 0.0111, 0.0056, 0.0061, 0.0062, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:38:50,890 INFO [train.py:904] (2/8) Epoch 3, batch 1700, loss[loss=0.2406, simple_loss=0.314, pruned_loss=0.08362, over 17186.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3146, pruned_loss=0.08846, over 3321266.43 frames. ], batch size: 46, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:39:11,659 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:13,044 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4160, 5.2419, 5.1960, 5.1753, 4.6816, 5.1983, 5.1976, 4.8993], device='cuda:2'), covar=tensor([0.0240, 0.0122, 0.0142, 0.0110, 0.0800, 0.0178, 0.0134, 0.0233], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0124, 0.0188, 0.0156, 0.0226, 0.0167, 0.0135, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:39:28,009 INFO [optim.py:368] (2/8) 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 21:39:30,844 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:57,292 INFO [train.py:904] (2/8) Epoch 3, batch 1750, loss[loss=0.2576, simple_loss=0.3397, pruned_loss=0.08773, over 17055.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3158, pruned_loss=0.08929, over 3313167.52 frames. ], batch size: 53, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:40:55,828 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1164, 4.7786, 4.8785, 3.5725, 4.8750, 1.8518, 4.5740, 4.9372], device='cuda:2'), covar=tensor([0.0099, 0.0100, 0.0089, 0.0582, 0.0085, 0.1748, 0.0121, 0.0149], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0064, 0.0100, 0.0116, 0.0073, 0.0119, 0.0089, 0.0100], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:41:06,921 INFO [train.py:904] (2/8) Epoch 3, batch 1800, loss[loss=0.2717, simple_loss=0.3203, pruned_loss=0.1115, over 16864.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3163, pruned_loss=0.0884, over 3318029.42 frames. ], batch size: 109, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:41:30,522 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22119.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:41:44,204 INFO [optim.py:368] (2/8) 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 21:41:45,816 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:41:51,838 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0623, 2.5756, 2.3999, 4.4912, 1.8590, 4.3181, 2.2855, 2.3377], device='cuda:2'), covar=tensor([0.0291, 0.0734, 0.0465, 0.0171, 0.1890, 0.0181, 0.0929, 0.1648], device='cuda:2'), in_proj_covar=tensor([0.0241, 0.0217, 0.0182, 0.0245, 0.0285, 0.0189, 0.0208, 0.0275], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:42:09,367 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.23 vs. limit=5.0 2023-04-27 21:42:14,095 INFO [train.py:904] (2/8) Epoch 3, batch 1850, loss[loss=0.2098, simple_loss=0.2958, pruned_loss=0.0619, over 17123.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3171, pruned_loss=0.08886, over 3317868.90 frames. ], batch size: 48, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:42:47,478 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:09,373 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:22,338 INFO [train.py:904] (2/8) Epoch 3, batch 1900, loss[loss=0.1944, simple_loss=0.2698, pruned_loss=0.05948, over 15884.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3159, pruned_loss=0.08778, over 3313326.72 frames. ], batch size: 35, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:43:31,012 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:42,649 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:02,290 INFO [optim.py:368] (2/8) 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 21:44:12,163 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:12,261 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:32,422 INFO [train.py:904] (2/8) Epoch 3, batch 1950, loss[loss=0.2172, simple_loss=0.2942, pruned_loss=0.07011, over 16828.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3148, pruned_loss=0.08641, over 3319215.01 frames. ], batch size: 42, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:44:32,726 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:58,040 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:06,092 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:35,259 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:36,246 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:38,311 INFO [train.py:904] (2/8) Epoch 3, batch 2000, loss[loss=0.245, simple_loss=0.3234, pruned_loss=0.08328, over 16673.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3141, pruned_loss=0.08616, over 3318820.89 frames. ], batch size: 57, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:46:01,055 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:13,073 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 21:46:27,327 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1577, 2.2361, 1.6198, 2.1048, 2.7691, 2.7173, 3.1318, 2.9073], device='cuda:2'), covar=tensor([0.0043, 0.0130, 0.0161, 0.0142, 0.0071, 0.0101, 0.0043, 0.0060], device='cuda:2'), in_proj_covar=tensor([0.0061, 0.0118, 0.0114, 0.0116, 0.0109, 0.0117, 0.0076, 0.0092], device='cuda:2'), out_proj_covar=tensor([8.8459e-05, 1.7525e-04, 1.6346e-04, 1.7108e-04, 1.6551e-04, 1.7662e-04, 1.1226e-04, 1.4209e-04], device='cuda:2') 2023-04-27 21:46:46,968 INFO [train.py:904] (2/8) Epoch 3, batch 2050, loss[loss=0.222, simple_loss=0.3037, pruned_loss=0.07017, over 17120.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3139, pruned_loss=0.08644, over 3326286.21 frames. ], batch size: 47, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:47:11,071 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 21:47:33,567 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:47:55,914 INFO [train.py:904] (2/8) Epoch 3, batch 2100, loss[loss=0.2565, simple_loss=0.3315, pruned_loss=0.09073, over 16708.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3157, pruned_loss=0.08788, over 3328028.56 frames. ], batch size: 57, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:48:20,726 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:48:31,786 INFO [optim.py:368] (2/8) 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 21:48:55,058 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:00,597 INFO [train.py:904] (2/8) Epoch 3, batch 2150, loss[loss=0.1992, simple_loss=0.2854, pruned_loss=0.05653, over 16795.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3164, pruned_loss=0.08833, over 3328297.94 frames. ], batch size: 42, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:49:22,488 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:29,577 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0458, 3.0621, 3.3086, 2.1791, 3.1761, 3.2574, 3.1214, 1.9178], device='cuda:2'), covar=tensor([0.0316, 0.0092, 0.0041, 0.0240, 0.0039, 0.0052, 0.0032, 0.0309], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0056, 0.0060, 0.0107, 0.0054, 0.0061, 0.0062, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:49:47,663 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:08,035 INFO [train.py:904] (2/8) Epoch 3, batch 2200, loss[loss=0.3483, simple_loss=0.3853, pruned_loss=0.1557, over 11987.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3173, pruned_loss=0.08912, over 3325568.00 frames. ], batch size: 246, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:50:14,155 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 21:50:48,945 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:56,093 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5972, 4.2425, 4.4192, 2.9768, 4.2073, 4.3331, 4.3158, 2.6618], device='cuda:2'), covar=tensor([0.0295, 0.0028, 0.0023, 0.0211, 0.0020, 0.0031, 0.0019, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0057, 0.0059, 0.0107, 0.0054, 0.0061, 0.0061, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:51:15,037 INFO [train.py:904] (2/8) Epoch 3, batch 2250, loss[loss=0.2634, simple_loss=0.3234, pruned_loss=0.1017, over 15405.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3185, pruned_loss=0.08989, over 3325621.43 frames. ], batch size: 191, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:51:18,565 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:23,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8967, 4.2745, 3.3819, 2.6303, 3.1465, 2.2778, 4.5565, 4.7597], device='cuda:2'), covar=tensor([0.1877, 0.0532, 0.1037, 0.1008, 0.2077, 0.1267, 0.0287, 0.0297], device='cuda:2'), in_proj_covar=tensor([0.0256, 0.0234, 0.0247, 0.0202, 0.0277, 0.0187, 0.0212, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:51:39,439 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-27 21:51:42,421 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:09,941 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:20,009 INFO [train.py:904] (2/8) Epoch 3, batch 2300, loss[loss=0.2624, simple_loss=0.3198, pruned_loss=0.1025, over 16720.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3176, pruned_loss=0.08903, over 3333910.67 frames. ], batch size: 134, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:52:57,414 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:59,963 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-04-27 21:53:01,630 INFO [optim.py:368] (2/8) 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 21:53:11,965 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:30,352 INFO [train.py:904] (2/8) Epoch 3, batch 2350, loss[loss=0.2533, simple_loss=0.3356, pruned_loss=0.08556, over 16746.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.319, pruned_loss=0.09059, over 3325080.59 frames. ], batch size: 57, lr: 2.16e-02, grad_scale: 4.0 2023-04-27 21:53:38,443 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 21:54:01,190 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:54:11,821 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5279, 4.4556, 3.7034, 1.5918, 2.6736, 2.2327, 3.9939, 4.4341], device='cuda:2'), covar=tensor([0.0246, 0.0443, 0.0515, 0.1831, 0.0893, 0.1060, 0.0573, 0.0523], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0125, 0.0158, 0.0147, 0.0138, 0.0132, 0.0150, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 21:54:33,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8418, 4.7044, 2.0673, 4.9127, 2.9084, 4.7848, 2.6881, 3.4162], device='cuda:2'), covar=tensor([0.0035, 0.0134, 0.1425, 0.0025, 0.0732, 0.0226, 0.1129, 0.0454], device='cuda:2'), in_proj_covar=tensor([0.0090, 0.0136, 0.0172, 0.0086, 0.0162, 0.0170, 0.0181, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-27 21:54:35,036 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 21:54:35,731 INFO [train.py:904] (2/8) Epoch 3, batch 2400, loss[loss=0.2963, simple_loss=0.3504, pruned_loss=0.1211, over 16382.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3196, pruned_loss=0.09088, over 3320659.74 frames. ], batch size: 146, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:54:38,649 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0037, 4.8603, 4.7501, 4.1569, 4.7704, 1.9650, 4.4780, 4.9977], device='cuda:2'), covar=tensor([0.0071, 0.0061, 0.0081, 0.0338, 0.0063, 0.1315, 0.0086, 0.0090], device='cuda:2'), in_proj_covar=tensor([0.0081, 0.0066, 0.0103, 0.0119, 0.0075, 0.0119, 0.0092, 0.0103], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:54:51,851 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0927, 5.6553, 5.5440, 5.4776, 5.3137, 5.9426, 5.6889, 5.4188], device='cuda:2'), covar=tensor([0.0578, 0.1085, 0.1302, 0.1307, 0.2666, 0.0851, 0.0833, 0.1834], device='cuda:2'), in_proj_covar=tensor([0.0229, 0.0331, 0.0308, 0.0286, 0.0377, 0.0324, 0.0262, 0.0376], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 21:55:17,391 INFO [optim.py:368] (2/8) 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 21:55:32,350 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:55:46,915 INFO [train.py:904] (2/8) Epoch 3, batch 2450, loss[loss=0.2114, simple_loss=0.2872, pruned_loss=0.06776, over 15954.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3202, pruned_loss=0.09033, over 3311318.94 frames. ], batch size: 35, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:56:33,299 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:34,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8708, 3.7060, 2.8870, 2.6643, 2.9046, 2.1688, 3.8800, 4.2029], device='cuda:2'), covar=tensor([0.1511, 0.0542, 0.0946, 0.0859, 0.1627, 0.1208, 0.0322, 0.0387], device='cuda:2'), in_proj_covar=tensor([0.0259, 0.0239, 0.0251, 0.0205, 0.0284, 0.0188, 0.0214, 0.0206], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 21:56:46,161 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:55,061 INFO [train.py:904] (2/8) Epoch 3, batch 2500, loss[loss=0.2158, simple_loss=0.2998, pruned_loss=0.06595, over 17197.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3191, pruned_loss=0.08911, over 3325553.62 frames. ], batch size: 46, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:57:27,650 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:33,675 INFO [optim.py:368] (2/8) 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 21:57:36,907 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:41,075 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:56,199 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7418, 4.3416, 4.0826, 1.3070, 4.5163, 4.1482, 3.3161, 3.3508], device='cuda:2'), covar=tensor([0.0949, 0.0084, 0.0180, 0.1736, 0.0054, 0.0071, 0.0336, 0.0433], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0084, 0.0085, 0.0148, 0.0077, 0.0076, 0.0117, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 21:58:03,439 INFO [train.py:904] (2/8) Epoch 3, batch 2550, loss[loss=0.2138, simple_loss=0.2994, pruned_loss=0.0641, over 17139.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3177, pruned_loss=0.08747, over 3323065.43 frames. ], batch size: 47, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:58:10,934 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:31,384 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:44,552 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:52,523 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:01,461 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:13,834 INFO [train.py:904] (2/8) Epoch 3, batch 2600, loss[loss=0.2581, simple_loss=0.3214, pruned_loss=0.09745, over 16901.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3177, pruned_loss=0.08694, over 3319877.51 frames. ], batch size: 109, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:59:32,280 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3398, 4.1335, 3.7239, 1.8008, 2.7785, 2.3671, 3.6184, 4.2252], device='cuda:2'), covar=tensor([0.0221, 0.0401, 0.0481, 0.1559, 0.0766, 0.0970, 0.0637, 0.0416], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0126, 0.0159, 0.0147, 0.0139, 0.0133, 0.0150, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 21:59:38,824 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:53,532 INFO [optim.py:368] (2/8) 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 22:00:09,152 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:00:22,555 INFO [train.py:904] (2/8) Epoch 3, batch 2650, loss[loss=0.2176, simple_loss=0.2901, pruned_loss=0.07251, over 16847.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3188, pruned_loss=0.08739, over 3321750.45 frames. ], batch size: 42, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:00:55,776 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0749, 4.9365, 4.9910, 3.7015, 4.8873, 1.7134, 4.6183, 5.0670], device='cuda:2'), covar=tensor([0.0097, 0.0067, 0.0078, 0.0520, 0.0072, 0.1721, 0.0111, 0.0146], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0065, 0.0101, 0.0117, 0.0075, 0.0116, 0.0091, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:01:22,673 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:01:30,281 INFO [train.py:904] (2/8) Epoch 3, batch 2700, loss[loss=0.2686, simple_loss=0.3492, pruned_loss=0.09398, over 17031.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3195, pruned_loss=0.08724, over 3326489.35 frames. ], batch size: 53, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:01:55,567 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9531, 4.8536, 4.6425, 4.0594, 4.7866, 2.0040, 4.4461, 4.7788], device='cuda:2'), covar=tensor([0.0049, 0.0040, 0.0074, 0.0280, 0.0050, 0.1167, 0.0074, 0.0081], device='cuda:2'), in_proj_covar=tensor([0.0077, 0.0064, 0.0102, 0.0117, 0.0075, 0.0115, 0.0091, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:02:09,771 INFO [optim.py:368] (2/8) 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 22:02:24,012 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:37,954 INFO [train.py:904] (2/8) Epoch 3, batch 2750, loss[loss=0.2616, simple_loss=0.3204, pruned_loss=0.1013, over 16328.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3202, pruned_loss=0.08747, over 3331571.37 frames. ], batch size: 165, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:03:28,906 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:03:45,505 INFO [train.py:904] (2/8) Epoch 3, batch 2800, loss[loss=0.2131, simple_loss=0.2932, pruned_loss=0.06652, over 16096.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3199, pruned_loss=0.08718, over 3330484.59 frames. ], batch size: 35, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:03:54,989 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:25,544 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:55,613 INFO [train.py:904] (2/8) Epoch 3, batch 2850, loss[loss=0.3738, simple_loss=0.4018, pruned_loss=0.1729, over 12098.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3202, pruned_loss=0.08792, over 3324196.51 frames. ], batch size: 247, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:05:06,838 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7935, 4.6673, 4.5807, 4.0064, 4.6206, 1.9148, 4.3684, 4.6976], device='cuda:2'), covar=tensor([0.0064, 0.0061, 0.0079, 0.0316, 0.0050, 0.1274, 0.0079, 0.0101], device='cuda:2'), in_proj_covar=tensor([0.0077, 0.0066, 0.0102, 0.0119, 0.0076, 0.0117, 0.0093, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:05:20,396 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:36,686 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 3, batch 2900, loss[loss=0.2023, simple_loss=0.2771, pruned_loss=0.06374, over 16798.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3188, pruned_loss=0.08812, over 3319091.36 frames. ], batch size: 39, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:06:07,294 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0114, 5.4355, 5.3788, 5.2808, 5.2766, 5.9172, 5.6174, 5.3013], device='cuda:2'), covar=tensor([0.0643, 0.1180, 0.1090, 0.1578, 0.2552, 0.0769, 0.0850, 0.2109], device='cuda:2'), in_proj_covar=tensor([0.0230, 0.0327, 0.0302, 0.0283, 0.0369, 0.0320, 0.0260, 0.0384], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:06:10,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3682, 4.0367, 3.3416, 1.9896, 2.9421, 2.2736, 3.8489, 3.9922], device='cuda:2'), covar=tensor([0.0196, 0.0421, 0.0470, 0.1447, 0.0601, 0.0949, 0.0412, 0.0415], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0121, 0.0156, 0.0144, 0.0135, 0.0130, 0.0146, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 22:06:25,778 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:42,323 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 22:07:12,200 INFO [train.py:904] (2/8) Epoch 3, batch 2950, loss[loss=0.2592, simple_loss=0.3145, pruned_loss=0.102, over 16880.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3182, pruned_loss=0.08925, over 3324484.92 frames. ], batch size: 90, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:07:49,173 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:07,114 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:12,745 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:19,857 INFO [train.py:904] (2/8) Epoch 3, batch 3000, loss[loss=0.2058, simple_loss=0.2886, pruned_loss=0.06151, over 16879.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3182, pruned_loss=0.0894, over 3318996.54 frames. ], batch size: 42, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:08:19,857 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 22:08:30,489 INFO [train.py:938] (2/8) Epoch 3, validation: loss=0.1721, simple_loss=0.279, pruned_loss=0.03262, over 944034.00 frames. 2023-04-27 22:08:30,489 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 22:09:10,134 INFO [optim.py:368] (2/8) 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 22:09:25,505 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:37,530 INFO [train.py:904] (2/8) Epoch 3, batch 3050, loss[loss=0.1967, simple_loss=0.2701, pruned_loss=0.06161, over 16790.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3169, pruned_loss=0.08867, over 3319681.40 frames. ], batch size: 39, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:10:35,187 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7444, 3.5060, 3.5107, 3.5645, 3.6018, 4.0830, 3.8878, 3.5814], device='cuda:2'), covar=tensor([0.2120, 0.1577, 0.1561, 0.2097, 0.2824, 0.1434, 0.1097, 0.2337], device='cuda:2'), in_proj_covar=tensor([0.0236, 0.0329, 0.0306, 0.0291, 0.0376, 0.0328, 0.0261, 0.0383], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:10:44,545 INFO [train.py:904] (2/8) Epoch 3, batch 3100, loss[loss=0.2287, simple_loss=0.29, pruned_loss=0.08372, over 16670.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3167, pruned_loss=0.0897, over 3305183.07 frames. ], batch size: 89, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:10:56,653 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 22:11:02,101 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 22:11:28,255 INFO [optim.py:368] (2/8) 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 22:11:45,952 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1957, 4.1500, 4.0412, 4.1246, 3.5927, 4.1699, 4.0138, 3.7859], device='cuda:2'), covar=tensor([0.0402, 0.0233, 0.0212, 0.0174, 0.0894, 0.0229, 0.0397, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0133, 0.0192, 0.0159, 0.0224, 0.0172, 0.0139, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:11:53,309 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:54,087 INFO [train.py:904] (2/8) Epoch 3, batch 3150, loss[loss=0.2557, simple_loss=0.3388, pruned_loss=0.08626, over 17052.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3155, pruned_loss=0.0887, over 3315554.85 frames. ], batch size: 55, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:12:05,819 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9414, 2.7621, 2.4864, 4.3291, 2.1342, 4.0915, 2.4743, 2.5910], device='cuda:2'), covar=tensor([0.0282, 0.0705, 0.0437, 0.0188, 0.1659, 0.0215, 0.0851, 0.1309], device='cuda:2'), in_proj_covar=tensor([0.0252, 0.0227, 0.0187, 0.0253, 0.0294, 0.0198, 0.0215, 0.0290], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 22:12:12,236 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:35,064 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:44,078 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:52,841 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9447, 3.9369, 3.8320, 3.3311, 3.9105, 1.8432, 3.7066, 3.8474], device='cuda:2'), covar=tensor([0.0076, 0.0052, 0.0088, 0.0272, 0.0055, 0.1224, 0.0079, 0.0101], device='cuda:2'), in_proj_covar=tensor([0.0080, 0.0068, 0.0105, 0.0123, 0.0077, 0.0118, 0.0095, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:12:58,215 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:01,351 INFO [train.py:904] (2/8) Epoch 3, batch 3200, loss[loss=0.2272, simple_loss=0.3117, pruned_loss=0.07135, over 17141.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3152, pruned_loss=0.08755, over 3308089.87 frames. ], batch size: 48, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:13:39,248 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 22:13:57,286 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0271, 3.5119, 3.9592, 2.9173, 3.9147, 3.9873, 3.8928, 2.1502], device='cuda:2'), covar=tensor([0.0362, 0.0167, 0.0035, 0.0187, 0.0030, 0.0042, 0.0030, 0.0283], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0052, 0.0058, 0.0105, 0.0052, 0.0060, 0.0059, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:14:06,505 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:14:08,943 INFO [train.py:904] (2/8) Epoch 3, batch 3250, loss[loss=0.2161, simple_loss=0.286, pruned_loss=0.07315, over 16877.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3152, pruned_loss=0.08784, over 3309850.96 frames. ], batch size: 42, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:14:21,010 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 22:14:38,902 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:56,662 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:02,783 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-27 22:15:18,374 INFO [train.py:904] (2/8) Epoch 3, batch 3300, loss[loss=0.2553, simple_loss=0.3198, pruned_loss=0.09537, over 16837.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3169, pruned_loss=0.08905, over 3311494.13 frames. ], batch size: 116, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:15:57,196 INFO [optim.py:368] (2/8) 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 22:16:11,609 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2635, 5.5491, 4.8556, 5.5610, 4.9284, 4.6419, 5.3173, 5.5474], device='cuda:2'), covar=tensor([0.0768, 0.0813, 0.1543, 0.0436, 0.0790, 0.0694, 0.0616, 0.0871], device='cuda:2'), in_proj_covar=tensor([0.0276, 0.0399, 0.0341, 0.0242, 0.0257, 0.0243, 0.0308, 0.0270], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 22:16:24,694 INFO [train.py:904] (2/8) Epoch 3, batch 3350, loss[loss=0.2654, simple_loss=0.3252, pruned_loss=0.1028, over 15535.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3176, pruned_loss=0.08918, over 3317350.70 frames. ], batch size: 191, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:16:34,689 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0741, 4.6925, 4.6290, 1.8328, 5.0170, 4.7932, 3.4133, 3.9474], device='cuda:2'), covar=tensor([0.0633, 0.0073, 0.0110, 0.1311, 0.0031, 0.0039, 0.0298, 0.0270], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0081, 0.0079, 0.0142, 0.0076, 0.0074, 0.0112, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:17:07,021 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 22:17:10,964 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-04-27 22:17:31,089 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1254, 3.3520, 3.3646, 1.4461, 3.5638, 3.4563, 2.8687, 2.6120], device='cuda:2'), covar=tensor([0.0792, 0.0098, 0.0151, 0.1247, 0.0062, 0.0068, 0.0337, 0.0432], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0081, 0.0080, 0.0141, 0.0076, 0.0074, 0.0111, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:17:33,460 INFO [train.py:904] (2/8) Epoch 3, batch 3400, loss[loss=0.2571, simple_loss=0.3408, pruned_loss=0.08666, over 17027.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3168, pruned_loss=0.08828, over 3316545.27 frames. ], batch size: 53, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:18:13,370 INFO [optim.py:368] (2/8) 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 22:18:40,209 INFO [train.py:904] (2/8) Epoch 3, batch 3450, loss[loss=0.235, simple_loss=0.2979, pruned_loss=0.08603, over 16813.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3164, pruned_loss=0.08819, over 3313614.99 frames. ], batch size: 102, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:18:48,775 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 22:18:58,630 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:08,112 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2050, 5.0868, 5.0029, 4.9369, 4.3793, 5.0119, 4.9128, 4.5718], device='cuda:2'), covar=tensor([0.0332, 0.0169, 0.0189, 0.0142, 0.0947, 0.0189, 0.0229, 0.0308], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0135, 0.0195, 0.0162, 0.0229, 0.0174, 0.0139, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:19:47,198 INFO [train.py:904] (2/8) Epoch 3, batch 3500, loss[loss=0.2064, simple_loss=0.2793, pruned_loss=0.06674, over 16779.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3142, pruned_loss=0.08678, over 3308005.43 frames. ], batch size: 39, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:20:04,595 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:31,310 INFO [optim.py:368] (2/8) 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 22:20:48,857 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:20:59,078 INFO [train.py:904] (2/8) Epoch 3, batch 3550, loss[loss=0.2574, simple_loss=0.3206, pruned_loss=0.09711, over 16869.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3113, pruned_loss=0.08463, over 3320728.31 frames. ], batch size: 109, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:21:02,150 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-27 22:21:04,144 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1915, 2.1077, 1.7133, 1.8414, 2.6092, 2.4469, 2.8559, 2.5967], device='cuda:2'), covar=tensor([0.0056, 0.0126, 0.0162, 0.0180, 0.0068, 0.0120, 0.0057, 0.0089], device='cuda:2'), in_proj_covar=tensor([0.0063, 0.0119, 0.0117, 0.0118, 0.0112, 0.0121, 0.0081, 0.0098], device='cuda:2'), out_proj_covar=tensor([9.3196e-05, 1.7270e-04, 1.6438e-04, 1.7178e-04, 1.6686e-04, 1.7935e-04, 1.2084e-04, 1.5055e-04], device='cuda:2') 2023-04-27 22:21:29,317 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:45,206 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:05,751 INFO [train.py:904] (2/8) Epoch 3, batch 3600, loss[loss=0.2258, simple_loss=0.3033, pruned_loss=0.07413, over 16668.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3091, pruned_loss=0.08394, over 3318300.36 frames. ], batch size: 57, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:22:33,415 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:47,017 INFO [optim.py:368] (2/8) 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 22:22:50,341 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:13,779 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4759, 4.4894, 4.4341, 4.4475, 3.7809, 4.5485, 4.4340, 4.1841], device='cuda:2'), covar=tensor([0.0461, 0.0273, 0.0246, 0.0218, 0.0867, 0.0259, 0.0290, 0.0304], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0136, 0.0193, 0.0161, 0.0227, 0.0171, 0.0139, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:23:14,441 INFO [train.py:904] (2/8) Epoch 3, batch 3650, loss[loss=0.2286, simple_loss=0.2849, pruned_loss=0.0861, over 16780.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3087, pruned_loss=0.08493, over 3300952.42 frames. ], batch size: 102, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:23:51,328 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-27 22:24:16,599 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-27 22:24:17,392 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6538, 4.0436, 4.2992, 1.4750, 4.4826, 4.5473, 3.1646, 3.1518], device='cuda:2'), covar=tensor([0.0704, 0.0110, 0.0151, 0.1505, 0.0059, 0.0048, 0.0321, 0.0423], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0082, 0.0080, 0.0143, 0.0079, 0.0074, 0.0115, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:24:22,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5515, 4.5282, 4.5290, 4.0199, 4.4785, 2.0150, 4.3520, 4.5251], device='cuda:2'), covar=tensor([0.0054, 0.0046, 0.0058, 0.0224, 0.0039, 0.1057, 0.0058, 0.0082], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0066, 0.0099, 0.0113, 0.0073, 0.0111, 0.0090, 0.0101], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:24:29,887 INFO [train.py:904] (2/8) Epoch 3, batch 3700, loss[loss=0.2542, simple_loss=0.3017, pruned_loss=0.1034, over 16710.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3077, pruned_loss=0.08646, over 3283422.52 frames. ], batch size: 83, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:25:13,796 INFO [optim.py:368] (2/8) 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 22:25:22,523 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:42,969 INFO [train.py:904] (2/8) Epoch 3, batch 3750, loss[loss=0.225, simple_loss=0.3024, pruned_loss=0.07375, over 17232.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3085, pruned_loss=0.08816, over 3265951.66 frames. ], batch size: 45, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:26:46,892 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:26:50,661 INFO [train.py:904] (2/8) Epoch 3, batch 3800, loss[loss=0.2213, simple_loss=0.2859, pruned_loss=0.07839, over 16641.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3096, pruned_loss=0.08961, over 3264778.91 frames. ], batch size: 76, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:27:20,393 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9218, 3.9801, 2.1314, 3.9087, 2.8067, 3.9383, 1.7755, 3.0638], device='cuda:2'), covar=tensor([0.0045, 0.0133, 0.1187, 0.0046, 0.0538, 0.0330, 0.1325, 0.0433], device='cuda:2'), in_proj_covar=tensor([0.0088, 0.0134, 0.0167, 0.0081, 0.0157, 0.0165, 0.0177, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-27 22:27:34,019 INFO [optim.py:368] (2/8) 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 22:27:52,796 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:28:01,891 INFO [train.py:904] (2/8) Epoch 3, batch 3850, loss[loss=0.249, simple_loss=0.3059, pruned_loss=0.09604, over 16862.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3084, pruned_loss=0.08915, over 3275423.04 frames. ], batch size: 116, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:28:02,292 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8617, 5.3048, 5.3045, 5.3841, 5.2441, 5.8289, 5.5283, 5.3027], device='cuda:2'), covar=tensor([0.0615, 0.1024, 0.1042, 0.1220, 0.1663, 0.0734, 0.0733, 0.1676], device='cuda:2'), in_proj_covar=tensor([0.0228, 0.0322, 0.0299, 0.0283, 0.0357, 0.0316, 0.0258, 0.0377], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:28:22,684 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 22:29:01,360 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:13,623 INFO [train.py:904] (2/8) Epoch 3, batch 3900, loss[loss=0.2604, simple_loss=0.3163, pruned_loss=0.1023, over 16762.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3078, pruned_loss=0.08961, over 3277068.40 frames. ], batch size: 124, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:29:56,950 INFO [optim.py:368] (2/8) 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 22:30:25,197 INFO [train.py:904] (2/8) Epoch 3, batch 3950, loss[loss=0.2537, simple_loss=0.3142, pruned_loss=0.09666, over 16546.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3071, pruned_loss=0.08999, over 3282800.79 frames. ], batch size: 68, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:30:39,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4428, 2.1670, 1.3913, 2.1882, 2.9422, 2.3784, 3.7724, 3.4030], device='cuda:2'), covar=tensor([0.0014, 0.0126, 0.0190, 0.0135, 0.0060, 0.0120, 0.0018, 0.0038], device='cuda:2'), in_proj_covar=tensor([0.0060, 0.0119, 0.0120, 0.0118, 0.0111, 0.0119, 0.0079, 0.0096], device='cuda:2'), out_proj_covar=tensor([8.9515e-05, 1.7342e-04, 1.6913e-04, 1.7078e-04, 1.6446e-04, 1.7488e-04, 1.1744e-04, 1.4541e-04], device='cuda:2') 2023-04-27 22:31:34,850 INFO [train.py:904] (2/8) Epoch 3, batch 4000, loss[loss=0.2575, simple_loss=0.3243, pruned_loss=0.0953, over 16425.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3063, pruned_loss=0.08979, over 3291487.86 frames. ], batch size: 146, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:31:55,063 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-04-27 22:32:17,081 INFO [optim.py:368] (2/8) 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 22:32:45,199 INFO [train.py:904] (2/8) Epoch 3, batch 4050, loss[loss=0.2147, simple_loss=0.2861, pruned_loss=0.07167, over 16706.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3045, pruned_loss=0.087, over 3293621.20 frames. ], batch size: 57, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:32:47,082 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9530, 4.2752, 3.5064, 2.5706, 3.2107, 2.4434, 4.5515, 5.0026], device='cuda:2'), covar=tensor([0.2019, 0.0539, 0.0985, 0.1037, 0.2158, 0.1170, 0.0303, 0.0144], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0242, 0.0258, 0.0213, 0.0300, 0.0196, 0.0224, 0.0218], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 22:33:46,370 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:33:58,259 INFO [train.py:904] (2/8) Epoch 3, batch 4100, loss[loss=0.2264, simple_loss=0.3027, pruned_loss=0.07504, over 16586.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3048, pruned_loss=0.08516, over 3287214.08 frames. ], batch size: 57, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:34:42,994 INFO [optim.py:368] (2/8) 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 22:35:13,089 INFO [train.py:904] (2/8) Epoch 3, batch 4150, loss[loss=0.2652, simple_loss=0.3466, pruned_loss=0.09195, over 16632.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3143, pruned_loss=0.08996, over 3244177.56 frames. ], batch size: 89, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:35:15,071 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 22:36:27,611 INFO [train.py:904] (2/8) Epoch 3, batch 4200, loss[loss=0.3261, simple_loss=0.374, pruned_loss=0.1392, over 11776.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3227, pruned_loss=0.09275, over 3224898.87 frames. ], batch size: 246, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:37:10,951 INFO [optim.py:368] (2/8) 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 22:37:30,992 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5134, 3.4997, 3.1358, 2.3688, 2.5605, 2.0917, 3.4860, 3.8526], device='cuda:2'), covar=tensor([0.1816, 0.0567, 0.0826, 0.1119, 0.1651, 0.1224, 0.0345, 0.0266], device='cuda:2'), in_proj_covar=tensor([0.0259, 0.0236, 0.0247, 0.0203, 0.0285, 0.0185, 0.0213, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 22:37:35,799 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:37:40,125 INFO [train.py:904] (2/8) Epoch 3, batch 4250, loss[loss=0.2133, simple_loss=0.3041, pruned_loss=0.06126, over 16774.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.325, pruned_loss=0.09208, over 3212796.66 frames. ], batch size: 83, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:37:43,074 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4706, 4.7758, 4.6762, 4.7128, 4.7804, 5.3163, 5.0028, 4.7305], device='cuda:2'), covar=tensor([0.0742, 0.0938, 0.0875, 0.1325, 0.1793, 0.0705, 0.0841, 0.1806], device='cuda:2'), in_proj_covar=tensor([0.0219, 0.0295, 0.0273, 0.0260, 0.0336, 0.0292, 0.0240, 0.0350], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:37:57,132 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9451, 4.1891, 3.6121, 2.6234, 3.3264, 2.6061, 4.1758, 4.7023], device='cuda:2'), covar=tensor([0.1685, 0.0435, 0.0804, 0.1006, 0.1529, 0.0939, 0.0304, 0.0158], device='cuda:2'), in_proj_covar=tensor([0.0261, 0.0235, 0.0247, 0.0204, 0.0285, 0.0185, 0.0214, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 22:38:06,904 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:21,542 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8707, 3.4176, 2.5274, 4.7028, 4.6027, 4.0117, 1.7745, 3.2598], device='cuda:2'), covar=tensor([0.1293, 0.0440, 0.1148, 0.0055, 0.0131, 0.0321, 0.1286, 0.0617], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0127, 0.0158, 0.0070, 0.0127, 0.0131, 0.0149, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:38:40,970 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5256, 3.5986, 2.8790, 2.3115, 2.8054, 2.1319, 3.6388, 4.0693], device='cuda:2'), covar=tensor([0.1871, 0.0630, 0.1023, 0.1046, 0.1732, 0.1142, 0.0371, 0.0272], device='cuda:2'), in_proj_covar=tensor([0.0260, 0.0237, 0.0249, 0.0204, 0.0286, 0.0185, 0.0214, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 22:38:53,670 INFO [train.py:904] (2/8) Epoch 3, batch 4300, loss[loss=0.2449, simple_loss=0.3324, pruned_loss=0.07869, over 16866.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.326, pruned_loss=0.09077, over 3207042.45 frames. ], batch size: 96, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:39:04,988 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:37,547 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:38,216 INFO [optim.py:368] (2/8) 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 22:40:06,091 INFO [train.py:904] (2/8) Epoch 3, batch 4350, loss[loss=0.2716, simple_loss=0.3423, pruned_loss=0.1004, over 16628.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3302, pruned_loss=0.09245, over 3208657.69 frames. ], batch size: 57, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:40:35,029 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8513, 1.7689, 1.2324, 1.5949, 2.3917, 2.1139, 2.6843, 2.6089], device='cuda:2'), covar=tensor([0.0014, 0.0121, 0.0185, 0.0156, 0.0074, 0.0112, 0.0027, 0.0053], device='cuda:2'), in_proj_covar=tensor([0.0054, 0.0116, 0.0119, 0.0116, 0.0109, 0.0118, 0.0074, 0.0094], device='cuda:2'), out_proj_covar=tensor([7.8762e-05, 1.6765e-04, 1.6709e-04, 1.6806e-04, 1.6124e-04, 1.7327e-04, 1.0741e-04, 1.4281e-04], device='cuda:2') 2023-04-27 22:41:10,540 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:41:21,563 INFO [train.py:904] (2/8) Epoch 3, batch 4400, loss[loss=0.2649, simple_loss=0.3469, pruned_loss=0.09142, over 16331.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.3324, pruned_loss=0.09319, over 3207491.39 frames. ], batch size: 35, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:42:05,387 INFO [optim.py:368] (2/8) 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 22:42:20,529 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:42:34,836 INFO [train.py:904] (2/8) Epoch 3, batch 4450, loss[loss=0.2684, simple_loss=0.3456, pruned_loss=0.09562, over 15408.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.335, pruned_loss=0.09339, over 3205183.51 frames. ], batch size: 191, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:42:53,424 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9267, 3.9953, 3.7542, 3.8117, 3.1846, 3.9052, 3.7916, 3.5094], device='cuda:2'), covar=tensor([0.0326, 0.0173, 0.0242, 0.0188, 0.0872, 0.0194, 0.0437, 0.0327], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0117, 0.0165, 0.0137, 0.0196, 0.0146, 0.0123, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:42:55,288 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:43:28,931 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1898, 4.1528, 1.6107, 4.4183, 2.7121, 4.3125, 2.2126, 2.9365], device='cuda:2'), covar=tensor([0.0043, 0.0124, 0.1683, 0.0020, 0.0658, 0.0180, 0.1209, 0.0531], device='cuda:2'), in_proj_covar=tensor([0.0088, 0.0127, 0.0171, 0.0077, 0.0159, 0.0158, 0.0179, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-27 22:43:47,147 INFO [train.py:904] (2/8) Epoch 3, batch 4500, loss[loss=0.2644, simple_loss=0.3251, pruned_loss=0.1018, over 11919.00 frames. ], tot_loss[loss=0.26, simple_loss=0.3343, pruned_loss=0.09287, over 3209005.49 frames. ], batch size: 246, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:44:22,408 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:24,104 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5765, 2.7721, 2.3790, 4.2163, 4.0252, 3.7416, 1.4723, 2.9008], device='cuda:2'), covar=tensor([0.1411, 0.0525, 0.1171, 0.0060, 0.0133, 0.0290, 0.1316, 0.0667], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0130, 0.0163, 0.0070, 0.0130, 0.0134, 0.0154, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:44:29,086 INFO [optim.py:368] (2/8) 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 22:44:52,826 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:56,540 INFO [train.py:904] (2/8) Epoch 3, batch 4550, loss[loss=0.2729, simple_loss=0.3461, pruned_loss=0.09985, over 16681.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3343, pruned_loss=0.09314, over 3218289.27 frames. ], batch size: 134, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:07,316 INFO [train.py:904] (2/8) Epoch 3, batch 4600, loss[loss=0.2654, simple_loss=0.3345, pruned_loss=0.09815, over 16648.00 frames. ], tot_loss[loss=0.259, simple_loss=0.334, pruned_loss=0.092, over 3225412.28 frames. ], batch size: 124, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:10,703 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:20,473 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:44,270 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:52,471 INFO [optim.py:368] (2/8) 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 22:47:20,338 INFO [train.py:904] (2/8) Epoch 3, batch 4650, loss[loss=0.2418, simple_loss=0.323, pruned_loss=0.08031, over 16695.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3329, pruned_loss=0.09142, over 3225256.92 frames. ], batch size: 124, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:47:44,255 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1638, 3.2531, 1.5217, 3.3029, 2.2676, 3.2543, 1.7834, 2.5507], device='cuda:2'), covar=tensor([0.0075, 0.0150, 0.1524, 0.0041, 0.0668, 0.0245, 0.1196, 0.0599], device='cuda:2'), in_proj_covar=tensor([0.0088, 0.0125, 0.0169, 0.0076, 0.0157, 0.0156, 0.0176, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-27 22:48:24,033 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:48:32,747 INFO [train.py:904] (2/8) Epoch 3, batch 4700, loss[loss=0.2726, simple_loss=0.3515, pruned_loss=0.09682, over 16771.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3301, pruned_loss=0.09, over 3212197.39 frames. ], batch size: 124, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:16,885 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 22:49:17,307 INFO [optim.py:368] (2/8) 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 22:49:31,905 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1898, 3.5857, 3.4934, 1.5417, 3.7619, 3.7642, 2.9431, 2.7268], device='cuda:2'), covar=tensor([0.0817, 0.0091, 0.0165, 0.1338, 0.0078, 0.0051, 0.0296, 0.0376], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0080, 0.0079, 0.0145, 0.0072, 0.0071, 0.0112, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:49:45,101 INFO [train.py:904] (2/8) Epoch 3, batch 4750, loss[loss=0.2102, simple_loss=0.2886, pruned_loss=0.06591, over 16931.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3259, pruned_loss=0.08788, over 3211938.97 frames. ], batch size: 96, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:52,751 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:50:58,607 INFO [train.py:904] (2/8) Epoch 3, batch 4800, loss[loss=0.2334, simple_loss=0.3123, pruned_loss=0.07731, over 16751.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3222, pruned_loss=0.08555, over 3204893.14 frames. ], batch size: 89, lr: 2.06e-02, grad_scale: 8.0 2023-04-27 22:51:15,612 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1614, 3.5928, 3.2237, 5.0253, 4.8409, 4.4774, 1.5362, 3.8383], device='cuda:2'), covar=tensor([0.1119, 0.0356, 0.0701, 0.0041, 0.0077, 0.0218, 0.1283, 0.0370], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0130, 0.0164, 0.0071, 0.0130, 0.0135, 0.0154, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:51:28,791 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:51:33,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3907, 4.2804, 4.8242, 4.8905, 4.8311, 4.2683, 4.4511, 4.2537], device='cuda:2'), covar=tensor([0.0186, 0.0269, 0.0273, 0.0288, 0.0304, 0.0235, 0.0598, 0.0281], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0165, 0.0184, 0.0178, 0.0223, 0.0183, 0.0280, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-27 22:51:47,435 INFO [optim.py:368] (2/8) 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 22:52:13,128 INFO [train.py:904] (2/8) Epoch 3, batch 4850, loss[loss=0.2472, simple_loss=0.327, pruned_loss=0.08367, over 16732.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3236, pruned_loss=0.08585, over 3184516.54 frames. ], batch size: 89, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:21,204 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6782, 3.0839, 2.3585, 4.4865, 4.2680, 4.0373, 1.9359, 2.8873], device='cuda:2'), covar=tensor([0.1606, 0.0500, 0.1265, 0.0068, 0.0142, 0.0218, 0.1331, 0.0745], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0127, 0.0162, 0.0069, 0.0128, 0.0133, 0.0152, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 22:53:25,040 INFO [train.py:904] (2/8) Epoch 3, batch 4900, loss[loss=0.2873, simple_loss=0.3594, pruned_loss=0.1076, over 11962.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.324, pruned_loss=0.08545, over 3161858.59 frames. ], batch size: 248, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:28,450 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:29,419 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:58,874 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:54:10,579 INFO [optim.py:368] (2/8) 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 22:54:35,308 INFO [train.py:904] (2/8) Epoch 3, batch 4950, loss[loss=0.2996, simple_loss=0.3764, pruned_loss=0.1114, over 16225.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3237, pruned_loss=0.08503, over 3172877.19 frames. ], batch size: 165, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:54:35,558 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:54:35,916 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-04-27 22:54:39,245 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 22:55:08,062 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:48,338 INFO [train.py:904] (2/8) Epoch 3, batch 5000, loss[loss=0.2337, simple_loss=0.3149, pruned_loss=0.07625, over 17197.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.325, pruned_loss=0.08531, over 3180199.47 frames. ], batch size: 44, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:35,248 INFO [optim.py:368] (2/8) 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 22:56:59,650 INFO [train.py:904] (2/8) Epoch 3, batch 5050, loss[loss=0.2299, simple_loss=0.3135, pruned_loss=0.07318, over 16522.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3247, pruned_loss=0.08442, over 3200456.25 frames. ], batch size: 75, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:59,957 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 22:57:50,426 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5054, 2.5942, 2.2161, 4.1938, 1.8329, 3.6557, 2.2569, 2.3376], device='cuda:2'), covar=tensor([0.0339, 0.0750, 0.0524, 0.0162, 0.1886, 0.0268, 0.1032, 0.1361], device='cuda:2'), in_proj_covar=tensor([0.0255, 0.0233, 0.0196, 0.0256, 0.0304, 0.0206, 0.0224, 0.0304], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 22:58:08,623 INFO [train.py:904] (2/8) Epoch 3, batch 5100, loss[loss=0.2182, simple_loss=0.2932, pruned_loss=0.07155, over 16664.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.323, pruned_loss=0.08388, over 3195422.95 frames. ], batch size: 62, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:58:11,904 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 22:58:38,834 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:58:47,802 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:58:57,532 INFO [optim.py:368] (2/8) 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 22:59:23,205 INFO [train.py:904] (2/8) Epoch 3, batch 5150, loss[loss=0.2625, simple_loss=0.3451, pruned_loss=0.08991, over 16935.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3225, pruned_loss=0.08288, over 3188771.92 frames. ], batch size: 109, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:59:25,501 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2691, 5.0070, 5.0371, 5.0893, 4.5584, 5.0525, 4.9951, 4.7331], device='cuda:2'), covar=tensor([0.0258, 0.0152, 0.0128, 0.0095, 0.0692, 0.0155, 0.0117, 0.0256], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0122, 0.0170, 0.0140, 0.0197, 0.0157, 0.0125, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 22:59:50,322 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:00:19,404 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:00:35,985 INFO [train.py:904] (2/8) Epoch 3, batch 5200, loss[loss=0.2108, simple_loss=0.2871, pruned_loss=0.06727, over 16480.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3214, pruned_loss=0.08281, over 3197663.12 frames. ], batch size: 68, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:00:40,381 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:01:22,143 INFO [optim.py:368] (2/8) 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 23:01:45,790 INFO [train.py:904] (2/8) Epoch 3, batch 5250, loss[loss=0.2727, simple_loss=0.3313, pruned_loss=0.107, over 12223.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3186, pruned_loss=0.08238, over 3195117.67 frames. ], batch size: 246, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:01:47,881 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:02:00,420 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-27 23:02:56,098 INFO [train.py:904] (2/8) Epoch 3, batch 5300, loss[loss=0.273, simple_loss=0.333, pruned_loss=0.1065, over 12242.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3154, pruned_loss=0.08117, over 3190654.58 frames. ], batch size: 247, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:03:14,348 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1582, 4.9955, 4.8569, 4.2670, 5.0878, 1.7252, 4.7387, 5.1019], device='cuda:2'), covar=tensor([0.0051, 0.0041, 0.0057, 0.0295, 0.0040, 0.1329, 0.0066, 0.0075], device='cuda:2'), in_proj_covar=tensor([0.0073, 0.0061, 0.0095, 0.0113, 0.0071, 0.0118, 0.0084, 0.0096], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:03:43,229 INFO [optim.py:368] (2/8) 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 23:04:08,022 INFO [train.py:904] (2/8) Epoch 3, batch 5350, loss[loss=0.2376, simple_loss=0.3173, pruned_loss=0.07896, over 16766.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3126, pruned_loss=0.07997, over 3193170.01 frames. ], batch size: 124, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:04:08,380 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:04:12,380 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 23:05:04,083 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.49 vs. limit=5.0 2023-04-27 23:05:17,183 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:05:19,288 INFO [train.py:904] (2/8) Epoch 3, batch 5400, loss[loss=0.2957, simple_loss=0.357, pruned_loss=0.1172, over 12114.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.316, pruned_loss=0.08123, over 3194696.83 frames. ], batch size: 246, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:06:06,451 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7183, 5.1480, 5.2654, 5.2176, 5.1009, 5.6767, 5.2562, 5.0187], device='cuda:2'), covar=tensor([0.0577, 0.0935, 0.0670, 0.1107, 0.1616, 0.0519, 0.0679, 0.1629], device='cuda:2'), in_proj_covar=tensor([0.0222, 0.0294, 0.0272, 0.0266, 0.0339, 0.0297, 0.0238, 0.0354], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:06:07,786 INFO [optim.py:368] (2/8) 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 23:06:34,320 INFO [train.py:904] (2/8) Epoch 3, batch 5450, loss[loss=0.3101, simple_loss=0.3771, pruned_loss=0.1216, over 16341.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.32, pruned_loss=0.08384, over 3185404.55 frames. ], batch size: 146, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:07:24,680 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:07:49,275 INFO [train.py:904] (2/8) Epoch 3, batch 5500, loss[loss=0.3848, simple_loss=0.4124, pruned_loss=0.1786, over 11478.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.3302, pruned_loss=0.09196, over 3143941.56 frames. ], batch size: 248, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:08:06,176 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 23:08:39,223 INFO [optim.py:368] (2/8) 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 23:08:57,740 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9479, 2.3775, 2.2085, 3.1317, 2.0465, 3.0003, 2.2863, 1.9679], device='cuda:2'), covar=tensor([0.0285, 0.0617, 0.0340, 0.0187, 0.1210, 0.0205, 0.0728, 0.1045], device='cuda:2'), in_proj_covar=tensor([0.0253, 0.0228, 0.0192, 0.0249, 0.0298, 0.0203, 0.0220, 0.0291], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:09:06,198 INFO [train.py:904] (2/8) Epoch 3, batch 5550, loss[loss=0.3553, simple_loss=0.4047, pruned_loss=0.153, over 16327.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3406, pruned_loss=0.1003, over 3125719.27 frames. ], batch size: 165, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:09:17,598 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:10:25,185 INFO [train.py:904] (2/8) Epoch 3, batch 5600, loss[loss=0.2768, simple_loss=0.3494, pruned_loss=0.1021, over 17104.00 frames. ], tot_loss[loss=0.2801, simple_loss=0.3473, pruned_loss=0.1065, over 3105752.46 frames. ], batch size: 47, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:10:37,557 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2042, 4.0390, 2.4735, 5.2135, 5.0148, 4.5491, 2.1390, 3.3180], device='cuda:2'), covar=tensor([0.1177, 0.0306, 0.1236, 0.0048, 0.0137, 0.0269, 0.1037, 0.0607], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0130, 0.0166, 0.0072, 0.0134, 0.0143, 0.0156, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 23:10:56,273 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:11:02,402 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1957, 2.9146, 2.6896, 2.0346, 2.5649, 2.0900, 2.7865, 2.9983], device='cuda:2'), covar=tensor([0.0266, 0.0430, 0.0377, 0.1224, 0.0577, 0.0804, 0.0517, 0.0426], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0114, 0.0154, 0.0144, 0.0138, 0.0130, 0.0145, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-27 23:11:14,477 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 23:11:21,465 INFO [optim.py:368] (2/8) 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 23:11:48,512 INFO [train.py:904] (2/8) Epoch 3, batch 5650, loss[loss=0.3363, simple_loss=0.3887, pruned_loss=0.1419, over 16432.00 frames. ], tot_loss[loss=0.2911, simple_loss=0.355, pruned_loss=0.1136, over 3065972.64 frames. ], batch size: 146, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:13:10,043 INFO [train.py:904] (2/8) Epoch 3, batch 5700, loss[loss=0.3005, simple_loss=0.3703, pruned_loss=0.1154, over 16255.00 frames. ], tot_loss[loss=0.292, simple_loss=0.3555, pruned_loss=0.1143, over 3071641.27 frames. ], batch size: 165, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:14:00,537 INFO [optim.py:368] (2/8) 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 23:14:27,438 INFO [train.py:904] (2/8) Epoch 3, batch 5750, loss[loss=0.3156, simple_loss=0.3756, pruned_loss=0.1278, over 16204.00 frames. ], tot_loss[loss=0.2964, simple_loss=0.3594, pruned_loss=0.1167, over 3040485.33 frames. ], batch size: 165, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:22,081 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:15:47,178 INFO [train.py:904] (2/8) Epoch 3, batch 5800, loss[loss=0.2462, simple_loss=0.3233, pruned_loss=0.0845, over 16562.00 frames. ], tot_loss[loss=0.2936, simple_loss=0.3584, pruned_loss=0.1144, over 3052487.30 frames. ], batch size: 62, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:48,987 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:16:36,164 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:16:38,587 INFO [optim.py:368] (2/8) 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 23:16:41,890 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-27 23:17:05,080 INFO [train.py:904] (2/8) Epoch 3, batch 5850, loss[loss=0.321, simple_loss=0.3733, pruned_loss=0.1343, over 16354.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3561, pruned_loss=0.112, over 3060284.76 frames. ], batch size: 146, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:17:24,195 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:31,371 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:37,194 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9258, 4.7077, 4.8786, 5.1744, 5.2574, 4.5741, 5.2758, 5.2122], device='cuda:2'), covar=tensor([0.0570, 0.0525, 0.0889, 0.0297, 0.0284, 0.0425, 0.0259, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0339, 0.0446, 0.0339, 0.0252, 0.0240, 0.0275, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:18:26,955 INFO [train.py:904] (2/8) Epoch 3, batch 5900, loss[loss=0.2828, simple_loss=0.3629, pruned_loss=0.1014, over 16685.00 frames. ], tot_loss[loss=0.2858, simple_loss=0.3533, pruned_loss=0.1091, over 3106527.18 frames. ], batch size: 57, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:18:51,473 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:19:15,328 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:22,627 INFO [optim.py:368] (2/8) 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 23:19:47,804 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:48,582 INFO [train.py:904] (2/8) Epoch 3, batch 5950, loss[loss=0.2855, simple_loss=0.3606, pruned_loss=0.1052, over 16902.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3546, pruned_loss=0.1081, over 3102837.63 frames. ], batch size: 109, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:20:25,196 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 23:20:57,498 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3781, 5.2839, 5.0015, 4.2933, 5.0732, 2.0458, 4.7937, 5.1135], device='cuda:2'), covar=tensor([0.0041, 0.0035, 0.0066, 0.0284, 0.0046, 0.1323, 0.0054, 0.0069], device='cuda:2'), in_proj_covar=tensor([0.0073, 0.0062, 0.0095, 0.0112, 0.0071, 0.0121, 0.0082, 0.0094], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:21:07,934 INFO [train.py:904] (2/8) Epoch 3, batch 6000, loss[loss=0.2704, simple_loss=0.3462, pruned_loss=0.09729, over 16688.00 frames. ], tot_loss[loss=0.2866, simple_loss=0.3549, pruned_loss=0.1092, over 3076404.34 frames. ], batch size: 124, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:21:07,934 INFO [train.py:929] (2/8) Computing validation loss 2023-04-27 23:21:18,884 INFO [train.py:938] (2/8) Epoch 3, validation: loss=0.2097, simple_loss=0.3184, pruned_loss=0.05055, over 944034.00 frames. 2023-04-27 23:21:18,884 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-27 23:21:34,587 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:21:41,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6729, 3.1084, 2.3687, 3.9884, 3.8170, 3.7528, 1.6827, 2.7696], device='cuda:2'), covar=tensor([0.1319, 0.0386, 0.1118, 0.0065, 0.0211, 0.0243, 0.1187, 0.0640], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0131, 0.0165, 0.0071, 0.0136, 0.0141, 0.0156, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-27 23:21:59,548 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6957, 2.6935, 1.5653, 2.7398, 2.0475, 2.7287, 1.8774, 2.4304], device='cuda:2'), covar=tensor([0.0118, 0.0324, 0.1283, 0.0070, 0.0727, 0.0480, 0.1056, 0.0493], device='cuda:2'), in_proj_covar=tensor([0.0088, 0.0133, 0.0172, 0.0080, 0.0164, 0.0164, 0.0180, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-27 23:22:07,348 INFO [optim.py:368] (2/8) 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 23:22:28,943 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-04-27 23:22:36,223 INFO [train.py:904] (2/8) Epoch 3, batch 6050, loss[loss=0.2598, simple_loss=0.3449, pruned_loss=0.08733, over 16790.00 frames. ], tot_loss[loss=0.2844, simple_loss=0.3528, pruned_loss=0.108, over 3080743.49 frames. ], batch size: 83, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:23:51,471 INFO [train.py:904] (2/8) Epoch 3, batch 6100, loss[loss=0.2601, simple_loss=0.3308, pruned_loss=0.0947, over 17179.00 frames. ], tot_loss[loss=0.2822, simple_loss=0.3522, pruned_loss=0.1061, over 3111551.14 frames. ], batch size: 46, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:23:59,706 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1202, 4.0987, 4.2155, 4.1889, 4.2040, 4.6859, 4.3891, 4.1379], device='cuda:2'), covar=tensor([0.1243, 0.1524, 0.1081, 0.1675, 0.2156, 0.0803, 0.0944, 0.2138], device='cuda:2'), in_proj_covar=tensor([0.0229, 0.0316, 0.0294, 0.0277, 0.0366, 0.0317, 0.0256, 0.0370], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:24:12,655 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:24:38,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7504, 2.6494, 2.6884, 1.8489, 2.3309, 2.5171, 2.2936, 1.7219], device='cuda:2'), covar=tensor([0.0288, 0.0031, 0.0045, 0.0210, 0.0049, 0.0062, 0.0035, 0.0271], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0049, 0.0057, 0.0106, 0.0049, 0.0061, 0.0056, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:24:42,456 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 3, batch 6150, loss[loss=0.3267, simple_loss=0.3791, pruned_loss=0.1371, over 11637.00 frames. ], tot_loss[loss=0.2808, simple_loss=0.3505, pruned_loss=0.1056, over 3101863.34 frames. ], batch size: 250, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:25:23,333 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:25:50,264 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:26:28,207 INFO [train.py:904] (2/8) Epoch 3, batch 6200, loss[loss=0.2547, simple_loss=0.323, pruned_loss=0.09323, over 16760.00 frames. ], tot_loss[loss=0.2784, simple_loss=0.3475, pruned_loss=0.1047, over 3102755.39 frames. ], batch size: 124, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:26:48,657 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:27:03,437 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:27:18,518 INFO [optim.py:368] (2/8) 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 23:27:41,891 INFO [train.py:904] (2/8) Epoch 3, batch 6250, loss[loss=0.2526, simple_loss=0.3307, pruned_loss=0.08726, over 17035.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3472, pruned_loss=0.1047, over 3088871.32 frames. ], batch size: 50, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:27:58,188 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:28:45,262 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2020, 4.5015, 4.2742, 4.4030, 3.1529, 4.4514, 4.4477, 3.9236], device='cuda:2'), covar=tensor([0.0814, 0.0325, 0.0381, 0.0291, 0.1717, 0.0395, 0.0339, 0.0504], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0131, 0.0169, 0.0142, 0.0200, 0.0162, 0.0126, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:28:54,857 INFO [train.py:904] (2/8) Epoch 3, batch 6300, loss[loss=0.2685, simple_loss=0.3439, pruned_loss=0.09654, over 16850.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3472, pruned_loss=0.1047, over 3089755.36 frames. ], batch size: 116, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:29:02,862 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:29:48,249 INFO [optim.py:368] (2/8) 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 23:30:11,908 INFO [train.py:904] (2/8) Epoch 3, batch 6350, loss[loss=0.2689, simple_loss=0.3464, pruned_loss=0.0957, over 16676.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3475, pruned_loss=0.1059, over 3092016.58 frames. ], batch size: 89, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:31:11,138 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6176, 3.4903, 3.6712, 3.6059, 3.6419, 4.0243, 3.9018, 3.5735], device='cuda:2'), covar=tensor([0.1769, 0.1842, 0.1357, 0.2114, 0.3040, 0.1350, 0.1223, 0.2450], device='cuda:2'), in_proj_covar=tensor([0.0230, 0.0311, 0.0287, 0.0273, 0.0355, 0.0313, 0.0254, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:31:25,030 INFO [train.py:904] (2/8) Epoch 3, batch 6400, loss[loss=0.2701, simple_loss=0.3431, pruned_loss=0.09855, over 16889.00 frames. ], tot_loss[loss=0.2814, simple_loss=0.3483, pruned_loss=0.1072, over 3098097.43 frames. ], batch size: 109, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:12,499 INFO [optim.py:368] (2/8) 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 23:32:35,289 INFO [train.py:904] (2/8) Epoch 3, batch 6450, loss[loss=0.2593, simple_loss=0.3306, pruned_loss=0.09398, over 17072.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.3472, pruned_loss=0.1058, over 3091842.63 frames. ], batch size: 53, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:47,308 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:05,687 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:53,754 INFO [train.py:904] (2/8) Epoch 3, batch 6500, loss[loss=0.2463, simple_loss=0.3163, pruned_loss=0.08821, over 16598.00 frames. ], tot_loss[loss=0.277, simple_loss=0.3448, pruned_loss=0.1045, over 3105216.64 frames. ], batch size: 62, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:34:01,035 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:29,502 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:44,550 INFO [optim.py:368] (2/8) 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 23:35:06,591 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4650, 3.4989, 2.7914, 2.2268, 2.6242, 2.1370, 3.6086, 3.8077], device='cuda:2'), covar=tensor([0.1938, 0.0568, 0.1035, 0.1131, 0.1650, 0.1234, 0.0317, 0.0348], device='cuda:2'), in_proj_covar=tensor([0.0267, 0.0238, 0.0253, 0.0212, 0.0294, 0.0191, 0.0215, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:35:12,110 INFO [train.py:904] (2/8) Epoch 3, batch 6550, loss[loss=0.3647, simple_loss=0.405, pruned_loss=0.1622, over 11525.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3477, pruned_loss=0.1056, over 3103769.46 frames. ], batch size: 246, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:35:45,473 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:36:28,904 INFO [train.py:904] (2/8) Epoch 3, batch 6600, loss[loss=0.2914, simple_loss=0.3517, pruned_loss=0.1156, over 16591.00 frames. ], tot_loss[loss=0.2804, simple_loss=0.3494, pruned_loss=0.1057, over 3104409.71 frames. ], batch size: 62, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:36:33,593 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:36:37,652 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:36:47,042 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5058, 4.5589, 5.0897, 5.0307, 5.0398, 4.6249, 4.6386, 4.4447], device='cuda:2'), covar=tensor([0.0203, 0.0212, 0.0213, 0.0298, 0.0352, 0.0234, 0.0757, 0.0367], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0169, 0.0188, 0.0186, 0.0224, 0.0192, 0.0290, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-27 23:36:52,040 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7214, 3.7328, 3.0558, 2.5300, 2.8953, 2.1047, 3.9130, 4.2296], device='cuda:2'), covar=tensor([0.1739, 0.0543, 0.0979, 0.0939, 0.1519, 0.1173, 0.0308, 0.0279], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0238, 0.0256, 0.0213, 0.0295, 0.0192, 0.0219, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:37:20,959 INFO [optim.py:368] (2/8) 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 23:37:45,809 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:46,401 INFO [train.py:904] (2/8) Epoch 3, batch 6650, loss[loss=0.27, simple_loss=0.3325, pruned_loss=0.1037, over 17045.00 frames. ], tot_loss[loss=0.2848, simple_loss=0.3521, pruned_loss=0.1088, over 3083414.49 frames. ], batch size: 53, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:37:51,225 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:38:09,028 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:39:02,818 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.84 vs. limit=5.0 2023-04-27 23:39:03,601 INFO [train.py:904] (2/8) Epoch 3, batch 6700, loss[loss=0.2618, simple_loss=0.3259, pruned_loss=0.09889, over 16557.00 frames. ], tot_loss[loss=0.2854, simple_loss=0.3515, pruned_loss=0.1096, over 3065942.76 frames. ], batch size: 68, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:39:05,240 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9139, 3.8575, 3.7580, 3.8301, 3.3933, 3.8456, 3.5781, 3.6177], device='cuda:2'), covar=tensor([0.0312, 0.0168, 0.0194, 0.0129, 0.0657, 0.0207, 0.0572, 0.0321], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0128, 0.0167, 0.0140, 0.0197, 0.0160, 0.0126, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:39:19,055 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:57,110 INFO [optim.py:368] (2/8) 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 23:40:21,184 INFO [train.py:904] (2/8) Epoch 3, batch 6750, loss[loss=0.3483, simple_loss=0.3991, pruned_loss=0.1487, over 12070.00 frames. ], tot_loss[loss=0.2849, simple_loss=0.351, pruned_loss=0.1094, over 3063038.67 frames. ], batch size: 246, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:40:51,196 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:41:37,189 INFO [train.py:904] (2/8) Epoch 3, batch 6800, loss[loss=0.2968, simple_loss=0.3532, pruned_loss=0.1202, over 11627.00 frames. ], tot_loss[loss=0.2837, simple_loss=0.3502, pruned_loss=0.1086, over 3063269.80 frames. ], batch size: 247, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:42:03,930 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:31,272 INFO [optim.py:368] (2/8) 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 23:42:51,775 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:55,394 INFO [train.py:904] (2/8) Epoch 3, batch 6850, loss[loss=0.251, simple_loss=0.3563, pruned_loss=0.07286, over 16801.00 frames. ], tot_loss[loss=0.2837, simple_loss=0.3513, pruned_loss=0.108, over 3080466.67 frames. ], batch size: 83, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:43:09,031 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7722, 5.2113, 5.2299, 5.2226, 5.0872, 5.8074, 5.3172, 5.0212], device='cuda:2'), covar=tensor([0.0657, 0.1282, 0.1096, 0.1502, 0.2143, 0.0787, 0.1047, 0.2252], device='cuda:2'), in_proj_covar=tensor([0.0237, 0.0320, 0.0301, 0.0275, 0.0362, 0.0324, 0.0263, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:44:01,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5884, 1.5613, 1.7173, 2.4263, 2.4539, 2.5385, 1.3315, 2.5692], device='cuda:2'), covar=tensor([0.0048, 0.0213, 0.0161, 0.0104, 0.0075, 0.0080, 0.0226, 0.0053], device='cuda:2'), in_proj_covar=tensor([0.0087, 0.0125, 0.0109, 0.0102, 0.0094, 0.0070, 0.0117, 0.0063], device='cuda:2'), out_proj_covar=tensor([1.3221e-04, 1.9713e-04, 1.7679e-04, 1.6356e-04, 1.4621e-04, 1.0700e-04, 1.8139e-04, 9.5264e-05], device='cuda:2') 2023-04-27 23:44:10,166 INFO [train.py:904] (2/8) Epoch 3, batch 6900, loss[loss=0.2923, simple_loss=0.3627, pruned_loss=0.1109, over 16438.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.3527, pruned_loss=0.107, over 3097019.91 frames. ], batch size: 146, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:44:22,773 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:44:47,732 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 23:45:02,576 INFO [optim.py:368] (2/8) 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 23:45:25,673 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9177, 3.7842, 3.7992, 3.8154, 3.3983, 3.8550, 3.6117, 3.6203], device='cuda:2'), covar=tensor([0.0337, 0.0229, 0.0179, 0.0138, 0.0621, 0.0228, 0.0510, 0.0342], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0129, 0.0168, 0.0141, 0.0199, 0.0161, 0.0126, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:45:28,455 INFO [train.py:904] (2/8) Epoch 3, batch 6950, loss[loss=0.3002, simple_loss=0.3649, pruned_loss=0.1178, over 16724.00 frames. ], tot_loss[loss=0.2878, simple_loss=0.356, pruned_loss=0.1098, over 3093357.54 frames. ], batch size: 124, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:45:41,657 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:45:52,721 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5850, 4.9004, 4.6186, 4.5901, 4.2112, 4.2063, 4.4166, 4.9239], device='cuda:2'), covar=tensor([0.0478, 0.0692, 0.0973, 0.0452, 0.0585, 0.0700, 0.0520, 0.0603], device='cuda:2'), in_proj_covar=tensor([0.0274, 0.0382, 0.0339, 0.0242, 0.0243, 0.0242, 0.0303, 0.0268], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:46:43,369 INFO [train.py:904] (2/8) Epoch 3, batch 7000, loss[loss=0.291, simple_loss=0.3659, pruned_loss=0.1081, over 15455.00 frames. ], tot_loss[loss=0.2867, simple_loss=0.3555, pruned_loss=0.109, over 3083846.27 frames. ], batch size: 191, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:46:50,998 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:11,568 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6839, 3.2321, 3.2184, 2.2069, 2.9930, 3.1100, 3.2299, 1.5645], device='cuda:2'), covar=tensor([0.0352, 0.0027, 0.0034, 0.0230, 0.0035, 0.0076, 0.0029, 0.0354], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0051, 0.0058, 0.0111, 0.0053, 0.0063, 0.0057, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-27 23:47:35,868 INFO [optim.py:368] (2/8) 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 23:47:38,994 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 23:48:01,107 INFO [train.py:904] (2/8) Epoch 3, batch 7050, loss[loss=0.2888, simple_loss=0.3558, pruned_loss=0.1109, over 16744.00 frames. ], tot_loss[loss=0.288, simple_loss=0.3566, pruned_loss=0.1097, over 3069589.10 frames. ], batch size: 124, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:48:32,883 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5126, 4.4853, 4.3426, 3.6173, 4.3409, 1.7661, 4.1625, 4.3163], device='cuda:2'), covar=tensor([0.0058, 0.0042, 0.0062, 0.0277, 0.0045, 0.1403, 0.0063, 0.0090], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0061, 0.0095, 0.0109, 0.0070, 0.0121, 0.0082, 0.0093], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:49:19,631 INFO [train.py:904] (2/8) Epoch 3, batch 7100, loss[loss=0.276, simple_loss=0.3496, pruned_loss=0.1012, over 16839.00 frames. ], tot_loss[loss=0.2862, simple_loss=0.3546, pruned_loss=0.1089, over 3071493.58 frames. ], batch size: 102, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:50:12,101 INFO [optim.py:368] (2/8) 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 23:50:35,660 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9870, 2.6838, 2.6248, 1.7469, 2.8152, 2.7927, 2.4442, 2.3224], device='cuda:2'), covar=tensor([0.0675, 0.0139, 0.0142, 0.0838, 0.0089, 0.0089, 0.0317, 0.0342], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0082, 0.0078, 0.0144, 0.0072, 0.0071, 0.0115, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-27 23:50:36,353 INFO [train.py:904] (2/8) Epoch 3, batch 7150, loss[loss=0.3232, simple_loss=0.3601, pruned_loss=0.1431, over 11404.00 frames. ], tot_loss[loss=0.2837, simple_loss=0.3517, pruned_loss=0.1079, over 3080701.33 frames. ], batch size: 247, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:51,150 INFO [train.py:904] (2/8) Epoch 3, batch 7200, loss[loss=0.256, simple_loss=0.3325, pruned_loss=0.08978, over 16767.00 frames. ], tot_loss[loss=0.2789, simple_loss=0.3481, pruned_loss=0.1048, over 3081626.09 frames. ], batch size: 39, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:51,911 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 23:51:55,800 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:52:45,515 INFO [optim.py:368] (2/8) 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 23:53:12,423 INFO [train.py:904] (2/8) Epoch 3, batch 7250, loss[loss=0.2815, simple_loss=0.3421, pruned_loss=0.1105, over 15209.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3453, pruned_loss=0.1035, over 3042418.67 frames. ], batch size: 190, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:53:26,576 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:53:50,403 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 23:54:26,523 INFO [train.py:904] (2/8) Epoch 3, batch 7300, loss[loss=0.3347, simple_loss=0.3778, pruned_loss=0.1457, over 11656.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.345, pruned_loss=0.1034, over 3048375.17 frames. ], batch size: 248, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:54:35,449 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:54:38,056 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:54:39,581 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 23:54:41,041 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6886, 4.8895, 4.6155, 4.6435, 4.2355, 4.1777, 4.4153, 4.9090], device='cuda:2'), covar=tensor([0.0376, 0.0493, 0.0731, 0.0361, 0.0444, 0.0654, 0.0427, 0.0516], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0372, 0.0329, 0.0239, 0.0241, 0.0243, 0.0298, 0.0260], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:55:17,368 INFO [optim.py:368] (2/8) 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 23:55:17,912 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3004, 2.1447, 1.4465, 1.8140, 2.7252, 2.5694, 3.3893, 3.1019], device='cuda:2'), covar=tensor([0.0014, 0.0154, 0.0210, 0.0175, 0.0090, 0.0132, 0.0030, 0.0053], device='cuda:2'), in_proj_covar=tensor([0.0054, 0.0121, 0.0123, 0.0122, 0.0113, 0.0126, 0.0077, 0.0100], device='cuda:2'), out_proj_covar=tensor([7.2703e-05, 1.7120e-04, 1.6858e-04, 1.7279e-04, 1.6383e-04, 1.8024e-04, 1.0931e-04, 1.4513e-04], device='cuda:2') 2023-04-27 23:55:23,148 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2819, 4.2214, 4.1528, 3.4544, 4.1173, 1.6034, 3.8911, 4.0060], device='cuda:2'), covar=tensor([0.0054, 0.0048, 0.0063, 0.0285, 0.0054, 0.1556, 0.0081, 0.0103], device='cuda:2'), in_proj_covar=tensor([0.0072, 0.0061, 0.0092, 0.0107, 0.0068, 0.0118, 0.0080, 0.0091], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-27 23:55:40,985 INFO [train.py:904] (2/8) Epoch 3, batch 7350, loss[loss=0.2565, simple_loss=0.3212, pruned_loss=0.09584, over 17036.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.344, pruned_loss=0.1029, over 3048589.87 frames. ], batch size: 53, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:55:46,794 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:56:44,465 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 23:56:59,424 INFO [train.py:904] (2/8) Epoch 3, batch 7400, loss[loss=0.2504, simple_loss=0.3357, pruned_loss=0.0826, over 16403.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3451, pruned_loss=0.1034, over 3065721.58 frames. ], batch size: 75, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:57:20,327 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:57:55,128 INFO [optim.py:368] (2/8) 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 23:58:18,299 INFO [train.py:904] (2/8) Epoch 3, batch 7450, loss[loss=0.2688, simple_loss=0.3476, pruned_loss=0.09498, over 16330.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3477, pruned_loss=0.1057, over 3046063.74 frames. ], batch size: 146, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:58:54,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9889, 1.5635, 1.3419, 1.3357, 1.7506, 1.6076, 1.7235, 1.8353], device='cuda:2'), covar=tensor([0.0023, 0.0111, 0.0157, 0.0143, 0.0076, 0.0118, 0.0058, 0.0076], device='cuda:2'), in_proj_covar=tensor([0.0053, 0.0120, 0.0124, 0.0121, 0.0114, 0.0123, 0.0078, 0.0099], device='cuda:2'), out_proj_covar=tensor([7.1515e-05, 1.6906e-04, 1.7058e-04, 1.6955e-04, 1.6432e-04, 1.7576e-04, 1.1002e-04, 1.4347e-04], device='cuda:2') 2023-04-27 23:58:58,286 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:59:39,385 INFO [train.py:904] (2/8) Epoch 3, batch 7500, loss[loss=0.2924, simple_loss=0.3658, pruned_loss=0.1095, over 16915.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3474, pruned_loss=0.1045, over 3072464.19 frames. ], batch size: 116, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:59:44,225 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:33,135 INFO [optim.py:368] (2/8) 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 00:00:35,123 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-28 00:00:44,673 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4682, 2.5736, 2.1582, 3.9336, 1.8861, 3.5216, 2.2800, 2.3059], device='cuda:2'), covar=tensor([0.0390, 0.0874, 0.0616, 0.0225, 0.1980, 0.0307, 0.1051, 0.1500], device='cuda:2'), in_proj_covar=tensor([0.0270, 0.0238, 0.0202, 0.0269, 0.0318, 0.0213, 0.0231, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:00:55,938 INFO [train.py:904] (2/8) Epoch 3, batch 7550, loss[loss=0.3237, simple_loss=0.3653, pruned_loss=0.141, over 11327.00 frames. ], tot_loss[loss=0.2773, simple_loss=0.3462, pruned_loss=0.1041, over 3073737.38 frames. ], batch size: 246, lr: 1.96e-02, grad_scale: 4.0 2023-04-28 00:00:58,759 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:15,862 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1612, 3.8423, 3.9376, 2.8920, 3.4926, 3.7367, 3.9245, 1.8649], device='cuda:2'), covar=tensor([0.0310, 0.0020, 0.0026, 0.0180, 0.0037, 0.0053, 0.0024, 0.0312], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0052, 0.0057, 0.0109, 0.0053, 0.0062, 0.0057, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 00:01:59,717 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 00:02:13,362 INFO [train.py:904] (2/8) Epoch 3, batch 7600, loss[loss=0.2767, simple_loss=0.3554, pruned_loss=0.09901, over 16847.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3451, pruned_loss=0.1034, over 3092504.67 frames. ], batch size: 102, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:02:55,530 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6988, 2.6575, 2.6745, 1.9647, 2.3171, 2.4134, 2.3865, 1.6991], device='cuda:2'), covar=tensor([0.0276, 0.0034, 0.0047, 0.0194, 0.0048, 0.0062, 0.0034, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0051, 0.0057, 0.0109, 0.0053, 0.0062, 0.0057, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 00:03:08,896 INFO [optim.py:368] (2/8) 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 00:03:31,739 INFO [train.py:904] (2/8) Epoch 3, batch 7650, loss[loss=0.3462, simple_loss=0.3776, pruned_loss=0.1574, over 11167.00 frames. ], tot_loss[loss=0.2783, simple_loss=0.3468, pruned_loss=0.1049, over 3082879.43 frames. ], batch size: 247, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:04:52,685 INFO [train.py:904] (2/8) Epoch 3, batch 7700, loss[loss=0.2573, simple_loss=0.336, pruned_loss=0.08928, over 16776.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.349, pruned_loss=0.1074, over 3069562.23 frames. ], batch size: 83, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:05:05,429 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 00:05:28,925 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6737, 3.5855, 3.6835, 3.9224, 3.9063, 3.5022, 3.9219, 3.9970], device='cuda:2'), covar=tensor([0.0640, 0.0612, 0.1018, 0.0410, 0.0462, 0.1157, 0.0476, 0.0340], device='cuda:2'), in_proj_covar=tensor([0.0283, 0.0357, 0.0455, 0.0349, 0.0266, 0.0251, 0.0285, 0.0288], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:05:46,976 INFO [optim.py:368] (2/8) 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 00:06:11,242 INFO [train.py:904] (2/8) Epoch 3, batch 7750, loss[loss=0.2853, simple_loss=0.3518, pruned_loss=0.1094, over 16370.00 frames. ], tot_loss[loss=0.2822, simple_loss=0.3494, pruned_loss=0.1075, over 3067509.87 frames. ], batch size: 146, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:06:40,888 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:07:03,202 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 00:07:22,020 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8301, 3.7627, 1.4346, 3.8629, 2.4624, 3.8729, 1.8787, 2.7255], device='cuda:2'), covar=tensor([0.0058, 0.0204, 0.1795, 0.0043, 0.0758, 0.0333, 0.1376, 0.0601], device='cuda:2'), in_proj_covar=tensor([0.0089, 0.0135, 0.0175, 0.0082, 0.0161, 0.0163, 0.0180, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 00:07:28,342 INFO [train.py:904] (2/8) Epoch 3, batch 7800, loss[loss=0.3721, simple_loss=0.396, pruned_loss=0.1741, over 11479.00 frames. ], tot_loss[loss=0.2836, simple_loss=0.3506, pruned_loss=0.1083, over 3078829.77 frames. ], batch size: 248, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:08:22,843 INFO [optim.py:368] (2/8) 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 00:08:45,060 INFO [train.py:904] (2/8) Epoch 3, batch 7850, loss[loss=0.2626, simple_loss=0.3414, pruned_loss=0.09188, over 16651.00 frames. ], tot_loss[loss=0.282, simple_loss=0.3506, pruned_loss=0.1067, over 3093774.19 frames. ], batch size: 57, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:00,834 INFO [train.py:904] (2/8) Epoch 3, batch 7900, loss[loss=0.2746, simple_loss=0.3503, pruned_loss=0.09948, over 16720.00 frames. ], tot_loss[loss=0.2819, simple_loss=0.35, pruned_loss=0.1069, over 3078065.80 frames. ], batch size: 124, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:45,177 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:55,796 INFO [optim.py:368] (2/8) 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 00:11:18,498 INFO [train.py:904] (2/8) Epoch 3, batch 7950, loss[loss=0.2976, simple_loss=0.3672, pruned_loss=0.114, over 15296.00 frames. ], tot_loss[loss=0.2822, simple_loss=0.3501, pruned_loss=0.1072, over 3080753.18 frames. ], batch size: 190, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:11:26,011 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:12:13,416 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:18,118 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:20,127 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 00:12:33,513 INFO [train.py:904] (2/8) Epoch 3, batch 8000, loss[loss=0.327, simple_loss=0.3677, pruned_loss=0.1431, over 11576.00 frames. ], tot_loss[loss=0.2833, simple_loss=0.3505, pruned_loss=0.108, over 3068678.28 frames. ], batch size: 247, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:12:56,324 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 00:13:26,665 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 00:13:27,063 INFO [optim.py:368] (2/8) 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 00:13:46,423 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:49,305 INFO [train.py:904] (2/8) Epoch 3, batch 8050, loss[loss=0.2816, simple_loss=0.3548, pruned_loss=0.1042, over 15402.00 frames. ], tot_loss[loss=0.2812, simple_loss=0.3496, pruned_loss=0.1065, over 3102707.16 frames. ], batch size: 191, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:14:18,760 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:14:41,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5745, 3.4897, 4.0039, 3.9692, 3.9502, 3.5962, 3.6726, 3.6802], device='cuda:2'), covar=tensor([0.0245, 0.0313, 0.0255, 0.0332, 0.0369, 0.0283, 0.0646, 0.0307], device='cuda:2'), in_proj_covar=tensor([0.0189, 0.0179, 0.0192, 0.0194, 0.0232, 0.0200, 0.0295, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 00:15:05,403 INFO [train.py:904] (2/8) Epoch 3, batch 8100, loss[loss=0.2956, simple_loss=0.3618, pruned_loss=0.1147, over 16328.00 frames. ], tot_loss[loss=0.2802, simple_loss=0.349, pruned_loss=0.1057, over 3094032.26 frames. ], batch size: 165, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:15:30,570 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:15:51,125 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-28 00:15:57,041 INFO [optim.py:368] (2/8) 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 00:16:20,000 INFO [train.py:904] (2/8) Epoch 3, batch 8150, loss[loss=0.3206, simple_loss=0.3672, pruned_loss=0.137, over 11852.00 frames. ], tot_loss[loss=0.2776, simple_loss=0.3461, pruned_loss=0.1046, over 3087617.54 frames. ], batch size: 247, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:16:40,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0066, 3.7567, 3.9800, 4.2285, 4.2684, 3.7604, 4.2533, 4.2566], device='cuda:2'), covar=tensor([0.0608, 0.0621, 0.0948, 0.0363, 0.0358, 0.0791, 0.0393, 0.0349], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0370, 0.0466, 0.0361, 0.0271, 0.0260, 0.0293, 0.0300], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:17:15,168 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:16,733 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-28 00:17:20,588 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0114, 2.6604, 2.6313, 1.6944, 2.8036, 2.7680, 2.4314, 2.3214], device='cuda:2'), covar=tensor([0.0683, 0.0124, 0.0185, 0.1011, 0.0109, 0.0120, 0.0324, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0082, 0.0080, 0.0144, 0.0072, 0.0074, 0.0113, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 00:17:23,243 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:36,817 INFO [train.py:904] (2/8) Epoch 3, batch 8200, loss[loss=0.3086, simple_loss=0.3549, pruned_loss=0.1312, over 11436.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3428, pruned_loss=0.1032, over 3083460.92 frames. ], batch size: 248, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:18:16,354 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5457, 3.3784, 2.9830, 1.8149, 2.5981, 2.0936, 2.9092, 3.4105], device='cuda:2'), covar=tensor([0.0389, 0.0618, 0.0484, 0.1559, 0.0788, 0.0933, 0.0770, 0.0504], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0119, 0.0157, 0.0145, 0.0139, 0.0130, 0.0146, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 00:18:31,947 INFO [optim.py:368] (2/8) 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 00:18:50,557 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:55,849 INFO [train.py:904] (2/8) Epoch 3, batch 8250, loss[loss=0.2387, simple_loss=0.3189, pruned_loss=0.07924, over 16795.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3423, pruned_loss=0.1014, over 3062125.07 frames. ], batch size: 102, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:19:00,253 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:19:51,813 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:16,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6521, 4.7310, 4.7359, 4.7013, 4.6459, 5.1852, 4.9789, 4.5721], device='cuda:2'), covar=tensor([0.0647, 0.1056, 0.0955, 0.1257, 0.1848, 0.0710, 0.0659, 0.1382], device='cuda:2'), in_proj_covar=tensor([0.0226, 0.0315, 0.0295, 0.0273, 0.0351, 0.0323, 0.0259, 0.0366], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 00:20:16,982 INFO [train.py:904] (2/8) Epoch 3, batch 8300, loss[loss=0.2177, simple_loss=0.3124, pruned_loss=0.06154, over 16731.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3374, pruned_loss=0.09644, over 3048069.19 frames. ], batch size: 124, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:20:33,879 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:21:14,648 INFO [optim.py:368] (2/8) 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 00:21:27,048 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:29,066 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4295, 4.3806, 4.4863, 4.5516, 4.5309, 5.0038, 4.7044, 4.3749], device='cuda:2'), covar=tensor([0.0794, 0.1259, 0.1089, 0.1223, 0.1872, 0.0782, 0.0823, 0.1876], device='cuda:2'), in_proj_covar=tensor([0.0222, 0.0311, 0.0286, 0.0269, 0.0344, 0.0318, 0.0253, 0.0360], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:21:39,704 INFO [train.py:904] (2/8) Epoch 3, batch 8350, loss[loss=0.2647, simple_loss=0.3453, pruned_loss=0.09203, over 15372.00 frames. ], tot_loss[loss=0.26, simple_loss=0.3348, pruned_loss=0.09255, over 3051233.23 frames. ], batch size: 190, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:00,530 INFO [train.py:904] (2/8) Epoch 3, batch 8400, loss[loss=0.2264, simple_loss=0.3001, pruned_loss=0.07633, over 11965.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3298, pruned_loss=0.08884, over 3033930.88 frames. ], batch size: 246, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:44,826 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-28 00:23:58,229 INFO [optim.py:368] (2/8) 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 00:24:11,889 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7009, 1.8940, 1.6774, 1.6324, 2.3814, 2.1807, 2.7830, 2.6824], device='cuda:2'), covar=tensor([0.0022, 0.0158, 0.0188, 0.0208, 0.0085, 0.0131, 0.0038, 0.0062], device='cuda:2'), in_proj_covar=tensor([0.0055, 0.0118, 0.0122, 0.0123, 0.0111, 0.0119, 0.0075, 0.0097], device='cuda:2'), out_proj_covar=tensor([7.2852e-05, 1.6492e-04, 1.6547e-04, 1.7206e-04, 1.5822e-04, 1.6727e-04, 1.0502e-04, 1.3765e-04], device='cuda:2') 2023-04-28 00:24:20,120 INFO [train.py:904] (2/8) Epoch 3, batch 8450, loss[loss=0.2287, simple_loss=0.3131, pruned_loss=0.07216, over 16741.00 frames. ], tot_loss[loss=0.2497, simple_loss=0.3273, pruned_loss=0.08607, over 3055034.53 frames. ], batch size: 83, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:24:57,850 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4192, 3.7697, 3.7933, 2.7809, 3.6303, 3.7331, 3.5906, 1.9790], device='cuda:2'), covar=tensor([0.0285, 0.0016, 0.0020, 0.0188, 0.0023, 0.0032, 0.0023, 0.0310], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0053, 0.0056, 0.0109, 0.0051, 0.0062, 0.0057, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 00:25:42,011 INFO [train.py:904] (2/8) Epoch 3, batch 8500, loss[loss=0.2269, simple_loss=0.3057, pruned_loss=0.074, over 16858.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3223, pruned_loss=0.0827, over 3061127.53 frames. ], batch size: 116, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:26:40,691 INFO [optim.py:368] (2/8) 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 00:26:49,999 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:00,694 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:05,698 INFO [train.py:904] (2/8) Epoch 3, batch 8550, loss[loss=0.2211, simple_loss=0.2915, pruned_loss=0.07539, over 12050.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3195, pruned_loss=0.08128, over 3045237.47 frames. ], batch size: 246, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:27:06,974 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 00:27:25,975 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:12,803 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:18,152 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9863, 1.6229, 1.4422, 1.3491, 1.8521, 1.6291, 1.9300, 1.9218], device='cuda:2'), covar=tensor([0.0023, 0.0135, 0.0166, 0.0155, 0.0083, 0.0128, 0.0050, 0.0078], device='cuda:2'), in_proj_covar=tensor([0.0054, 0.0116, 0.0118, 0.0119, 0.0109, 0.0117, 0.0073, 0.0093], device='cuda:2'), out_proj_covar=tensor([7.0779e-05, 1.6191e-04, 1.5935e-04, 1.6495e-04, 1.5415e-04, 1.6465e-04, 1.0123e-04, 1.3104e-04], device='cuda:2') 2023-04-28 00:28:46,470 INFO [train.py:904] (2/8) Epoch 3, batch 8600, loss[loss=0.2395, simple_loss=0.3325, pruned_loss=0.07327, over 16867.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3208, pruned_loss=0.0807, over 3052760.16 frames. ], batch size: 96, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:29:07,424 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:29:30,798 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0112, 3.7615, 4.0925, 4.2466, 4.3221, 3.8813, 4.3907, 4.3011], device='cuda:2'), covar=tensor([0.0751, 0.0718, 0.1023, 0.0493, 0.0463, 0.0649, 0.0349, 0.0437], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0345, 0.0438, 0.0338, 0.0257, 0.0244, 0.0281, 0.0281], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:29:30,848 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:47,749 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 00:29:50,787 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:57,553 INFO [optim.py:368] (2/8) 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 00:30:11,136 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:30:25,441 INFO [train.py:904] (2/8) Epoch 3, batch 8650, loss[loss=0.2168, simple_loss=0.3086, pruned_loss=0.06253, over 16656.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3181, pruned_loss=0.07853, over 3055864.19 frames. ], batch size: 134, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:30:36,709 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8771, 4.1321, 3.8847, 3.9903, 3.5885, 3.6881, 3.8725, 4.0743], device='cuda:2'), covar=tensor([0.0602, 0.0696, 0.0806, 0.0370, 0.0570, 0.1066, 0.0512, 0.0768], device='cuda:2'), in_proj_covar=tensor([0.0260, 0.0367, 0.0312, 0.0238, 0.0234, 0.0241, 0.0288, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:30:45,513 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:31:06,316 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6675, 2.9802, 2.7732, 4.3277, 3.9819, 4.1766, 1.4601, 3.2614], device='cuda:2'), covar=tensor([0.1480, 0.0471, 0.0910, 0.0061, 0.0163, 0.0228, 0.1365, 0.0526], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0134, 0.0162, 0.0071, 0.0134, 0.0143, 0.0152, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-04-28 00:31:52,871 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:12,024 INFO [train.py:904] (2/8) Epoch 3, batch 8700, loss[loss=0.2338, simple_loss=0.3003, pruned_loss=0.08364, over 12660.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3131, pruned_loss=0.07589, over 3041344.27 frames. ], batch size: 250, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:32:15,232 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:33:18,499 INFO [optim.py:368] (2/8) 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 00:33:46,755 INFO [train.py:904] (2/8) Epoch 3, batch 8750, loss[loss=0.2294, simple_loss=0.3233, pruned_loss=0.06778, over 16758.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3121, pruned_loss=0.07419, over 3064277.97 frames. ], batch size: 83, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:33:51,108 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-04-28 00:34:18,785 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:35:29,212 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8508, 3.8904, 4.3980, 4.3387, 4.3547, 3.8963, 4.0887, 3.9055], device='cuda:2'), covar=tensor([0.0223, 0.0245, 0.0275, 0.0363, 0.0354, 0.0272, 0.0641, 0.0338], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0164, 0.0178, 0.0182, 0.0209, 0.0186, 0.0272, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-28 00:35:38,457 INFO [train.py:904] (2/8) Epoch 3, batch 8800, loss[loss=0.2405, simple_loss=0.3191, pruned_loss=0.08096, over 12861.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3101, pruned_loss=0.07246, over 3082512.65 frames. ], batch size: 247, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:35:52,497 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3424, 3.8645, 3.8520, 1.4460, 4.0657, 4.0301, 3.2092, 2.8524], device='cuda:2'), covar=tensor([0.0797, 0.0127, 0.0163, 0.1489, 0.0049, 0.0041, 0.0255, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0081, 0.0077, 0.0144, 0.0069, 0.0069, 0.0109, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 00:36:52,045 INFO [optim.py:368] (2/8) 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 00:37:03,485 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:07,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0313, 1.5039, 1.3220, 1.2133, 1.8145, 1.4245, 1.8621, 1.8620], device='cuda:2'), covar=tensor([0.0021, 0.0142, 0.0169, 0.0205, 0.0083, 0.0169, 0.0053, 0.0069], device='cuda:2'), in_proj_covar=tensor([0.0052, 0.0122, 0.0121, 0.0123, 0.0111, 0.0121, 0.0075, 0.0097], device='cuda:2'), out_proj_covar=tensor([6.7726e-05, 1.7033e-04, 1.6410e-04, 1.6945e-04, 1.5656e-04, 1.6986e-04, 1.0284e-04, 1.3647e-04], device='cuda:2') 2023-04-28 00:37:16,943 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:21,268 INFO [train.py:904] (2/8) Epoch 3, batch 8850, loss[loss=0.2269, simple_loss=0.3233, pruned_loss=0.06526, over 16102.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3122, pruned_loss=0.0716, over 3064990.08 frames. ], batch size: 165, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:38:45,270 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:38:57,572 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 3, batch 8900, loss[loss=0.1971, simple_loss=0.2916, pruned_loss=0.05128, over 16679.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3119, pruned_loss=0.07035, over 3072871.49 frames. ], batch size: 89, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:39:17,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1487, 3.0228, 2.6522, 1.8538, 2.4947, 2.1554, 2.7350, 3.0677], device='cuda:2'), covar=tensor([0.0228, 0.0398, 0.0477, 0.1376, 0.0658, 0.0753, 0.0615, 0.0373], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0108, 0.0150, 0.0143, 0.0133, 0.0129, 0.0139, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 00:39:40,209 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:35,250 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:36,245 INFO [optim.py:368] (2/8) 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 00:40:49,551 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-28 00:41:11,852 INFO [train.py:904] (2/8) Epoch 3, batch 8950, loss[loss=0.1985, simple_loss=0.2893, pruned_loss=0.05389, over 16211.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3123, pruned_loss=0.07149, over 3083355.98 frames. ], batch size: 165, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:42:53,668 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:43:00,670 INFO [train.py:904] (2/8) Epoch 3, batch 9000, loss[loss=0.2244, simple_loss=0.2979, pruned_loss=0.07541, over 12045.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.308, pruned_loss=0.06926, over 3084311.04 frames. ], batch size: 248, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:43:00,671 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 00:43:11,832 INFO [train.py:938] (2/8) Epoch 3, validation: loss=0.1886, simple_loss=0.2904, pruned_loss=0.04341, over 944034.00 frames. 2023-04-28 00:43:11,833 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 00:44:26,933 INFO [optim.py:368] (2/8) 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 00:44:57,071 INFO [train.py:904] (2/8) Epoch 3, batch 9050, loss[loss=0.2249, simple_loss=0.3076, pruned_loss=0.0711, over 16367.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3096, pruned_loss=0.07048, over 3081173.45 frames. ], batch size: 146, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:45:12,170 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:46:10,104 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-28 00:46:35,348 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1921, 3.2163, 1.3406, 3.2512, 2.0916, 3.1448, 1.5441, 2.4533], device='cuda:2'), covar=tensor([0.0099, 0.0216, 0.1809, 0.0053, 0.0812, 0.0395, 0.1691, 0.0633], device='cuda:2'), in_proj_covar=tensor([0.0088, 0.0127, 0.0173, 0.0076, 0.0154, 0.0154, 0.0180, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 00:46:41,627 INFO [train.py:904] (2/8) Epoch 3, batch 9100, loss[loss=0.2392, simple_loss=0.3231, pruned_loss=0.07765, over 15180.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3097, pruned_loss=0.07135, over 3073428.56 frames. ], batch size: 190, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:47:11,942 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5119, 2.3319, 2.0997, 3.9333, 1.6315, 3.5404, 2.1107, 2.1085], device='cuda:2'), covar=tensor([0.0359, 0.1010, 0.0611, 0.0206, 0.2180, 0.0295, 0.1135, 0.1631], device='cuda:2'), in_proj_covar=tensor([0.0263, 0.0247, 0.0208, 0.0265, 0.0317, 0.0213, 0.0233, 0.0293], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:48:03,077 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0652, 3.9990, 4.5729, 4.5639, 4.5044, 4.1517, 4.2213, 4.0310], device='cuda:2'), covar=tensor([0.0203, 0.0324, 0.0227, 0.0258, 0.0309, 0.0221, 0.0514, 0.0310], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0172, 0.0178, 0.0183, 0.0214, 0.0190, 0.0277, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 00:48:08,508 INFO [optim.py:368] (2/8) 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 00:48:30,251 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-28 00:48:40,643 INFO [train.py:904] (2/8) Epoch 3, batch 9150, loss[loss=0.2166, simple_loss=0.3056, pruned_loss=0.06382, over 16797.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3102, pruned_loss=0.0709, over 3082024.00 frames. ], batch size: 124, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:50:24,877 INFO [train.py:904] (2/8) Epoch 3, batch 9200, loss[loss=0.2185, simple_loss=0.3091, pruned_loss=0.06397, over 16219.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3056, pruned_loss=0.0693, over 3091944.11 frames. ], batch size: 165, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:50:26,358 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 00:50:53,686 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8263, 2.9615, 2.2082, 3.8145, 3.6131, 3.6993, 1.5343, 2.9480], device='cuda:2'), covar=tensor([0.1388, 0.0458, 0.1232, 0.0094, 0.0195, 0.0342, 0.1450, 0.0638], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0133, 0.0160, 0.0069, 0.0130, 0.0142, 0.0155, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-04-28 00:50:54,896 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:32,421 INFO [optim.py:368] (2/8) 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 00:52:01,381 INFO [train.py:904] (2/8) Epoch 3, batch 9250, loss[loss=0.2179, simple_loss=0.3077, pruned_loss=0.06409, over 16292.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3054, pruned_loss=0.06984, over 3065738.98 frames. ], batch size: 146, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:52:02,461 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:26,186 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-28 00:52:29,199 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:31,860 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:50,314 INFO [train.py:904] (2/8) Epoch 3, batch 9300, loss[loss=0.2005, simple_loss=0.2866, pruned_loss=0.05713, over 16612.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3029, pruned_loss=0.06838, over 3062484.45 frames. ], batch size: 134, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:54:16,006 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:55:11,308 INFO [optim.py:368] (2/8) 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 00:55:35,075 INFO [train.py:904] (2/8) Epoch 3, batch 9350, loss[loss=0.2376, simple_loss=0.3142, pruned_loss=0.08048, over 16950.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.303, pruned_loss=0.06818, over 3083126.57 frames. ], batch size: 109, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:55:51,606 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:56:00,146 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:56:04,062 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3119, 3.2127, 3.2710, 3.4792, 3.4754, 3.1995, 3.5192, 3.5380], device='cuda:2'), covar=tensor([0.0516, 0.0605, 0.1167, 0.0516, 0.0476, 0.1267, 0.0494, 0.0354], device='cuda:2'), in_proj_covar=tensor([0.0283, 0.0347, 0.0440, 0.0342, 0.0257, 0.0241, 0.0273, 0.0281], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:56:21,699 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4888, 3.5585, 2.9658, 2.1874, 2.5502, 2.1093, 3.7268, 3.9638], device='cuda:2'), covar=tensor([0.2153, 0.0633, 0.1137, 0.1296, 0.1628, 0.1360, 0.0353, 0.0408], device='cuda:2'), in_proj_covar=tensor([0.0263, 0.0240, 0.0253, 0.0213, 0.0231, 0.0193, 0.0216, 0.0204], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:56:40,382 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9079, 3.7432, 3.8529, 3.8724, 3.9414, 4.2226, 4.0344, 3.8148], device='cuda:2'), covar=tensor([0.1119, 0.1488, 0.1039, 0.1525, 0.1893, 0.0930, 0.0786, 0.1602], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0298, 0.0284, 0.0263, 0.0340, 0.0316, 0.0246, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:57:16,961 INFO [train.py:904] (2/8) Epoch 3, batch 9400, loss[loss=0.1868, simple_loss=0.2672, pruned_loss=0.05324, over 12439.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3024, pruned_loss=0.06795, over 3061604.28 frames. ], batch size: 248, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:57:27,902 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:28,470 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-28 00:58:01,501 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:58:08,371 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0910, 4.0945, 3.8674, 3.9451, 3.5497, 4.0495, 3.7580, 3.7179], device='cuda:2'), covar=tensor([0.0310, 0.0160, 0.0193, 0.0137, 0.0610, 0.0168, 0.0438, 0.0305], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0122, 0.0165, 0.0135, 0.0188, 0.0151, 0.0117, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 00:58:32,711 INFO [optim.py:368] (2/8) 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 00:58:58,179 INFO [train.py:904] (2/8) Epoch 3, batch 9450, loss[loss=0.2401, simple_loss=0.3105, pruned_loss=0.08488, over 12666.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3053, pruned_loss=0.06901, over 3064580.78 frames. ], batch size: 250, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:58:59,046 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4949, 3.7607, 1.6696, 3.7753, 2.5193, 3.6753, 1.9931, 2.7788], device='cuda:2'), covar=tensor([0.0062, 0.0166, 0.1495, 0.0030, 0.0629, 0.0345, 0.1196, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0091, 0.0126, 0.0167, 0.0075, 0.0151, 0.0150, 0.0174, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 01:00:26,873 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-28 01:00:37,296 INFO [train.py:904] (2/8) Epoch 3, batch 9500, loss[loss=0.2196, simple_loss=0.3055, pruned_loss=0.06681, over 15333.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3039, pruned_loss=0.06808, over 3076367.71 frames. ], batch size: 191, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:41,457 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:01:03,906 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4354, 3.4785, 2.8525, 2.2290, 2.5308, 2.0682, 3.5887, 3.8942], device='cuda:2'), covar=tensor([0.2219, 0.0721, 0.1228, 0.1333, 0.1780, 0.1547, 0.0439, 0.0342], device='cuda:2'), in_proj_covar=tensor([0.0256, 0.0237, 0.0250, 0.0211, 0.0226, 0.0191, 0.0210, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:01:50,890 INFO [optim.py:368] (2/8) 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 01:02:19,709 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5836, 4.3925, 4.3294, 4.3798, 3.9575, 4.4432, 4.3962, 4.0727], device='cuda:2'), covar=tensor([0.0271, 0.0239, 0.0167, 0.0114, 0.0609, 0.0169, 0.0198, 0.0312], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0124, 0.0164, 0.0133, 0.0189, 0.0151, 0.0116, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:02:22,330 INFO [train.py:904] (2/8) Epoch 3, batch 9550, loss[loss=0.2422, simple_loss=0.3261, pruned_loss=0.07912, over 16356.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3042, pruned_loss=0.06829, over 3092551.37 frames. ], batch size: 146, lr: 1.89e-02, grad_scale: 4.0 2023-04-28 01:02:49,277 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:41,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6246, 4.9343, 4.8662, 5.0286, 4.9573, 5.4735, 5.1478, 4.8803], device='cuda:2'), covar=tensor([0.0643, 0.1242, 0.0989, 0.1367, 0.1871, 0.0730, 0.0882, 0.1803], device='cuda:2'), in_proj_covar=tensor([0.0212, 0.0298, 0.0286, 0.0262, 0.0340, 0.0310, 0.0241, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:03:48,966 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:51,890 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8414, 4.1738, 4.0015, 4.0404, 3.6298, 3.7502, 3.8743, 4.1364], device='cuda:2'), covar=tensor([0.0543, 0.0612, 0.0636, 0.0359, 0.0475, 0.0945, 0.0486, 0.0661], device='cuda:2'), in_proj_covar=tensor([0.0260, 0.0362, 0.0308, 0.0237, 0.0235, 0.0243, 0.0290, 0.0252], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:04:04,157 INFO [train.py:904] (2/8) Epoch 3, batch 9600, loss[loss=0.2088, simple_loss=0.2866, pruned_loss=0.06549, over 17161.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.306, pruned_loss=0.06969, over 3079876.86 frames. ], batch size: 49, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:04:16,860 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:32,030 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 01:05:18,577 INFO [optim.py:368] (2/8) 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 01:05:23,704 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:51,206 INFO [train.py:904] (2/8) Epoch 3, batch 9650, loss[loss=0.2124, simple_loss=0.2926, pruned_loss=0.06615, over 12353.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3082, pruned_loss=0.07006, over 3067261.02 frames. ], batch size: 248, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:05:55,225 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8325, 5.1546, 4.8758, 4.9774, 4.5859, 4.4370, 4.6653, 5.2490], device='cuda:2'), covar=tensor([0.0542, 0.0741, 0.0904, 0.0355, 0.0501, 0.0670, 0.0488, 0.0561], device='cuda:2'), in_proj_covar=tensor([0.0256, 0.0360, 0.0303, 0.0232, 0.0235, 0.0241, 0.0287, 0.0251], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:07:03,793 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2033, 1.5207, 2.1466, 2.8903, 2.8971, 3.1620, 1.7256, 2.9492], device='cuda:2'), covar=tensor([0.0035, 0.0230, 0.0133, 0.0090, 0.0056, 0.0046, 0.0192, 0.0056], device='cuda:2'), in_proj_covar=tensor([0.0092, 0.0127, 0.0111, 0.0104, 0.0099, 0.0073, 0.0117, 0.0064], device='cuda:2'), out_proj_covar=tensor([1.3660e-04, 1.9357e-04, 1.7404e-04, 1.6059e-04, 1.5034e-04, 1.0707e-04, 1.7642e-04, 9.5554e-05], device='cuda:2') 2023-04-28 01:07:41,652 INFO [train.py:904] (2/8) Epoch 3, batch 9700, loss[loss=0.2185, simple_loss=0.3009, pruned_loss=0.0681, over 16788.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3065, pruned_loss=0.06946, over 3063360.39 frames. ], batch size: 124, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:08:16,000 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:08:24,498 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9818, 2.7968, 2.6762, 1.6184, 2.8371, 2.8443, 2.5741, 2.3927], device='cuda:2'), covar=tensor([0.0736, 0.0124, 0.0211, 0.1191, 0.0108, 0.0103, 0.0300, 0.0406], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0079, 0.0076, 0.0146, 0.0071, 0.0069, 0.0108, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 01:08:59,978 INFO [optim.py:368] (2/8) 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 01:09:00,770 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2034, 4.0471, 3.5537, 2.0117, 2.9670, 2.4284, 3.4573, 3.8476], device='cuda:2'), covar=tensor([0.0212, 0.0386, 0.0495, 0.1430, 0.0705, 0.0891, 0.0686, 0.0582], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0108, 0.0151, 0.0143, 0.0135, 0.0130, 0.0139, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 01:09:24,292 INFO [train.py:904] (2/8) Epoch 3, batch 9750, loss[loss=0.2069, simple_loss=0.2971, pruned_loss=0.05833, over 16210.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3056, pruned_loss=0.06948, over 3063797.51 frames. ], batch size: 166, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:10:32,392 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 01:11:02,931 INFO [train.py:904] (2/8) Epoch 3, batch 9800, loss[loss=0.194, simple_loss=0.2986, pruned_loss=0.04475, over 16754.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3063, pruned_loss=0.06802, over 3089130.79 frames. ], batch size: 83, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:11:48,330 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:12:14,773 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 3, batch 9850, loss[loss=0.2301, simple_loss=0.3195, pruned_loss=0.07034, over 16242.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3082, pruned_loss=0.06796, over 3102322.53 frames. ], batch size: 165, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:13:02,425 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:13:46,575 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 01:14:06,760 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:39,080 INFO [train.py:904] (2/8) Epoch 3, batch 9900, loss[loss=0.2096, simple_loss=0.3093, pruned_loss=0.05499, over 16685.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3088, pruned_loss=0.0682, over 3092837.11 frames. ], batch size: 76, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:14:54,258 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:59,658 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 01:16:35,560 INFO [train.py:904] (2/8) Epoch 3, batch 9950, loss[loss=0.208, simple_loss=0.3034, pruned_loss=0.05627, over 16922.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3095, pruned_loss=0.06812, over 3084822.66 frames. ], batch size: 102, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:16:46,522 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:17:23,102 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:18:37,069 INFO [train.py:904] (2/8) Epoch 3, batch 10000, loss[loss=0.2091, simple_loss=0.2906, pruned_loss=0.06378, over 12736.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3074, pruned_loss=0.06729, over 3095800.05 frames. ], batch size: 250, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:19:11,875 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:19:55,680 INFO [optim.py:368] (2/8) 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 01:20:19,869 INFO [train.py:904] (2/8) Epoch 3, batch 10050, loss[loss=0.2364, simple_loss=0.3334, pruned_loss=0.06969, over 16896.00 frames. ], tot_loss[loss=0.219, simple_loss=0.306, pruned_loss=0.066, over 3102693.89 frames. ], batch size: 96, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:20:38,383 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:20:54,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3714, 4.5030, 4.2476, 4.2993, 3.7302, 4.3814, 4.2903, 4.0814], device='cuda:2'), covar=tensor([0.0427, 0.0200, 0.0190, 0.0139, 0.0759, 0.0222, 0.0222, 0.0325], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0124, 0.0164, 0.0135, 0.0190, 0.0152, 0.0116, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:21:54,399 INFO [train.py:904] (2/8) Epoch 3, batch 10100, loss[loss=0.2208, simple_loss=0.291, pruned_loss=0.0753, over 12539.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3063, pruned_loss=0.0664, over 3109635.81 frames. ], batch size: 249, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:22:36,856 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:23:00,425 INFO [optim.py:368] (2/8) 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 01:23:38,627 INFO [train.py:904] (2/8) Epoch 4, batch 0, loss[loss=0.2995, simple_loss=0.3599, pruned_loss=0.1196, over 17095.00 frames. ], tot_loss[loss=0.2995, simple_loss=0.3599, pruned_loss=0.1196, over 17095.00 frames. ], batch size: 53, lr: 1.75e-02, grad_scale: 8.0 2023-04-28 01:23:38,627 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 01:23:46,510 INFO [train.py:938] (2/8) Epoch 4, validation: loss=0.188, simple_loss=0.2904, pruned_loss=0.04284, over 944034.00 frames. 2023-04-28 01:23:46,511 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 01:23:56,887 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0504, 5.0702, 4.9139, 5.0028, 4.2383, 5.0898, 5.0324, 4.5973], device='cuda:2'), covar=tensor([0.0485, 0.0279, 0.0221, 0.0145, 0.0818, 0.0206, 0.0188, 0.0349], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0125, 0.0163, 0.0134, 0.0187, 0.0153, 0.0115, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:23:57,937 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:14,003 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0103, 4.9042, 4.7907, 4.8159, 4.3593, 4.8783, 4.9233, 4.4917], device='cuda:2'), covar=tensor([0.0367, 0.0245, 0.0169, 0.0128, 0.0721, 0.0219, 0.0180, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0129, 0.0167, 0.0138, 0.0193, 0.0157, 0.0118, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:24:29,163 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:40,975 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1642, 4.2881, 2.0536, 4.6570, 2.8753, 4.4136, 2.2384, 3.2193], device='cuda:2'), covar=tensor([0.0087, 0.0161, 0.1471, 0.0020, 0.0643, 0.0299, 0.1374, 0.0538], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0137, 0.0175, 0.0081, 0.0161, 0.0160, 0.0182, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 01:24:56,020 INFO [train.py:904] (2/8) Epoch 4, batch 50, loss[loss=0.2945, simple_loss=0.3516, pruned_loss=0.1188, over 16339.00 frames. ], tot_loss[loss=0.2813, simple_loss=0.3386, pruned_loss=0.112, over 739885.96 frames. ], batch size: 165, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:25:02,247 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:49,876 INFO [optim.py:368] (2/8) 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 01:26:02,162 INFO [train.py:904] (2/8) Epoch 4, batch 100, loss[loss=0.2165, simple_loss=0.2857, pruned_loss=0.07363, over 16978.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3251, pruned_loss=0.09794, over 1317648.88 frames. ], batch size: 41, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:26:23,714 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:27:11,705 INFO [train.py:904] (2/8) Epoch 4, batch 150, loss[loss=0.2588, simple_loss=0.335, pruned_loss=0.09132, over 16647.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3221, pruned_loss=0.09436, over 1755034.12 frames. ], batch size: 57, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:27:50,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9150, 4.6048, 4.8378, 5.1762, 5.2467, 4.6795, 5.2606, 5.1897], device='cuda:2'), covar=tensor([0.0680, 0.0625, 0.1099, 0.0410, 0.0371, 0.0374, 0.0400, 0.0329], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0388, 0.0509, 0.0386, 0.0295, 0.0272, 0.0313, 0.0317], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:28:04,890 INFO [optim.py:368] (2/8) 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 01:28:19,173 INFO [train.py:904] (2/8) Epoch 4, batch 200, loss[loss=0.2286, simple_loss=0.3124, pruned_loss=0.07237, over 17123.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3226, pruned_loss=0.0927, over 2105326.16 frames. ], batch size: 47, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:28:24,011 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:28:26,233 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3702, 5.2137, 5.1045, 5.0885, 4.6504, 5.1090, 5.2357, 4.7858], device='cuda:2'), covar=tensor([0.0359, 0.0211, 0.0145, 0.0133, 0.0784, 0.0247, 0.0175, 0.0338], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0146, 0.0187, 0.0155, 0.0220, 0.0177, 0.0133, 0.0192], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:28:56,754 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2614, 4.0288, 4.2389, 4.5487, 4.5920, 4.1013, 4.5925, 4.5449], device='cuda:2'), covar=tensor([0.0786, 0.0733, 0.1306, 0.0522, 0.0479, 0.0668, 0.0529, 0.0410], device='cuda:2'), in_proj_covar=tensor([0.0321, 0.0390, 0.0516, 0.0389, 0.0298, 0.0275, 0.0312, 0.0319], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:29:19,546 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1425, 2.5530, 2.5069, 4.7972, 1.9556, 4.1502, 2.4235, 2.4286], device='cuda:2'), covar=tensor([0.0328, 0.1194, 0.0606, 0.0169, 0.2309, 0.0351, 0.1223, 0.2078], device='cuda:2'), in_proj_covar=tensor([0.0275, 0.0264, 0.0218, 0.0279, 0.0326, 0.0231, 0.0245, 0.0324], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:29:26,979 INFO [train.py:904] (2/8) Epoch 4, batch 250, loss[loss=0.258, simple_loss=0.3158, pruned_loss=0.1001, over 16427.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3191, pruned_loss=0.09085, over 2382825.78 frames. ], batch size: 75, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:29:48,572 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:48,732 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:30:21,642 INFO [optim.py:368] (2/8) 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 01:30:36,156 INFO [train.py:904] (2/8) Epoch 4, batch 300, loss[loss=0.2504, simple_loss=0.306, pruned_loss=0.09744, over 16881.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3145, pruned_loss=0.08754, over 2602590.52 frames. ], batch size: 109, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:30:47,923 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6874, 6.1069, 5.7500, 5.9424, 5.2916, 5.0009, 5.5990, 6.1676], device='cuda:2'), covar=tensor([0.0528, 0.0494, 0.0773, 0.0353, 0.0532, 0.0535, 0.0424, 0.0519], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0421, 0.0362, 0.0272, 0.0274, 0.0273, 0.0334, 0.0294], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:31:16,665 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:31:43,553 INFO [train.py:904] (2/8) Epoch 4, batch 350, loss[loss=0.2158, simple_loss=0.2987, pruned_loss=0.06647, over 17266.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3103, pruned_loss=0.08482, over 2762285.02 frames. ], batch size: 45, lr: 1.74e-02, grad_scale: 1.0 2023-04-28 01:31:45,081 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6504, 4.7828, 4.7849, 4.7983, 4.7999, 5.3868, 4.9603, 4.6189], device='cuda:2'), covar=tensor([0.0997, 0.1528, 0.1026, 0.1680, 0.2332, 0.0963, 0.1174, 0.2230], device='cuda:2'), in_proj_covar=tensor([0.0246, 0.0354, 0.0324, 0.0300, 0.0399, 0.0361, 0.0280, 0.0405], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:32:20,668 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:38,673 INFO [optim.py:368] (2/8) 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 01:32:51,066 INFO [train.py:904] (2/8) Epoch 4, batch 400, loss[loss=0.2519, simple_loss=0.3042, pruned_loss=0.09976, over 16696.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3077, pruned_loss=0.08409, over 2888319.46 frames. ], batch size: 134, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:32:55,334 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1634, 1.5609, 2.2617, 2.8533, 2.7514, 3.2888, 1.7399, 3.1040], device='cuda:2'), covar=tensor([0.0054, 0.0200, 0.0135, 0.0098, 0.0076, 0.0087, 0.0198, 0.0057], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0129, 0.0114, 0.0107, 0.0103, 0.0080, 0.0121, 0.0067], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 01:33:11,868 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:33:22,040 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 01:33:31,662 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-04-28 01:33:39,124 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 01:33:43,474 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0914, 3.7086, 2.9628, 5.2302, 5.1028, 4.6241, 1.7989, 3.4833], device='cuda:2'), covar=tensor([0.1080, 0.0414, 0.0926, 0.0057, 0.0142, 0.0240, 0.1148, 0.0522], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0132, 0.0160, 0.0069, 0.0148, 0.0146, 0.0150, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 01:34:01,537 INFO [train.py:904] (2/8) Epoch 4, batch 450, loss[loss=0.2445, simple_loss=0.2973, pruned_loss=0.0959, over 16732.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3044, pruned_loss=0.08176, over 2989472.02 frames. ], batch size: 124, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:34:05,074 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:17,423 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:35,341 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:45,892 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:50,748 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 01:34:56,469 INFO [optim.py:368] (2/8) 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 01:35:09,304 INFO [train.py:904] (2/8) Epoch 4, batch 500, loss[loss=0.2127, simple_loss=0.2973, pruned_loss=0.06409, over 16644.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3033, pruned_loss=0.08083, over 3068461.34 frames. ], batch size: 57, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:35:28,333 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:59,027 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:03,597 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3416, 5.1720, 5.1162, 5.0754, 4.5392, 5.1714, 5.1870, 4.8826], device='cuda:2'), covar=tensor([0.0370, 0.0271, 0.0166, 0.0134, 0.0964, 0.0211, 0.0148, 0.0290], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0154, 0.0201, 0.0166, 0.0237, 0.0188, 0.0141, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:36:09,087 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:17,831 INFO [train.py:904] (2/8) Epoch 4, batch 550, loss[loss=0.2256, simple_loss=0.3042, pruned_loss=0.07349, over 16377.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3023, pruned_loss=0.0802, over 3114303.21 frames. ], batch size: 68, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:36:33,175 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:36:33,261 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:36:39,322 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:52,754 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0376, 2.7639, 2.6796, 1.7431, 2.8109, 2.8381, 2.3828, 2.2961], device='cuda:2'), covar=tensor([0.0675, 0.0143, 0.0190, 0.1018, 0.0097, 0.0097, 0.0404, 0.0441], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0082, 0.0080, 0.0143, 0.0074, 0.0075, 0.0112, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 01:37:13,481 INFO [optim.py:368] (2/8) 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 01:37:26,136 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 01:37:28,709 INFO [train.py:904] (2/8) Epoch 4, batch 600, loss[loss=0.1934, simple_loss=0.2662, pruned_loss=0.06029, over 16762.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3011, pruned_loss=0.07994, over 3161972.48 frames. ], batch size: 39, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:37:46,198 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:51,682 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5485, 4.4757, 5.0889, 5.0829, 5.0712, 4.6228, 4.6356, 4.4801], device='cuda:2'), covar=tensor([0.0239, 0.0309, 0.0290, 0.0333, 0.0366, 0.0277, 0.0694, 0.0317], device='cuda:2'), in_proj_covar=tensor([0.0209, 0.0208, 0.0214, 0.0213, 0.0257, 0.0228, 0.0321, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 01:37:58,248 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1054, 4.3262, 2.1229, 4.6400, 2.8584, 4.5321, 1.9548, 3.1606], device='cuda:2'), covar=tensor([0.0078, 0.0172, 0.1436, 0.0025, 0.0661, 0.0256, 0.1404, 0.0523], device='cuda:2'), in_proj_covar=tensor([0.0099, 0.0140, 0.0169, 0.0080, 0.0157, 0.0166, 0.0174, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 01:37:58,264 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:38:36,813 INFO [train.py:904] (2/8) Epoch 4, batch 650, loss[loss=0.2236, simple_loss=0.2854, pruned_loss=0.08089, over 12039.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3, pruned_loss=0.07952, over 3187078.67 frames. ], batch size: 247, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:38:55,171 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5852, 4.2055, 4.3099, 1.6523, 4.5412, 4.5384, 3.0651, 3.4932], device='cuda:2'), covar=tensor([0.0739, 0.0123, 0.0187, 0.1395, 0.0060, 0.0047, 0.0345, 0.0335], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0081, 0.0081, 0.0142, 0.0074, 0.0074, 0.0112, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 01:39:30,534 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 4, batch 700, loss[loss=0.2067, simple_loss=0.278, pruned_loss=0.06773, over 16681.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.2997, pruned_loss=0.07902, over 3215858.78 frames. ], batch size: 89, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:39:46,436 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9730, 2.4648, 2.3940, 3.2706, 3.1308, 3.2448, 1.9202, 2.7603], device='cuda:2'), covar=tensor([0.1112, 0.0512, 0.0946, 0.0090, 0.0296, 0.0365, 0.1059, 0.0657], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0136, 0.0165, 0.0073, 0.0158, 0.0155, 0.0155, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 01:40:53,737 INFO [train.py:904] (2/8) Epoch 4, batch 750, loss[loss=0.2293, simple_loss=0.2895, pruned_loss=0.08453, over 16725.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.2997, pruned_loss=0.07975, over 3240449.09 frames. ], batch size: 89, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:41:48,310 INFO [optim.py:368] (2/8) 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 01:42:01,619 INFO [train.py:904] (2/8) Epoch 4, batch 800, loss[loss=0.2312, simple_loss=0.2938, pruned_loss=0.0843, over 16744.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.2994, pruned_loss=0.07905, over 3259344.01 frames. ], batch size: 89, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:42:14,288 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:36,352 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 01:42:44,446 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:55,440 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:43:11,602 INFO [train.py:904] (2/8) Epoch 4, batch 850, loss[loss=0.2208, simple_loss=0.287, pruned_loss=0.07733, over 15627.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.299, pruned_loss=0.07894, over 3274116.96 frames. ], batch size: 191, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:43:24,613 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:07,361 INFO [optim.py:368] (2/8) 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 01:44:19,701 INFO [train.py:904] (2/8) Epoch 4, batch 900, loss[loss=0.2037, simple_loss=0.2784, pruned_loss=0.06451, over 16022.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.2979, pruned_loss=0.07785, over 3275953.91 frames. ], batch size: 35, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:44:23,395 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4881, 4.2489, 3.8054, 1.7820, 3.0027, 2.4377, 3.7586, 4.0830], device='cuda:2'), covar=tensor([0.0280, 0.0537, 0.0483, 0.1691, 0.0758, 0.0992, 0.0658, 0.0751], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0127, 0.0156, 0.0145, 0.0137, 0.0129, 0.0140, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 01:44:32,185 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:42,744 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:45:31,311 INFO [train.py:904] (2/8) Epoch 4, batch 950, loss[loss=0.239, simple_loss=0.3046, pruned_loss=0.08667, over 15391.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.2987, pruned_loss=0.07841, over 3284536.43 frames. ], batch size: 190, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:45:41,880 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:45:43,136 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8624, 2.5715, 2.5991, 4.2786, 2.0586, 3.6358, 2.3932, 2.5520], device='cuda:2'), covar=tensor([0.0354, 0.1022, 0.0539, 0.0213, 0.1920, 0.0381, 0.1213, 0.1502], device='cuda:2'), in_proj_covar=tensor([0.0284, 0.0266, 0.0220, 0.0284, 0.0329, 0.0241, 0.0248, 0.0336], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:45:54,345 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0960, 3.9849, 3.7281, 1.6111, 3.9402, 3.9145, 3.0705, 3.0028], device='cuda:2'), covar=tensor([0.1189, 0.0110, 0.0249, 0.1483, 0.0097, 0.0091, 0.0360, 0.0493], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0081, 0.0081, 0.0143, 0.0076, 0.0075, 0.0112, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 01:46:13,254 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5434, 4.3256, 3.8749, 1.8886, 2.9586, 2.4078, 3.7053, 4.0934], device='cuda:2'), covar=tensor([0.0243, 0.0385, 0.0408, 0.1530, 0.0677, 0.0921, 0.0636, 0.0794], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0126, 0.0154, 0.0145, 0.0136, 0.0129, 0.0141, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 01:46:26,066 INFO [optim.py:368] (2/8) 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 01:46:38,672 INFO [train.py:904] (2/8) Epoch 4, batch 1000, loss[loss=0.2064, simple_loss=0.2729, pruned_loss=0.06995, over 16552.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.2977, pruned_loss=0.07831, over 3287608.91 frames. ], batch size: 75, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:47:04,155 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8344, 4.3116, 3.2380, 2.5172, 3.0534, 2.2172, 4.3599, 4.4220], device='cuda:2'), covar=tensor([0.2125, 0.0588, 0.1291, 0.1370, 0.2571, 0.1578, 0.0393, 0.0493], device='cuda:2'), in_proj_covar=tensor([0.0271, 0.0245, 0.0257, 0.0222, 0.0290, 0.0200, 0.0221, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 01:47:06,286 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:07,486 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2537, 3.9949, 3.2873, 1.8752, 2.9389, 2.2135, 3.7994, 3.8887], device='cuda:2'), covar=tensor([0.0255, 0.0438, 0.0548, 0.1600, 0.0626, 0.0966, 0.0520, 0.0641], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0127, 0.0155, 0.0146, 0.0137, 0.0130, 0.0142, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 01:47:36,272 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:47:48,996 INFO [train.py:904] (2/8) Epoch 4, batch 1050, loss[loss=0.2576, simple_loss=0.3188, pruned_loss=0.09823, over 16545.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.2973, pruned_loss=0.07854, over 3293962.80 frames. ], batch size: 75, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:48:45,340 INFO [optim.py:368] (2/8) 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 01:49:00,846 INFO [train.py:904] (2/8) Epoch 4, batch 1100, loss[loss=0.2361, simple_loss=0.2989, pruned_loss=0.0867, over 15920.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.2951, pruned_loss=0.07706, over 3297783.27 frames. ], batch size: 35, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:49:02,385 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:49:11,024 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:41,943 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:50,417 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1715, 4.1679, 4.6456, 4.6221, 4.6006, 4.2586, 4.2685, 4.2024], device='cuda:2'), covar=tensor([0.0217, 0.0372, 0.0255, 0.0353, 0.0332, 0.0256, 0.0636, 0.0337], device='cuda:2'), in_proj_covar=tensor([0.0215, 0.0212, 0.0219, 0.0216, 0.0258, 0.0231, 0.0328, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 01:49:53,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0134, 5.0564, 4.8566, 4.7968, 3.9801, 4.9251, 5.0410, 4.4846], device='cuda:2'), covar=tensor([0.0506, 0.0239, 0.0223, 0.0183, 0.1295, 0.0264, 0.0198, 0.0362], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0160, 0.0208, 0.0173, 0.0245, 0.0192, 0.0150, 0.0209], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:49:55,720 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:09,609 INFO [train.py:904] (2/8) Epoch 4, batch 1150, loss[loss=0.1828, simple_loss=0.257, pruned_loss=0.05433, over 16966.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2944, pruned_loss=0.07578, over 3305094.43 frames. ], batch size: 41, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:50:18,982 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:49,049 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:56,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2044, 5.0655, 4.9444, 4.9174, 4.4450, 4.9523, 5.0382, 4.6294], device='cuda:2'), covar=tensor([0.0346, 0.0204, 0.0171, 0.0146, 0.0856, 0.0258, 0.0188, 0.0373], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0160, 0.0208, 0.0173, 0.0246, 0.0193, 0.0149, 0.0210], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:51:00,549 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:04,258 INFO [optim.py:368] (2/8) 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 01:51:19,069 INFO [train.py:904] (2/8) Epoch 4, batch 1200, loss[loss=0.2521, simple_loss=0.3107, pruned_loss=0.09681, over 16869.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2933, pruned_loss=0.07493, over 3308843.07 frames. ], batch size: 96, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:51:41,067 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:51:48,039 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 01:52:12,255 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-28 01:52:27,378 INFO [train.py:904] (2/8) Epoch 4, batch 1250, loss[loss=0.23, simple_loss=0.3091, pruned_loss=0.07551, over 17093.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.2944, pruned_loss=0.07584, over 3313770.16 frames. ], batch size: 47, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:52:31,863 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:40,756 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 01:52:47,962 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:53:22,524 INFO [optim.py:368] (2/8) 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 01:53:35,321 INFO [train.py:904] (2/8) Epoch 4, batch 1300, loss[loss=0.2183, simple_loss=0.2967, pruned_loss=0.06997, over 17217.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.2945, pruned_loss=0.07558, over 3316144.95 frames. ], batch size: 44, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:53:54,673 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:53:54,833 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:26,689 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:44,573 INFO [train.py:904] (2/8) Epoch 4, batch 1350, loss[loss=0.2033, simple_loss=0.2845, pruned_loss=0.06108, over 16838.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.2952, pruned_loss=0.07555, over 3307738.96 frames. ], batch size: 42, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:04,633 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 01:55:14,531 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:39,665 INFO [optim.py:368] (2/8) 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 01:55:45,317 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1982, 4.1506, 4.1476, 4.2114, 4.1066, 4.6783, 4.3274, 3.9566], device='cuda:2'), covar=tensor([0.1328, 0.1382, 0.1143, 0.1605, 0.2430, 0.0972, 0.1027, 0.2123], device='cuda:2'), in_proj_covar=tensor([0.0258, 0.0364, 0.0337, 0.0310, 0.0413, 0.0367, 0.0288, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:55:48,222 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:55:50,726 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:52,765 INFO [train.py:904] (2/8) Epoch 4, batch 1400, loss[loss=0.2061, simple_loss=0.2963, pruned_loss=0.05792, over 17056.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.2945, pruned_loss=0.07505, over 3316349.92 frames. ], batch size: 55, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:54,859 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3682, 4.0517, 3.8931, 1.7094, 4.1589, 4.0624, 3.2421, 3.0168], device='cuda:2'), covar=tensor([0.0796, 0.0066, 0.0179, 0.1250, 0.0047, 0.0057, 0.0266, 0.0404], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0080, 0.0079, 0.0143, 0.0072, 0.0076, 0.0109, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 01:56:32,139 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:40,510 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:46,657 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2573, 4.1713, 4.1668, 3.6658, 4.1971, 1.8389, 3.9323, 3.9923], device='cuda:2'), covar=tensor([0.0054, 0.0055, 0.0077, 0.0237, 0.0054, 0.1353, 0.0075, 0.0115], device='cuda:2'), in_proj_covar=tensor([0.0086, 0.0077, 0.0115, 0.0127, 0.0085, 0.0131, 0.0102, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:57:02,742 INFO [train.py:904] (2/8) Epoch 4, batch 1450, loss[loss=0.2367, simple_loss=0.2889, pruned_loss=0.09225, over 16443.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.293, pruned_loss=0.07454, over 3319657.36 frames. ], batch size: 146, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:57:22,473 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:56,660 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:00,353 INFO [optim.py:368] (2/8) 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 01:58:12,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2089, 5.2407, 4.9641, 4.3372, 5.0660, 2.0844, 4.6941, 5.1701], device='cuda:2'), covar=tensor([0.0051, 0.0044, 0.0067, 0.0303, 0.0050, 0.1350, 0.0076, 0.0087], device='cuda:2'), in_proj_covar=tensor([0.0085, 0.0076, 0.0112, 0.0124, 0.0083, 0.0128, 0.0101, 0.0114], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 01:58:12,762 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:13,459 INFO [train.py:904] (2/8) Epoch 4, batch 1500, loss[loss=0.1969, simple_loss=0.2759, pruned_loss=0.0589, over 17188.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.2937, pruned_loss=0.07562, over 3319179.35 frames. ], batch size: 46, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:58:47,441 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:26,619 INFO [train.py:904] (2/8) Epoch 4, batch 1550, loss[loss=0.2537, simple_loss=0.2994, pruned_loss=0.104, over 16707.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.2961, pruned_loss=0.07857, over 3320259.87 frames. ], batch size: 89, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:59:30,861 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 01:59:40,585 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:21,779 INFO [optim.py:368] (2/8) 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 02:00:34,399 INFO [train.py:904] (2/8) Epoch 4, batch 1600, loss[loss=0.2139, simple_loss=0.296, pruned_loss=0.06591, over 17088.00 frames. ], tot_loss[loss=0.228, simple_loss=0.2979, pruned_loss=0.07901, over 3322370.23 frames. ], batch size: 47, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:00:46,072 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:52,705 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:01:41,676 INFO [train.py:904] (2/8) Epoch 4, batch 1650, loss[loss=0.1994, simple_loss=0.2846, pruned_loss=0.05708, over 17211.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.2995, pruned_loss=0.07981, over 3308476.53 frames. ], batch size: 45, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:01:58,771 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:17,852 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1537, 5.4492, 5.0535, 5.3492, 4.8739, 4.5755, 4.9916, 5.4898], device='cuda:2'), covar=tensor([0.0607, 0.0624, 0.0902, 0.0378, 0.0600, 0.0658, 0.0623, 0.0660], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0453, 0.0388, 0.0282, 0.0288, 0.0284, 0.0353, 0.0312], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:02:37,464 INFO [optim.py:368] (2/8) 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 02:02:40,735 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:45,018 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:02:50,054 INFO [train.py:904] (2/8) Epoch 4, batch 1700, loss[loss=0.2012, simple_loss=0.2961, pruned_loss=0.05312, over 17131.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3007, pruned_loss=0.07983, over 3311862.71 frames. ], batch size: 49, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:03:02,609 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 02:03:29,670 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:04:01,131 INFO [train.py:904] (2/8) Epoch 4, batch 1750, loss[loss=0.2468, simple_loss=0.3116, pruned_loss=0.09103, over 16305.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3035, pruned_loss=0.08111, over 3291584.35 frames. ], batch size: 165, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:04:02,537 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-04-28 02:04:47,676 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:58,762 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 4, batch 1800, loss[loss=0.2101, simple_loss=0.2979, pruned_loss=0.0612, over 17119.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3047, pruned_loss=0.08111, over 3298658.86 frames. ], batch size: 47, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:05:39,577 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:04,366 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0338, 4.9690, 4.7897, 4.8178, 4.3856, 4.8961, 4.8253, 4.5116], device='cuda:2'), covar=tensor([0.0328, 0.0207, 0.0172, 0.0137, 0.0798, 0.0213, 0.0190, 0.0351], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0162, 0.0208, 0.0175, 0.0248, 0.0195, 0.0151, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:06:18,353 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4825, 4.5907, 4.0102, 2.1099, 3.0405, 2.5233, 3.8030, 4.0867], device='cuda:2'), covar=tensor([0.0243, 0.0365, 0.0372, 0.1392, 0.0651, 0.0852, 0.0590, 0.0629], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0128, 0.0154, 0.0141, 0.0134, 0.0126, 0.0139, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:06:18,946 INFO [train.py:904] (2/8) Epoch 4, batch 1850, loss[loss=0.1986, simple_loss=0.2884, pruned_loss=0.05445, over 17276.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3052, pruned_loss=0.08046, over 3312651.04 frames. ], batch size: 52, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:06:26,551 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:54,114 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 02:07:18,152 INFO [optim.py:368] (2/8) 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 02:07:29,364 INFO [train.py:904] (2/8) Epoch 4, batch 1900, loss[loss=0.2411, simple_loss=0.3176, pruned_loss=0.08229, over 17129.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3048, pruned_loss=0.07957, over 3313884.20 frames. ], batch size: 49, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:07:43,114 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:00,282 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0864, 1.8285, 2.3062, 2.7377, 2.8311, 3.4838, 1.8109, 3.2024], device='cuda:2'), covar=tensor([0.0068, 0.0193, 0.0128, 0.0134, 0.0089, 0.0049, 0.0196, 0.0043], device='cuda:2'), in_proj_covar=tensor([0.0101, 0.0128, 0.0116, 0.0114, 0.0107, 0.0080, 0.0124, 0.0070], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 02:08:41,339 INFO [train.py:904] (2/8) Epoch 4, batch 1950, loss[loss=0.2172, simple_loss=0.2951, pruned_loss=0.06969, over 16820.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3045, pruned_loss=0.07877, over 3308730.90 frames. ], batch size: 42, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:08:50,794 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:51,315 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-28 02:09:15,937 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8878, 4.3368, 3.4050, 2.4329, 3.1703, 2.3923, 4.5263, 4.3966], device='cuda:2'), covar=tensor([0.2033, 0.0585, 0.1056, 0.1364, 0.2306, 0.1367, 0.0314, 0.0554], device='cuda:2'), in_proj_covar=tensor([0.0270, 0.0250, 0.0259, 0.0226, 0.0296, 0.0196, 0.0227, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:09:37,979 INFO [optim.py:368] (2/8) 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 02:09:39,320 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:47,991 INFO [train.py:904] (2/8) Epoch 4, batch 2000, loss[loss=0.3305, simple_loss=0.3726, pruned_loss=0.1442, over 12056.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3031, pruned_loss=0.07804, over 3316947.41 frames. ], batch size: 246, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:10:04,157 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0053, 5.6144, 5.5769, 5.5763, 5.5742, 6.1165, 5.7965, 5.5176], device='cuda:2'), covar=tensor([0.0662, 0.1404, 0.1450, 0.1527, 0.2290, 0.0883, 0.0868, 0.2210], device='cuda:2'), in_proj_covar=tensor([0.0263, 0.0370, 0.0348, 0.0313, 0.0426, 0.0375, 0.0298, 0.0422], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:10:28,197 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:45,038 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:58,283 INFO [train.py:904] (2/8) Epoch 4, batch 2050, loss[loss=0.2456, simple_loss=0.3024, pruned_loss=0.09441, over 16664.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3029, pruned_loss=0.07844, over 3316392.33 frames. ], batch size: 89, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:11:19,003 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 02:11:34,833 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:44,268 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:57,877 INFO [optim.py:368] (2/8) 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 02:12:07,692 INFO [train.py:904] (2/8) Epoch 4, batch 2100, loss[loss=0.2023, simple_loss=0.2737, pruned_loss=0.06548, over 16027.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.303, pruned_loss=0.0787, over 3317873.38 frames. ], batch size: 35, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:12:36,415 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:51,000 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:17,653 INFO [train.py:904] (2/8) Epoch 4, batch 2150, loss[loss=0.208, simple_loss=0.2879, pruned_loss=0.064, over 17211.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3052, pruned_loss=0.08028, over 3314637.42 frames. ], batch size: 45, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:13:24,704 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:42,049 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:14:15,065 INFO [optim.py:368] (2/8) 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 02:14:26,823 INFO [train.py:904] (2/8) Epoch 4, batch 2200, loss[loss=0.218, simple_loss=0.3062, pruned_loss=0.06487, over 17148.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3049, pruned_loss=0.07974, over 3326156.09 frames. ], batch size: 48, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:14:28,668 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 02:14:30,433 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:15:34,029 INFO [train.py:904] (2/8) Epoch 4, batch 2250, loss[loss=0.2738, simple_loss=0.3256, pruned_loss=0.111, over 16901.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3065, pruned_loss=0.08141, over 3325584.68 frames. ], batch size: 109, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:16:28,736 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9788, 3.3393, 3.5051, 3.4664, 3.4192, 3.1493, 3.1854, 3.2646], device='cuda:2'), covar=tensor([0.0339, 0.0352, 0.0381, 0.0463, 0.0448, 0.0416, 0.0810, 0.0411], device='cuda:2'), in_proj_covar=tensor([0.0224, 0.0219, 0.0226, 0.0228, 0.0274, 0.0242, 0.0345, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 02:16:32,021 INFO [optim.py:368] (2/8) 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 02:16:41,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2461, 4.2497, 4.2766, 4.2871, 4.2688, 4.8690, 4.6063, 4.2329], device='cuda:2'), covar=tensor([0.1463, 0.1534, 0.1294, 0.1841, 0.2646, 0.1065, 0.1046, 0.2371], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0371, 0.0347, 0.0313, 0.0421, 0.0380, 0.0297, 0.0419], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:16:42,204 INFO [train.py:904] (2/8) Epoch 4, batch 2300, loss[loss=0.2266, simple_loss=0.2946, pruned_loss=0.07932, over 16282.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3061, pruned_loss=0.08118, over 3319645.11 frames. ], batch size: 165, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:16:48,907 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 02:17:51,370 INFO [train.py:904] (2/8) Epoch 4, batch 2350, loss[loss=0.1715, simple_loss=0.254, pruned_loss=0.04452, over 16822.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3051, pruned_loss=0.07968, over 3333130.81 frames. ], batch size: 39, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:18:48,983 INFO [optim.py:368] (2/8) 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 02:18:57,192 INFO [train.py:904] (2/8) Epoch 4, batch 2400, loss[loss=0.1881, simple_loss=0.2662, pruned_loss=0.055, over 17026.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3058, pruned_loss=0.07968, over 3328078.74 frames. ], batch size: 41, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:20:06,099 INFO [train.py:904] (2/8) Epoch 4, batch 2450, loss[loss=0.2369, simple_loss=0.3025, pruned_loss=0.08567, over 16773.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3068, pruned_loss=0.07974, over 3328735.56 frames. ], batch size: 83, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:03,694 INFO [optim.py:368] (2/8) 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 02:21:07,163 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2558, 4.3417, 1.7961, 4.6393, 2.5305, 4.5798, 2.2571, 3.0316], device='cuda:2'), covar=tensor([0.0069, 0.0138, 0.1342, 0.0028, 0.0694, 0.0198, 0.1104, 0.0512], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0147, 0.0170, 0.0084, 0.0158, 0.0178, 0.0180, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:21:13,283 INFO [train.py:904] (2/8) Epoch 4, batch 2500, loss[loss=0.1856, simple_loss=0.2664, pruned_loss=0.05242, over 16969.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3056, pruned_loss=0.07871, over 3333707.18 frames. ], batch size: 41, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:18,157 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6600, 5.9125, 5.5921, 5.8786, 5.2665, 4.8731, 5.4636, 6.0154], device='cuda:2'), covar=tensor([0.0511, 0.0611, 0.0900, 0.0315, 0.0516, 0.0595, 0.0439, 0.0609], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0450, 0.0381, 0.0277, 0.0282, 0.0280, 0.0349, 0.0307], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:21:34,449 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:21:39,027 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3948, 3.9857, 4.3751, 2.9854, 3.7817, 4.3565, 3.8535, 2.5608], device='cuda:2'), covar=tensor([0.0304, 0.0032, 0.0018, 0.0208, 0.0032, 0.0023, 0.0024, 0.0261], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0055, 0.0057, 0.0107, 0.0058, 0.0064, 0.0058, 0.0103], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:22:21,140 INFO [train.py:904] (2/8) Epoch 4, batch 2550, loss[loss=0.2047, simple_loss=0.2882, pruned_loss=0.06057, over 16931.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3056, pruned_loss=0.07876, over 3339867.99 frames. ], batch size: 41, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:22:47,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0185, 3.2922, 3.6517, 2.4421, 3.4127, 3.6230, 3.4866, 1.7421], device='cuda:2'), covar=tensor([0.0294, 0.0086, 0.0040, 0.0215, 0.0052, 0.0061, 0.0045, 0.0348], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0056, 0.0058, 0.0107, 0.0059, 0.0065, 0.0058, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:22:57,628 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 02:23:30,981 INFO [train.py:904] (2/8) Epoch 4, batch 2600, loss[loss=0.2569, simple_loss=0.3224, pruned_loss=0.09573, over 16531.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3066, pruned_loss=0.07864, over 3332591.51 frames. ], batch size: 75, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:23:31,479 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2545, 3.8707, 3.1758, 1.9966, 2.7250, 2.2177, 3.7112, 3.7166], device='cuda:2'), covar=tensor([0.0219, 0.0482, 0.0527, 0.1400, 0.0751, 0.0933, 0.0479, 0.0585], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0128, 0.0152, 0.0141, 0.0132, 0.0126, 0.0139, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:23:55,308 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:24:39,065 INFO [train.py:904] (2/8) Epoch 4, batch 2650, loss[loss=0.2442, simple_loss=0.3107, pruned_loss=0.0889, over 16863.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3068, pruned_loss=0.07831, over 3326234.03 frames. ], batch size: 96, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:04,922 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2736, 4.1639, 4.7263, 4.6886, 4.6792, 4.2601, 4.3526, 4.1270], device='cuda:2'), covar=tensor([0.0230, 0.0385, 0.0248, 0.0347, 0.0389, 0.0290, 0.0686, 0.0424], device='cuda:2'), in_proj_covar=tensor([0.0223, 0.0219, 0.0225, 0.0229, 0.0274, 0.0237, 0.0344, 0.0206], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 02:25:19,563 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:25:38,560 INFO [optim.py:368] (2/8) 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 02:25:46,694 INFO [train.py:904] (2/8) Epoch 4, batch 2700, loss[loss=0.2431, simple_loss=0.3082, pruned_loss=0.089, over 16664.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3074, pruned_loss=0.07801, over 3320397.40 frames. ], batch size: 134, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:54,285 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 4, batch 2750, loss[loss=0.2055, simple_loss=0.2931, pruned_loss=0.05898, over 17196.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3071, pruned_loss=0.07742, over 3318570.91 frames. ], batch size: 46, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:27:16,305 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:27:54,083 INFO [optim.py:368] (2/8) 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 02:28:04,360 INFO [train.py:904] (2/8) Epoch 4, batch 2800, loss[loss=0.2429, simple_loss=0.3131, pruned_loss=0.08635, over 16743.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3075, pruned_loss=0.07711, over 3325458.35 frames. ], batch size: 83, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:29:10,594 INFO [train.py:904] (2/8) Epoch 4, batch 2850, loss[loss=0.211, simple_loss=0.2886, pruned_loss=0.06677, over 17109.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3058, pruned_loss=0.07659, over 3331782.46 frames. ], batch size: 47, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:29:39,600 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:30:09,813 INFO [optim.py:368] (2/8) 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 02:30:20,484 INFO [train.py:904] (2/8) Epoch 4, batch 2900, loss[loss=0.327, simple_loss=0.3561, pruned_loss=0.1489, over 11953.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.305, pruned_loss=0.07773, over 3323031.97 frames. ], batch size: 247, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:31:05,666 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 02:31:21,942 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1792, 4.6167, 2.0761, 4.8168, 2.9817, 4.7670, 2.4665, 3.2335], device='cuda:2'), covar=tensor([0.0096, 0.0135, 0.1461, 0.0047, 0.0672, 0.0290, 0.1259, 0.0516], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0148, 0.0170, 0.0085, 0.0156, 0.0180, 0.0182, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:31:29,005 INFO [train.py:904] (2/8) Epoch 4, batch 2950, loss[loss=0.2514, simple_loss=0.3228, pruned_loss=0.09003, over 16676.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3039, pruned_loss=0.07802, over 3319529.42 frames. ], batch size: 68, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:01,322 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:32:27,160 INFO [optim.py:368] (2/8) 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 02:32:35,894 INFO [train.py:904] (2/8) Epoch 4, batch 3000, loss[loss=0.2273, simple_loss=0.2927, pruned_loss=0.08095, over 16868.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3043, pruned_loss=0.07905, over 3313261.18 frames. ], batch size: 96, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:35,894 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 02:32:45,547 INFO [train.py:938] (2/8) Epoch 4, validation: loss=0.1627, simple_loss=0.2694, pruned_loss=0.02796, over 944034.00 frames. 2023-04-28 02:32:45,547 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 02:32:54,616 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5123, 4.5550, 4.0970, 1.9834, 3.0093, 2.4936, 3.7873, 4.0750], device='cuda:2'), covar=tensor([0.0297, 0.0450, 0.0401, 0.1529, 0.0688, 0.0984, 0.0697, 0.0907], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0130, 0.0154, 0.0143, 0.0132, 0.0128, 0.0144, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:33:12,877 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:54,593 INFO [train.py:904] (2/8) Epoch 4, batch 3050, loss[loss=0.2067, simple_loss=0.2965, pruned_loss=0.0584, over 17117.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3037, pruned_loss=0.07851, over 3316049.60 frames. ], batch size: 48, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:33:58,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7827, 5.1051, 5.2003, 5.1927, 4.9565, 5.6668, 5.4429, 5.0982], device='cuda:2'), covar=tensor([0.0957, 0.1310, 0.1239, 0.1376, 0.2426, 0.0781, 0.0980, 0.1962], device='cuda:2'), in_proj_covar=tensor([0.0270, 0.0372, 0.0346, 0.0315, 0.0421, 0.0373, 0.0293, 0.0427], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:34:07,856 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:29,600 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9885, 5.5704, 5.5880, 5.6383, 5.6897, 6.1353, 5.9506, 5.6226], device='cuda:2'), covar=tensor([0.0681, 0.1534, 0.1454, 0.1688, 0.2462, 0.0804, 0.1023, 0.2028], device='cuda:2'), in_proj_covar=tensor([0.0270, 0.0376, 0.0350, 0.0318, 0.0427, 0.0379, 0.0296, 0.0432], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:34:37,998 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:54,167 INFO [optim.py:368] (2/8) 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 02:34:54,666 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9217, 4.4925, 4.4753, 1.9120, 4.8254, 4.8137, 3.3993, 3.8701], device='cuda:2'), covar=tensor([0.0692, 0.0099, 0.0136, 0.1151, 0.0033, 0.0041, 0.0264, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0083, 0.0083, 0.0141, 0.0073, 0.0075, 0.0113, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 02:35:02,736 INFO [train.py:904] (2/8) Epoch 4, batch 3100, loss[loss=0.2306, simple_loss=0.2928, pruned_loss=0.08425, over 16277.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3036, pruned_loss=0.07902, over 3304740.46 frames. ], batch size: 165, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:35:51,165 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3528, 4.2502, 4.8196, 4.8218, 4.7799, 4.3672, 4.3826, 4.2460], device='cuda:2'), covar=tensor([0.0260, 0.0336, 0.0256, 0.0317, 0.0358, 0.0274, 0.0760, 0.0374], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0224, 0.0227, 0.0234, 0.0277, 0.0240, 0.0348, 0.0209], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 02:36:10,679 INFO [train.py:904] (2/8) Epoch 4, batch 3150, loss[loss=0.2387, simple_loss=0.3181, pruned_loss=0.07965, over 17127.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3017, pruned_loss=0.07715, over 3315687.25 frames. ], batch size: 53, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:36:28,580 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8006, 4.1063, 3.7670, 3.9820, 3.5732, 3.6559, 3.7542, 4.0347], device='cuda:2'), covar=tensor([0.0714, 0.0713, 0.1004, 0.0444, 0.0733, 0.1304, 0.0710, 0.0943], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0465, 0.0402, 0.0294, 0.0298, 0.0291, 0.0368, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:36:39,536 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:06,233 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 02:37:11,327 INFO [optim.py:368] (2/8) 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 02:37:18,478 INFO [train.py:904] (2/8) Epoch 4, batch 3200, loss[loss=0.2524, simple_loss=0.3173, pruned_loss=0.09379, over 15521.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3003, pruned_loss=0.07636, over 3310423.32 frames. ], batch size: 191, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:37:44,850 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:54,516 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:38:25,339 INFO [train.py:904] (2/8) Epoch 4, batch 3250, loss[loss=0.2826, simple_loss=0.3586, pruned_loss=0.1033, over 15569.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3011, pruned_loss=0.07599, over 3315707.11 frames. ], batch size: 191, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:38:58,270 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:16,401 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:39:26,580 INFO [optim.py:368] (2/8) 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 02:39:36,477 INFO [train.py:904] (2/8) Epoch 4, batch 3300, loss[loss=0.32, simple_loss=0.3739, pruned_loss=0.133, over 12328.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3035, pruned_loss=0.07724, over 3313652.10 frames. ], batch size: 246, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:39:45,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8934, 4.4384, 4.5554, 3.2369, 4.2852, 4.6437, 4.2358, 2.5427], device='cuda:2'), covar=tensor([0.0245, 0.0023, 0.0020, 0.0180, 0.0025, 0.0021, 0.0021, 0.0237], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0053, 0.0056, 0.0106, 0.0057, 0.0063, 0.0058, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:39:57,885 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2425, 5.5422, 5.2356, 5.4016, 4.9299, 4.5860, 5.0664, 5.6154], device='cuda:2'), covar=tensor([0.0661, 0.0681, 0.0863, 0.0409, 0.0557, 0.0641, 0.0496, 0.0665], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0459, 0.0391, 0.0289, 0.0293, 0.0289, 0.0362, 0.0319], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:40:07,409 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:40:45,511 INFO [train.py:904] (2/8) Epoch 4, batch 3350, loss[loss=0.2254, simple_loss=0.2931, pruned_loss=0.07885, over 16825.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3033, pruned_loss=0.07733, over 3314368.88 frames. ], batch size: 102, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:51,107 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4137, 2.3231, 1.8819, 1.9922, 2.8752, 2.7237, 3.5764, 3.1837], device='cuda:2'), covar=tensor([0.0028, 0.0151, 0.0173, 0.0199, 0.0089, 0.0128, 0.0050, 0.0078], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0140, 0.0137, 0.0137, 0.0134, 0.0141, 0.0107, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-28 02:40:58,957 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:21,438 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:44,267 INFO [optim.py:368] (2/8) 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 02:41:52,910 INFO [train.py:904] (2/8) Epoch 4, batch 3400, loss[loss=0.199, simple_loss=0.2849, pruned_loss=0.0566, over 17228.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3026, pruned_loss=0.07657, over 3319865.94 frames. ], batch size: 45, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:42:04,788 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:42:56,427 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5658, 3.3856, 2.7403, 2.2622, 2.4974, 2.0248, 3.4455, 3.5059], device='cuda:2'), covar=tensor([0.1666, 0.0572, 0.0945, 0.1294, 0.1916, 0.1371, 0.0378, 0.0537], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0249, 0.0262, 0.0231, 0.0303, 0.0197, 0.0231, 0.0247], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:43:03,517 INFO [train.py:904] (2/8) Epoch 4, batch 3450, loss[loss=0.1904, simple_loss=0.2739, pruned_loss=0.05349, over 17237.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3011, pruned_loss=0.0765, over 3317003.31 frames. ], batch size: 45, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:43:53,878 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1802, 4.6888, 2.2212, 4.8735, 2.8946, 4.8492, 2.3076, 3.2475], device='cuda:2'), covar=tensor([0.0095, 0.0142, 0.1276, 0.0028, 0.0664, 0.0206, 0.1229, 0.0520], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0152, 0.0174, 0.0086, 0.0158, 0.0183, 0.0184, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:43:57,712 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:05,158 INFO [optim.py:368] (2/8) 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 02:44:13,345 INFO [train.py:904] (2/8) Epoch 4, batch 3500, loss[loss=0.1861, simple_loss=0.2642, pruned_loss=0.05401, over 16824.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.2984, pruned_loss=0.0751, over 3323913.30 frames. ], batch size: 42, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:45:26,889 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:45:27,574 INFO [train.py:904] (2/8) Epoch 4, batch 3550, loss[loss=0.2124, simple_loss=0.2784, pruned_loss=0.07317, over 16356.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.2989, pruned_loss=0.0759, over 3315184.45 frames. ], batch size: 146, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:45:29,528 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 02:46:11,015 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:46:24,682 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9930, 3.2907, 3.4717, 3.4760, 3.4430, 3.1849, 3.2431, 3.3112], device='cuda:2'), covar=tensor([0.0352, 0.0438, 0.0343, 0.0396, 0.0392, 0.0403, 0.0666, 0.0411], device='cuda:2'), in_proj_covar=tensor([0.0231, 0.0224, 0.0231, 0.0236, 0.0282, 0.0244, 0.0353, 0.0208], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 02:46:27,193 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:46:27,925 INFO [optim.py:368] (2/8) 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 02:46:30,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5043, 4.3222, 4.3586, 4.3113, 4.0489, 4.4060, 4.2297, 4.1862], device='cuda:2'), covar=tensor([0.0473, 0.0384, 0.0243, 0.0175, 0.0788, 0.0288, 0.0359, 0.0372], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0172, 0.0219, 0.0186, 0.0257, 0.0207, 0.0158, 0.0221], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:46:35,601 INFO [train.py:904] (2/8) Epoch 4, batch 3600, loss[loss=0.2368, simple_loss=0.3185, pruned_loss=0.07754, over 16748.00 frames. ], tot_loss[loss=0.224, simple_loss=0.2973, pruned_loss=0.07532, over 3306872.11 frames. ], batch size: 57, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:46:39,039 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9387, 4.3199, 3.6218, 2.4726, 3.2779, 2.3673, 4.6266, 4.5682], device='cuda:2'), covar=tensor([0.2032, 0.0619, 0.1049, 0.1351, 0.2207, 0.1288, 0.0338, 0.0464], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0249, 0.0262, 0.0233, 0.0302, 0.0197, 0.0231, 0.0250], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:46:43,508 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 02:47:45,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7210, 3.7687, 1.8011, 3.7049, 2.6681, 3.8279, 1.9326, 2.6466], device='cuda:2'), covar=tensor([0.0094, 0.0284, 0.1541, 0.0138, 0.0695, 0.0403, 0.1302, 0.0616], device='cuda:2'), in_proj_covar=tensor([0.0108, 0.0153, 0.0176, 0.0087, 0.0159, 0.0185, 0.0185, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:47:48,226 INFO [train.py:904] (2/8) Epoch 4, batch 3650, loss[loss=0.2299, simple_loss=0.3099, pruned_loss=0.07497, over 16678.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.2973, pruned_loss=0.07683, over 3290028.63 frames. ], batch size: 57, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:54,694 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:29,174 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:51,845 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8196, 3.8999, 1.8628, 3.9114, 2.7034, 3.9147, 1.9410, 2.8151], device='cuda:2'), covar=tensor([0.0075, 0.0212, 0.1427, 0.0059, 0.0619, 0.0333, 0.1253, 0.0509], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0150, 0.0175, 0.0086, 0.0157, 0.0182, 0.0182, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 02:48:53,739 INFO [optim.py:368] (2/8) 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 02:48:54,233 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:02,642 INFO [train.py:904] (2/8) Epoch 4, batch 3700, loss[loss=0.2001, simple_loss=0.2656, pruned_loss=0.06735, over 16761.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.2954, pruned_loss=0.0781, over 3283183.91 frames. ], batch size: 83, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:49:37,064 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-04-28 02:49:41,423 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:50:11,076 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 02:50:17,552 INFO [train.py:904] (2/8) Epoch 4, batch 3750, loss[loss=0.2394, simple_loss=0.3034, pruned_loss=0.08767, over 11386.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.296, pruned_loss=0.07961, over 3268968.64 frames. ], batch size: 247, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:50:25,287 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:50:48,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2983, 2.3798, 1.7078, 2.2069, 2.9444, 2.7193, 3.3649, 3.0927], device='cuda:2'), covar=tensor([0.0021, 0.0160, 0.0213, 0.0175, 0.0079, 0.0132, 0.0029, 0.0096], device='cuda:2'), in_proj_covar=tensor([0.0077, 0.0139, 0.0137, 0.0134, 0.0130, 0.0138, 0.0107, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-28 02:51:15,996 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3299, 3.2514, 3.7438, 2.6719, 3.5203, 3.8197, 3.6720, 1.8716], device='cuda:2'), covar=tensor([0.0275, 0.0055, 0.0028, 0.0201, 0.0037, 0.0036, 0.0030, 0.0286], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0055, 0.0057, 0.0108, 0.0059, 0.0064, 0.0059, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:51:21,419 INFO [optim.py:368] (2/8) 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 02:51:29,935 INFO [train.py:904] (2/8) Epoch 4, batch 3800, loss[loss=0.2796, simple_loss=0.3407, pruned_loss=0.1092, over 12492.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.2965, pruned_loss=0.08046, over 3275441.66 frames. ], batch size: 247, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:52:36,318 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 4, batch 3850, loss[loss=0.2195, simple_loss=0.2816, pruned_loss=0.07876, over 16670.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.2962, pruned_loss=0.0806, over 3284241.95 frames. ], batch size: 89, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:52:45,045 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6241, 2.1647, 1.5942, 2.0031, 2.5852, 2.4594, 2.7475, 2.7717], device='cuda:2'), covar=tensor([0.0037, 0.0143, 0.0186, 0.0178, 0.0079, 0.0109, 0.0060, 0.0077], device='cuda:2'), in_proj_covar=tensor([0.0076, 0.0138, 0.0137, 0.0133, 0.0130, 0.0138, 0.0104, 0.0115], device='cuda:2'), out_proj_covar=tensor([9.9450e-05, 1.8436e-04, 1.7668e-04, 1.7391e-04, 1.7396e-04, 1.8521e-04, 1.3956e-04, 1.5582e-04], device='cuda:2') 2023-04-28 02:52:54,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8370, 4.7149, 4.6631, 4.1053, 4.6807, 2.1760, 4.4695, 4.6867], device='cuda:2'), covar=tensor([0.0071, 0.0056, 0.0089, 0.0314, 0.0068, 0.1378, 0.0087, 0.0131], device='cuda:2'), in_proj_covar=tensor([0.0089, 0.0079, 0.0118, 0.0128, 0.0088, 0.0128, 0.0104, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:53:30,456 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 02:53:31,029 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:53:49,430 INFO [optim.py:368] (2/8) 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 02:53:57,007 INFO [train.py:904] (2/8) Epoch 4, batch 3900, loss[loss=0.2371, simple_loss=0.3227, pruned_loss=0.07573, over 17097.00 frames. ], tot_loss[loss=0.228, simple_loss=0.2955, pruned_loss=0.08029, over 3290364.09 frames. ], batch size: 48, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:54:40,913 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:55:09,053 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:55:10,261 INFO [train.py:904] (2/8) Epoch 4, batch 3950, loss[loss=0.2453, simple_loss=0.2994, pruned_loss=0.09558, over 16741.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.2951, pruned_loss=0.08036, over 3291322.12 frames. ], batch size: 134, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:56:16,758 INFO [optim.py:368] (2/8) 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 02:56:24,482 INFO [train.py:904] (2/8) Epoch 4, batch 4000, loss[loss=0.2569, simple_loss=0.3323, pruned_loss=0.09075, over 12854.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.295, pruned_loss=0.08075, over 3291044.04 frames. ], batch size: 246, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:36,428 INFO [train.py:904] (2/8) Epoch 4, batch 4050, loss[loss=0.2043, simple_loss=0.2838, pruned_loss=0.06243, over 16666.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.2937, pruned_loss=0.07826, over 3281985.92 frames. ], batch size: 62, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:36,745 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:57:54,809 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2488, 3.9297, 4.0375, 2.5716, 3.8784, 3.9411, 3.9151, 1.9364], device='cuda:2'), covar=tensor([0.0288, 0.0022, 0.0015, 0.0207, 0.0023, 0.0035, 0.0017, 0.0280], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0052, 0.0055, 0.0107, 0.0059, 0.0062, 0.0057, 0.0102], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 02:57:59,804 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9827, 2.2045, 1.6549, 2.0807, 2.6908, 2.4706, 2.8936, 2.9463], device='cuda:2'), covar=tensor([0.0027, 0.0148, 0.0197, 0.0186, 0.0087, 0.0131, 0.0043, 0.0068], device='cuda:2'), in_proj_covar=tensor([0.0074, 0.0136, 0.0137, 0.0134, 0.0128, 0.0138, 0.0103, 0.0115], device='cuda:2'), out_proj_covar=tensor([9.7638e-05, 1.8087e-04, 1.7633e-04, 1.7590e-04, 1.7168e-04, 1.8563e-04, 1.3761e-04, 1.5501e-04], device='cuda:2') 2023-04-28 02:58:08,144 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 02:58:16,986 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7581, 5.0051, 4.6944, 4.7563, 4.4062, 4.3316, 4.4783, 5.0379], device='cuda:2'), covar=tensor([0.0591, 0.0605, 0.0880, 0.0415, 0.0621, 0.0642, 0.0563, 0.0606], device='cuda:2'), in_proj_covar=tensor([0.0325, 0.0441, 0.0367, 0.0277, 0.0285, 0.0279, 0.0349, 0.0307], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:58:39,022 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9131, 3.7416, 3.8703, 4.1133, 4.1761, 3.8000, 4.0963, 4.1861], device='cuda:2'), covar=tensor([0.0651, 0.0596, 0.1144, 0.0416, 0.0429, 0.0963, 0.0544, 0.0368], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0415, 0.0548, 0.0434, 0.0325, 0.0312, 0.0341, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 02:58:41,514 INFO [optim.py:368] (2/8) 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 02:58:48,881 INFO [train.py:904] (2/8) Epoch 4, batch 4100, loss[loss=0.2529, simple_loss=0.3215, pruned_loss=0.09218, over 16651.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.2949, pruned_loss=0.07741, over 3260487.67 frames. ], batch size: 62, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:59:53,943 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:59:56,608 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:06,228 INFO [train.py:904] (2/8) Epoch 4, batch 4150, loss[loss=0.3093, simple_loss=0.3598, pruned_loss=0.1294, over 11155.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3034, pruned_loss=0.08136, over 3223149.02 frames. ], batch size: 247, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:00:38,842 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:44,946 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:01,424 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8106, 2.7324, 2.5848, 1.8846, 2.6192, 2.5564, 2.7188, 1.7335], device='cuda:2'), covar=tensor([0.0247, 0.0026, 0.0026, 0.0192, 0.0031, 0.0049, 0.0024, 0.0243], device='cuda:2'), in_proj_covar=tensor([0.0106, 0.0049, 0.0052, 0.0104, 0.0055, 0.0060, 0.0055, 0.0099], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 03:01:10,642 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:15,081 INFO [optim.py:368] (2/8) 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 03:01:22,988 INFO [train.py:904] (2/8) Epoch 4, batch 4200, loss[loss=0.2493, simple_loss=0.3297, pruned_loss=0.08439, over 16929.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3121, pruned_loss=0.08415, over 3192380.82 frames. ], batch size: 116, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:01:28,611 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:13,233 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:02:19,308 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:39,211 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:40,350 INFO [train.py:904] (2/8) Epoch 4, batch 4250, loss[loss=0.2252, simple_loss=0.3098, pruned_loss=0.07025, over 16397.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.314, pruned_loss=0.08333, over 3191430.55 frames. ], batch size: 35, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:02:53,639 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:02:59,100 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-28 03:03:47,976 INFO [optim.py:368] (2/8) 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 03:03:51,550 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:03:56,213 INFO [train.py:904] (2/8) Epoch 4, batch 4300, loss[loss=0.2843, simple_loss=0.3485, pruned_loss=0.11, over 11521.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3151, pruned_loss=0.08211, over 3192383.67 frames. ], batch size: 248, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:04:27,309 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:04:38,110 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8557, 4.5129, 4.8065, 5.0923, 5.1588, 4.4875, 5.1124, 5.1251], device='cuda:2'), covar=tensor([0.0643, 0.0723, 0.1027, 0.0309, 0.0352, 0.0453, 0.0384, 0.0327], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0396, 0.0522, 0.0412, 0.0310, 0.0301, 0.0323, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:05:10,252 INFO [train.py:904] (2/8) Epoch 4, batch 4350, loss[loss=0.2336, simple_loss=0.3163, pruned_loss=0.07548, over 16233.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3184, pruned_loss=0.08281, over 3205471.37 frames. ], batch size: 165, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:05:10,559 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:05:13,176 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2945, 3.1151, 3.2783, 3.4315, 3.4634, 3.1632, 3.4005, 3.4792], device='cuda:2'), covar=tensor([0.0566, 0.0556, 0.0810, 0.0384, 0.0412, 0.1927, 0.0655, 0.0396], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0398, 0.0525, 0.0414, 0.0310, 0.0304, 0.0325, 0.0337], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:06:15,520 INFO [optim.py:368] (2/8) 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 03:06:20,186 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:22,169 INFO [train.py:904] (2/8) Epoch 4, batch 4400, loss[loss=0.2224, simple_loss=0.3118, pruned_loss=0.06654, over 17129.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3199, pruned_loss=0.08332, over 3214858.28 frames. ], batch size: 49, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:07:32,093 INFO [train.py:904] (2/8) Epoch 4, batch 4450, loss[loss=0.2459, simple_loss=0.3299, pruned_loss=0.08098, over 17263.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3238, pruned_loss=0.08431, over 3205536.20 frames. ], batch size: 52, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:08:04,400 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6087, 4.5930, 4.3856, 4.3026, 4.0226, 4.5315, 4.3364, 4.1554], device='cuda:2'), covar=tensor([0.0274, 0.0129, 0.0146, 0.0136, 0.0626, 0.0137, 0.0240, 0.0342], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0149, 0.0193, 0.0160, 0.0224, 0.0175, 0.0138, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 03:08:36,347 INFO [optim.py:368] (2/8) 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 03:08:36,886 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6833, 3.9706, 3.2307, 2.5155, 2.9715, 2.2938, 4.2479, 4.1910], device='cuda:2'), covar=tensor([0.2140, 0.0524, 0.1024, 0.1098, 0.2054, 0.1275, 0.0283, 0.0361], device='cuda:2'), in_proj_covar=tensor([0.0276, 0.0245, 0.0263, 0.0233, 0.0314, 0.0197, 0.0228, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:08:40,977 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:08:41,339 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 03:08:43,009 INFO [train.py:904] (2/8) Epoch 4, batch 4500, loss[loss=0.2224, simple_loss=0.3068, pruned_loss=0.06905, over 16660.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3241, pruned_loss=0.08451, over 3209515.00 frames. ], batch size: 62, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:09:23,016 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:09:28,650 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:09:41,294 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8361, 4.1584, 3.5311, 2.4511, 3.2186, 2.4427, 4.7001, 4.4289], device='cuda:2'), covar=tensor([0.2128, 0.0522, 0.0939, 0.1190, 0.1935, 0.1204, 0.0274, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0248, 0.0266, 0.0235, 0.0319, 0.0198, 0.0231, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:09:54,754 INFO [train.py:904] (2/8) Epoch 4, batch 4550, loss[loss=0.2376, simple_loss=0.3157, pruned_loss=0.07974, over 17111.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.324, pruned_loss=0.08474, over 3217329.68 frames. ], batch size: 47, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:10:57,619 INFO [optim.py:368] (2/8) 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 03:11:04,052 INFO [train.py:904] (2/8) Epoch 4, batch 4600, loss[loss=0.2141, simple_loss=0.2986, pruned_loss=0.06475, over 16467.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3232, pruned_loss=0.08358, over 3230246.91 frames. ], batch size: 75, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:11:19,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1601, 4.3141, 4.2326, 2.8482, 3.7813, 4.2040, 4.0672, 2.2037], device='cuda:2'), covar=tensor([0.0325, 0.0009, 0.0018, 0.0200, 0.0027, 0.0028, 0.0016, 0.0278], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0050, 0.0054, 0.0108, 0.0057, 0.0062, 0.0057, 0.0101], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 03:11:25,504 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:12:01,877 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:12:15,705 INFO [train.py:904] (2/8) Epoch 4, batch 4650, loss[loss=0.3091, simple_loss=0.3684, pruned_loss=0.1249, over 12168.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3229, pruned_loss=0.08401, over 3213803.55 frames. ], batch size: 248, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:12:40,708 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1354, 3.5788, 3.5404, 1.5067, 3.8358, 3.6969, 3.0170, 2.7315], device='cuda:2'), covar=tensor([0.0883, 0.0116, 0.0150, 0.1338, 0.0047, 0.0052, 0.0344, 0.0473], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0084, 0.0082, 0.0143, 0.0070, 0.0074, 0.0113, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 03:13:20,503 INFO [optim.py:368] (2/8) 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 03:13:28,256 INFO [train.py:904] (2/8) Epoch 4, batch 4700, loss[loss=0.2185, simple_loss=0.3012, pruned_loss=0.0679, over 16755.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3198, pruned_loss=0.08246, over 3210083.59 frames. ], batch size: 124, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:13:32,449 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:13:32,901 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 03:14:06,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6855, 2.7508, 1.7203, 2.8053, 2.1474, 2.7387, 1.9244, 2.3864], device='cuda:2'), covar=tensor([0.0116, 0.0240, 0.1036, 0.0072, 0.0560, 0.0352, 0.0997, 0.0427], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0142, 0.0173, 0.0079, 0.0159, 0.0169, 0.0179, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 03:14:37,258 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8162, 4.0072, 1.8143, 4.3265, 2.5433, 4.1797, 2.1495, 2.7910], device='cuda:2'), covar=tensor([0.0123, 0.0195, 0.1858, 0.0030, 0.0916, 0.0217, 0.1424, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0106, 0.0144, 0.0175, 0.0080, 0.0161, 0.0171, 0.0180, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 03:14:41,628 INFO [train.py:904] (2/8) Epoch 4, batch 4750, loss[loss=0.2449, simple_loss=0.3181, pruned_loss=0.08585, over 15408.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3155, pruned_loss=0.08048, over 3198609.14 frames. ], batch size: 190, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:15:45,876 INFO [optim.py:368] (2/8) 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 03:15:51,953 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:15:53,750 INFO [train.py:904] (2/8) Epoch 4, batch 4800, loss[loss=0.2737, simple_loss=0.3501, pruned_loss=0.09867, over 16251.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3122, pruned_loss=0.07867, over 3213161.23 frames. ], batch size: 165, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:16:33,467 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:16:37,063 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6691, 1.2382, 1.5439, 1.6356, 1.8584, 1.9300, 1.2952, 1.7567], device='cuda:2'), covar=tensor([0.0067, 0.0162, 0.0086, 0.0115, 0.0074, 0.0048, 0.0152, 0.0038], device='cuda:2'), in_proj_covar=tensor([0.0103, 0.0129, 0.0117, 0.0113, 0.0114, 0.0080, 0.0130, 0.0073], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 03:16:38,631 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:16:59,980 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:05,714 INFO [train.py:904] (2/8) Epoch 4, batch 4850, loss[loss=0.224, simple_loss=0.3133, pruned_loss=0.06733, over 16714.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3134, pruned_loss=0.07845, over 3198752.99 frames. ], batch size: 76, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:17:16,496 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-28 03:17:35,197 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-28 03:17:41,753 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:47,234 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 03:17:48,532 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:18:11,908 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35345.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:18:12,515 INFO [optim.py:368] (2/8) 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 03:18:19,089 INFO [train.py:904] (2/8) Epoch 4, batch 4900, loss[loss=0.2136, simple_loss=0.2985, pruned_loss=0.06432, over 16850.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3134, pruned_loss=0.07791, over 3169669.73 frames. ], batch size: 102, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:18:42,122 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:18:53,776 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:19:02,816 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4501, 2.2818, 1.7272, 2.1208, 2.7808, 2.6305, 3.4702, 3.2096], device='cuda:2'), covar=tensor([0.0018, 0.0151, 0.0216, 0.0191, 0.0088, 0.0134, 0.0032, 0.0066], device='cuda:2'), in_proj_covar=tensor([0.0068, 0.0136, 0.0140, 0.0137, 0.0131, 0.0143, 0.0099, 0.0116], device='cuda:2'), out_proj_covar=tensor([8.8579e-05, 1.7994e-04, 1.7954e-04, 1.7829e-04, 1.7576e-04, 1.9013e-04, 1.2905e-04, 1.5557e-04], device='cuda:2') 2023-04-28 03:19:32,910 INFO [train.py:904] (2/8) Epoch 4, batch 4950, loss[loss=0.2478, simple_loss=0.3249, pruned_loss=0.08541, over 17115.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3145, pruned_loss=0.07833, over 3176420.89 frames. ], batch size: 48, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:19:41,843 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:19:51,939 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4612, 2.3139, 2.1506, 4.0438, 1.6711, 3.2941, 2.0668, 2.2346], device='cuda:2'), covar=tensor([0.0549, 0.1553, 0.0876, 0.0328, 0.3148, 0.0712, 0.1767, 0.2088], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0277, 0.0232, 0.0292, 0.0344, 0.0252, 0.0256, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:19:53,434 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:20:22,795 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:37,433 INFO [optim.py:368] (2/8) 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 03:20:41,761 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 4, batch 5000, loss[loss=0.2289, simple_loss=0.3154, pruned_loss=0.07125, over 16691.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3158, pruned_loss=0.07852, over 3182541.56 frames. ], batch size: 134, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:21:57,699 INFO [train.py:904] (2/8) Epoch 4, batch 5050, loss[loss=0.2288, simple_loss=0.2975, pruned_loss=0.08004, over 16405.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3159, pruned_loss=0.07857, over 3187179.97 frames. ], batch size: 35, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:22:33,533 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-28 03:22:39,993 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:22:50,030 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1497, 4.0518, 3.5670, 1.8538, 2.9023, 2.4526, 3.5831, 3.8474], device='cuda:2'), covar=tensor([0.0187, 0.0356, 0.0430, 0.1544, 0.0721, 0.0822, 0.0562, 0.0548], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0122, 0.0151, 0.0141, 0.0135, 0.0127, 0.0141, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 03:22:57,921 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5643, 5.8767, 5.4913, 5.6881, 5.2497, 4.9359, 5.3858, 6.0149], device='cuda:2'), covar=tensor([0.0474, 0.0599, 0.0836, 0.0350, 0.0488, 0.0450, 0.0477, 0.0531], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0419, 0.0363, 0.0273, 0.0270, 0.0273, 0.0339, 0.0295], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:23:03,492 INFO [optim.py:368] (2/8) 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 03:23:10,554 INFO [train.py:904] (2/8) Epoch 4, batch 5100, loss[loss=0.2076, simple_loss=0.3002, pruned_loss=0.05752, over 16722.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3127, pruned_loss=0.07692, over 3193915.37 frames. ], batch size: 76, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:24:08,068 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:24:11,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8106, 1.2357, 1.4827, 1.6809, 1.8552, 1.9350, 1.3540, 1.8320], device='cuda:2'), covar=tensor([0.0066, 0.0160, 0.0086, 0.0103, 0.0082, 0.0045, 0.0154, 0.0030], device='cuda:2'), in_proj_covar=tensor([0.0103, 0.0131, 0.0120, 0.0114, 0.0117, 0.0081, 0.0132, 0.0075], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 03:24:22,834 INFO [train.py:904] (2/8) Epoch 4, batch 5150, loss[loss=0.2419, simple_loss=0.3309, pruned_loss=0.07648, over 16206.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3133, pruned_loss=0.07628, over 3187539.98 frames. ], batch size: 165, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:25:01,953 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6703, 2.4487, 1.8669, 2.3146, 3.0642, 2.9348, 3.5974, 3.3871], device='cuda:2'), covar=tensor([0.0013, 0.0161, 0.0205, 0.0163, 0.0070, 0.0105, 0.0038, 0.0057], device='cuda:2'), in_proj_covar=tensor([0.0067, 0.0137, 0.0138, 0.0137, 0.0129, 0.0144, 0.0098, 0.0116], device='cuda:2'), out_proj_covar=tensor([8.7681e-05, 1.8164e-04, 1.7590e-04, 1.7689e-04, 1.7121e-04, 1.9127e-04, 1.2848e-04, 1.5580e-04], device='cuda:2') 2023-04-28 03:25:29,063 INFO [optim.py:368] (2/8) 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 03:25:36,102 INFO [train.py:904] (2/8) Epoch 4, batch 5200, loss[loss=0.194, simple_loss=0.2723, pruned_loss=0.05788, over 16390.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3106, pruned_loss=0.07481, over 3213224.96 frames. ], batch size: 35, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:25:36,469 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2079, 4.1817, 3.9669, 3.9710, 3.5463, 4.1404, 3.9168, 3.8244], device='cuda:2'), covar=tensor([0.0379, 0.0223, 0.0224, 0.0175, 0.0939, 0.0234, 0.0397, 0.0413], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0159, 0.0201, 0.0165, 0.0230, 0.0188, 0.0144, 0.0206], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 03:26:46,332 INFO [train.py:904] (2/8) Epoch 4, batch 5250, loss[loss=0.2023, simple_loss=0.2921, pruned_loss=0.05626, over 16863.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3082, pruned_loss=0.07462, over 3204109.88 frames. ], batch size: 96, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:26:46,804 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:26:55,508 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6234, 4.5292, 4.3870, 3.7396, 4.4792, 1.7128, 4.2665, 4.4509], device='cuda:2'), covar=tensor([0.0053, 0.0054, 0.0068, 0.0309, 0.0047, 0.1477, 0.0071, 0.0099], device='cuda:2'), in_proj_covar=tensor([0.0081, 0.0071, 0.0106, 0.0119, 0.0081, 0.0126, 0.0094, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:27:13,180 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 03:27:28,540 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:27:52,066 INFO [optim.py:368] (2/8) 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 03:27:55,753 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:00,137 INFO [train.py:904] (2/8) Epoch 4, batch 5300, loss[loss=0.1938, simple_loss=0.2772, pruned_loss=0.05518, over 16570.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3046, pruned_loss=0.07326, over 3205557.66 frames. ], batch size: 75, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:28:06,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4478, 2.4815, 1.8497, 2.2944, 2.9760, 2.7648, 3.5156, 3.2163], device='cuda:2'), covar=tensor([0.0017, 0.0163, 0.0220, 0.0190, 0.0085, 0.0121, 0.0035, 0.0063], device='cuda:2'), in_proj_covar=tensor([0.0068, 0.0142, 0.0141, 0.0140, 0.0132, 0.0147, 0.0101, 0.0118], device='cuda:2'), out_proj_covar=tensor([8.9690e-05, 1.8778e-04, 1.7930e-04, 1.8194e-04, 1.7628e-04, 1.9463e-04, 1.3083e-04, 1.5849e-04], device='cuda:2') 2023-04-28 03:28:17,066 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5720, 3.7001, 1.6820, 3.8487, 2.4772, 3.8382, 1.9062, 2.7804], device='cuda:2'), covar=tensor([0.0101, 0.0196, 0.1573, 0.0036, 0.0721, 0.0249, 0.1304, 0.0582], device='cuda:2'), in_proj_covar=tensor([0.0106, 0.0142, 0.0174, 0.0078, 0.0158, 0.0170, 0.0179, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 03:28:28,840 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:29,136 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 03:29:02,750 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:11,333 INFO [train.py:904] (2/8) Epoch 4, batch 5350, loss[loss=0.2603, simple_loss=0.3407, pruned_loss=0.08997, over 15428.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3033, pruned_loss=0.07211, over 3216837.04 frames. ], batch size: 190, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:29:24,886 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:56,951 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:03,117 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1910, 3.0368, 2.6971, 1.8534, 2.5490, 1.9549, 2.7738, 2.9618], device='cuda:2'), covar=tensor([0.0259, 0.0385, 0.0478, 0.1362, 0.0602, 0.0902, 0.0527, 0.0406], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0122, 0.0153, 0.0142, 0.0135, 0.0128, 0.0140, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 03:30:15,736 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 03:30:17,961 INFO [optim.py:368] (2/8) 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 03:30:23,337 INFO [train.py:904] (2/8) Epoch 4, batch 5400, loss[loss=0.2306, simple_loss=0.3115, pruned_loss=0.07483, over 17060.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3062, pruned_loss=0.07263, over 3231984.98 frames. ], batch size: 55, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:30:53,768 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:57,134 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0976, 3.2219, 3.5232, 3.4728, 3.4631, 3.1876, 3.3025, 3.2986], device='cuda:2'), covar=tensor([0.0261, 0.0411, 0.0309, 0.0403, 0.0380, 0.0334, 0.0566, 0.0385], device='cuda:2'), in_proj_covar=tensor([0.0207, 0.0200, 0.0206, 0.0214, 0.0251, 0.0222, 0.0313, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 03:31:13,806 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:38,572 INFO [train.py:904] (2/8) Epoch 4, batch 5450, loss[loss=0.3104, simple_loss=0.3737, pruned_loss=0.1235, over 15320.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3107, pruned_loss=0.07576, over 3207655.68 frames. ], batch size: 190, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:32:00,956 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1839, 3.6931, 3.7219, 1.6186, 3.9132, 3.8647, 3.0852, 2.6794], device='cuda:2'), covar=tensor([0.0765, 0.0082, 0.0119, 0.1142, 0.0045, 0.0054, 0.0275, 0.0419], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0083, 0.0078, 0.0139, 0.0068, 0.0073, 0.0110, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 03:32:50,256 INFO [optim.py:368] (2/8) 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 03:32:56,433 INFO [train.py:904] (2/8) Epoch 4, batch 5500, loss[loss=0.2849, simple_loss=0.3597, pruned_loss=0.1051, over 16900.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3206, pruned_loss=0.08352, over 3165459.99 frames. ], batch size: 96, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:33:23,726 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6482, 4.6102, 4.4696, 3.7547, 4.4344, 1.6141, 4.2775, 4.4585], device='cuda:2'), covar=tensor([0.0062, 0.0050, 0.0069, 0.0310, 0.0056, 0.1560, 0.0078, 0.0109], device='cuda:2'), in_proj_covar=tensor([0.0080, 0.0071, 0.0107, 0.0120, 0.0081, 0.0127, 0.0095, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:33:44,727 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8175, 4.0847, 3.7951, 3.9110, 3.4941, 3.5547, 3.7617, 4.0088], device='cuda:2'), covar=tensor([0.0564, 0.0614, 0.0782, 0.0428, 0.0569, 0.1180, 0.0516, 0.0777], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0422, 0.0373, 0.0279, 0.0273, 0.0277, 0.0343, 0.0297], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:33:47,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7992, 4.8588, 5.4220, 5.3670, 5.4135, 4.8668, 4.9692, 4.5967], device='cuda:2'), covar=tensor([0.0214, 0.0217, 0.0260, 0.0371, 0.0351, 0.0222, 0.0664, 0.0358], device='cuda:2'), in_proj_covar=tensor([0.0207, 0.0199, 0.0207, 0.0212, 0.0251, 0.0220, 0.0317, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 03:34:16,678 INFO [train.py:904] (2/8) Epoch 4, batch 5550, loss[loss=0.2962, simple_loss=0.3581, pruned_loss=0.1171, over 16283.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3298, pruned_loss=0.09085, over 3142186.60 frames. ], batch size: 165, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:34:17,786 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:35:02,201 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:35:21,245 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5305, 4.2194, 4.1461, 2.9653, 3.8350, 4.0516, 3.8853, 2.5443], device='cuda:2'), covar=tensor([0.0262, 0.0011, 0.0024, 0.0196, 0.0032, 0.0044, 0.0025, 0.0224], device='cuda:2'), in_proj_covar=tensor([0.0114, 0.0050, 0.0055, 0.0112, 0.0058, 0.0064, 0.0059, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 03:35:29,647 INFO [optim.py:368] (2/8) 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 03:35:33,006 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:35:35,569 INFO [train.py:904] (2/8) Epoch 4, batch 5600, loss[loss=0.2692, simple_loss=0.3372, pruned_loss=0.1007, over 16680.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3369, pruned_loss=0.09761, over 3110990.83 frames. ], batch size: 57, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:36:20,418 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:36:33,288 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0577, 1.6746, 1.4121, 1.4426, 1.8782, 1.7154, 1.7031, 1.8861], device='cuda:2'), covar=tensor([0.0023, 0.0104, 0.0147, 0.0144, 0.0073, 0.0102, 0.0055, 0.0070], device='cuda:2'), in_proj_covar=tensor([0.0066, 0.0138, 0.0138, 0.0136, 0.0130, 0.0141, 0.0099, 0.0117], device='cuda:2'), out_proj_covar=tensor([8.5973e-05, 1.8124e-04, 1.7625e-04, 1.7613e-04, 1.7204e-04, 1.8650e-04, 1.2787e-04, 1.5647e-04], device='cuda:2') 2023-04-28 03:36:57,479 INFO [train.py:904] (2/8) Epoch 4, batch 5650, loss[loss=0.3288, simple_loss=0.3908, pruned_loss=0.1334, over 15344.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3442, pruned_loss=0.1041, over 3085580.54 frames. ], batch size: 191, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:37:05,768 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8114, 4.1033, 3.7970, 3.9690, 3.5575, 3.6738, 3.8426, 4.0150], device='cuda:2'), covar=tensor([0.0603, 0.0626, 0.0909, 0.0414, 0.0527, 0.1176, 0.0543, 0.0831], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0415, 0.0369, 0.0274, 0.0269, 0.0275, 0.0336, 0.0296], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:37:41,229 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 03:38:17,750 INFO [train.py:904] (2/8) Epoch 4, batch 5700, loss[loss=0.3247, simple_loss=0.3659, pruned_loss=0.1417, over 11340.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3455, pruned_loss=0.1055, over 3063812.28 frames. ], batch size: 248, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:38:42,662 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:04,974 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:13,399 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:36,187 INFO [train.py:904] (2/8) Epoch 4, batch 5750, loss[loss=0.2745, simple_loss=0.3488, pruned_loss=0.1001, over 16802.00 frames. ], tot_loss[loss=0.2804, simple_loss=0.3484, pruned_loss=0.1063, over 3073466.66 frames. ], batch size: 83, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:40:12,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3862, 1.4091, 1.7925, 1.9167, 2.1996, 2.2379, 1.4566, 2.1088], device='cuda:2'), covar=tensor([0.0046, 0.0185, 0.0096, 0.0109, 0.0074, 0.0061, 0.0160, 0.0039], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0128, 0.0115, 0.0109, 0.0112, 0.0080, 0.0127, 0.0075], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 03:40:29,639 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:42,460 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:49,464 INFO [optim.py:368] (2/8) 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 03:40:55,810 INFO [train.py:904] (2/8) Epoch 4, batch 5800, loss[loss=0.2416, simple_loss=0.3168, pruned_loss=0.08321, over 16718.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3474, pruned_loss=0.1044, over 3065552.93 frames. ], batch size: 76, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:42:12,358 INFO [train.py:904] (2/8) Epoch 4, batch 5850, loss[loss=0.2342, simple_loss=0.3173, pruned_loss=0.07552, over 16397.00 frames. ], tot_loss[loss=0.2738, simple_loss=0.3443, pruned_loss=0.1016, over 3060920.41 frames. ], batch size: 68, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:42:19,037 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8388, 1.5022, 2.1249, 2.4817, 2.6492, 2.8957, 1.7105, 2.7354], device='cuda:2'), covar=tensor([0.0060, 0.0192, 0.0126, 0.0105, 0.0072, 0.0056, 0.0169, 0.0048], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0128, 0.0114, 0.0109, 0.0111, 0.0080, 0.0127, 0.0075], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 03:43:29,068 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 4, batch 5900, loss[loss=0.2633, simple_loss=0.333, pruned_loss=0.09681, over 15470.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3426, pruned_loss=0.1, over 3083506.03 frames. ], batch size: 191, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:44:56,221 INFO [train.py:904] (2/8) Epoch 4, batch 5950, loss[loss=0.2625, simple_loss=0.3407, pruned_loss=0.09212, over 16745.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3441, pruned_loss=0.09941, over 3081427.61 frames. ], batch size: 83, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:45:37,439 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:46:10,128 INFO [optim.py:368] (2/8) 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 03:46:14,579 INFO [train.py:904] (2/8) Epoch 4, batch 6000, loss[loss=0.2764, simple_loss=0.3536, pruned_loss=0.09964, over 16348.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3435, pruned_loss=0.09922, over 3086500.80 frames. ], batch size: 146, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:46:14,580 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 03:46:25,211 INFO [train.py:938] (2/8) Epoch 4, validation: loss=0.1966, simple_loss=0.3069, pruned_loss=0.04309, over 944034.00 frames. 2023-04-28 03:46:25,211 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 03:46:46,808 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-28 03:46:49,506 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:01,320 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:10,890 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 03:47:44,054 INFO [train.py:904] (2/8) Epoch 4, batch 6050, loss[loss=0.2244, simple_loss=0.3073, pruned_loss=0.07078, over 16468.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3407, pruned_loss=0.09761, over 3097306.48 frames. ], batch size: 68, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:48:06,438 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:42,914 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:58,310 INFO [optim.py:368] (2/8) 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 03:49:03,037 INFO [train.py:904] (2/8) Epoch 4, batch 6100, loss[loss=0.2498, simple_loss=0.3283, pruned_loss=0.0857, over 16773.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.34, pruned_loss=0.09662, over 3105592.93 frames. ], batch size: 83, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:49:34,801 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 03:50:03,675 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7136, 2.5194, 2.3011, 4.2185, 1.8700, 3.4481, 2.3971, 2.4378], device='cuda:2'), covar=tensor([0.0509, 0.1320, 0.0819, 0.0257, 0.2722, 0.0594, 0.1438, 0.1886], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0283, 0.0234, 0.0296, 0.0352, 0.0256, 0.0260, 0.0349], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:50:12,544 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9112, 3.4994, 2.7101, 4.9246, 4.4984, 4.2959, 1.8720, 3.2409], device='cuda:2'), covar=tensor([0.1254, 0.0405, 0.1009, 0.0058, 0.0196, 0.0277, 0.1147, 0.0692], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0136, 0.0164, 0.0074, 0.0156, 0.0158, 0.0157, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 03:50:22,075 INFO [train.py:904] (2/8) Epoch 4, batch 6150, loss[loss=0.2825, simple_loss=0.3548, pruned_loss=0.1051, over 16684.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3381, pruned_loss=0.09596, over 3114746.67 frames. ], batch size: 76, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:50:29,629 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:34,989 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:41,780 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:58,763 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 03:51:15,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8419, 2.9643, 2.4788, 4.3662, 4.1238, 4.0969, 1.4212, 3.0594], device='cuda:2'), covar=tensor([0.1197, 0.0448, 0.1095, 0.0057, 0.0176, 0.0239, 0.1293, 0.0643], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0136, 0.0164, 0.0074, 0.0156, 0.0158, 0.0156, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 03:51:24,024 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0645, 2.3804, 1.9740, 1.9321, 2.8515, 2.5598, 3.1878, 3.0768], device='cuda:2'), covar=tensor([0.0027, 0.0157, 0.0220, 0.0218, 0.0087, 0.0148, 0.0059, 0.0083], device='cuda:2'), in_proj_covar=tensor([0.0068, 0.0139, 0.0144, 0.0141, 0.0132, 0.0143, 0.0101, 0.0122], device='cuda:2'), out_proj_covar=tensor([8.8744e-05, 1.8216e-04, 1.8245e-04, 1.8181e-04, 1.7475e-04, 1.8875e-04, 1.2906e-04, 1.6152e-04], device='cuda:2') 2023-04-28 03:51:38,876 INFO [optim.py:368] (2/8) 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 03:51:44,323 INFO [train.py:904] (2/8) Epoch 4, batch 6200, loss[loss=0.2631, simple_loss=0.3417, pruned_loss=0.09223, over 16697.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.3363, pruned_loss=0.09538, over 3101067.08 frames. ], batch size: 89, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:51:57,815 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3792, 1.3530, 1.7684, 2.1776, 2.2427, 2.5531, 1.2741, 2.4290], device='cuda:2'), covar=tensor([0.0060, 0.0213, 0.0148, 0.0113, 0.0103, 0.0062, 0.0222, 0.0045], device='cuda:2'), in_proj_covar=tensor([0.0099, 0.0132, 0.0117, 0.0112, 0.0117, 0.0083, 0.0130, 0.0076], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 03:52:07,207 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:11,561 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:17,429 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:57,808 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:52:59,113 INFO [train.py:904] (2/8) Epoch 4, batch 6250, loss[loss=0.2921, simple_loss=0.3634, pruned_loss=0.1104, over 16799.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3353, pruned_loss=0.09456, over 3116970.65 frames. ], batch size: 124, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:53:36,121 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6820, 2.8687, 2.4575, 4.2098, 3.8292, 3.9208, 1.7189, 2.8569], device='cuda:2'), covar=tensor([0.1274, 0.0443, 0.1012, 0.0063, 0.0220, 0.0277, 0.1178, 0.0689], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0134, 0.0162, 0.0075, 0.0156, 0.0158, 0.0155, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 03:54:09,135 INFO [optim.py:368] (2/8) 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 03:54:14,280 INFO [train.py:904] (2/8) Epoch 4, batch 6300, loss[loss=0.2419, simple_loss=0.3226, pruned_loss=0.08062, over 16780.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3358, pruned_loss=0.0951, over 3096657.82 frames. ], batch size: 124, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:54:31,271 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:55:17,896 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6273, 4.6241, 4.7617, 4.7425, 4.7355, 5.2812, 4.9203, 4.6898], device='cuda:2'), covar=tensor([0.0846, 0.1709, 0.1280, 0.1560, 0.2255, 0.0948, 0.1171, 0.2227], device='cuda:2'), in_proj_covar=tensor([0.0252, 0.0349, 0.0335, 0.0311, 0.0407, 0.0366, 0.0281, 0.0412], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 03:55:32,004 INFO [train.py:904] (2/8) Epoch 4, batch 6350, loss[loss=0.3551, simple_loss=0.3847, pruned_loss=0.1627, over 11426.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.337, pruned_loss=0.09625, over 3104433.18 frames. ], batch size: 248, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:55:38,843 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:55:41,191 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4352, 3.3603, 3.3519, 2.8952, 3.3257, 2.1736, 3.1692, 2.8973], device='cuda:2'), covar=tensor([0.0094, 0.0063, 0.0094, 0.0224, 0.0058, 0.1222, 0.0083, 0.0112], device='cuda:2'), in_proj_covar=tensor([0.0082, 0.0070, 0.0108, 0.0118, 0.0079, 0.0130, 0.0096, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 03:55:47,920 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9086, 3.2847, 2.4890, 4.7574, 4.2647, 4.2328, 1.9493, 3.1371], device='cuda:2'), covar=tensor([0.1297, 0.0453, 0.1174, 0.0067, 0.0225, 0.0297, 0.1205, 0.0728], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0136, 0.0162, 0.0075, 0.0157, 0.0159, 0.0155, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 03:56:28,025 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:43,683 INFO [optim.py:368] (2/8) 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 03:56:48,091 INFO [train.py:904] (2/8) Epoch 4, batch 6400, loss[loss=0.2411, simple_loss=0.311, pruned_loss=0.08563, over 16461.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3361, pruned_loss=0.09623, over 3119953.94 frames. ], batch size: 68, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:57:10,924 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:57:16,214 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6726, 2.6472, 1.7501, 2.7150, 2.1720, 2.7275, 1.9323, 2.3976], device='cuda:2'), covar=tensor([0.0153, 0.0389, 0.1146, 0.0089, 0.0674, 0.0569, 0.1193, 0.0635], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0147, 0.0176, 0.0076, 0.0159, 0.0179, 0.0185, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 03:57:39,964 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:04,368 INFO [train.py:904] (2/8) Epoch 4, batch 6450, loss[loss=0.2352, simple_loss=0.3153, pruned_loss=0.07753, over 16337.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3357, pruned_loss=0.09526, over 3121511.02 frames. ], batch size: 146, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:58:06,861 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8531, 2.0597, 1.6058, 1.8577, 2.5919, 2.3134, 3.0209, 2.8952], device='cuda:2'), covar=tensor([0.0025, 0.0171, 0.0234, 0.0207, 0.0098, 0.0166, 0.0055, 0.0106], device='cuda:2'), in_proj_covar=tensor([0.0068, 0.0137, 0.0141, 0.0138, 0.0131, 0.0143, 0.0101, 0.0119], device='cuda:2'), out_proj_covar=tensor([8.7023e-05, 1.7962e-04, 1.7927e-04, 1.7609e-04, 1.7265e-04, 1.8742e-04, 1.2910e-04, 1.5791e-04], device='cuda:2') 2023-04-28 03:58:20,161 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:29,066 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6433, 2.4179, 2.0940, 3.4791, 2.9177, 3.3768, 1.5298, 2.6111], device='cuda:2'), covar=tensor([0.1612, 0.0697, 0.1382, 0.0123, 0.0417, 0.0401, 0.1646, 0.0893], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0134, 0.0160, 0.0074, 0.0152, 0.0157, 0.0153, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 03:59:12,533 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7375, 3.5242, 3.0397, 1.7396, 2.5957, 2.0676, 3.0615, 3.3383], device='cuda:2'), covar=tensor([0.0285, 0.0423, 0.0506, 0.1663, 0.0781, 0.0944, 0.0654, 0.0594], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0121, 0.0153, 0.0141, 0.0136, 0.0126, 0.0140, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 03:59:15,895 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-28 03:59:18,064 INFO [optim.py:368] (2/8) 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 03:59:23,780 INFO [train.py:904] (2/8) Epoch 4, batch 6500, loss[loss=0.292, simple_loss=0.3717, pruned_loss=0.1061, over 15342.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3329, pruned_loss=0.09388, over 3126994.08 frames. ], batch size: 191, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:59:34,541 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:38,826 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:44,110 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:49,575 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:56,659 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:32,226 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6330, 2.6548, 1.5353, 2.6530, 1.9816, 2.7977, 1.8566, 2.3844], device='cuda:2'), covar=tensor([0.0136, 0.0243, 0.1080, 0.0070, 0.0574, 0.0308, 0.1053, 0.0382], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0146, 0.0175, 0.0077, 0.0160, 0.0176, 0.0185, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 04:00:44,410 INFO [train.py:904] (2/8) Epoch 4, batch 6550, loss[loss=0.2714, simple_loss=0.3495, pruned_loss=0.09662, over 17037.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3362, pruned_loss=0.09553, over 3091457.34 frames. ], batch size: 53, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:00:50,640 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:50,874 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 04:00:55,404 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 04:01:13,041 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:17,264 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1125, 5.3892, 5.0209, 5.0963, 4.7222, 4.6291, 4.8558, 5.4490], device='cuda:2'), covar=tensor([0.0482, 0.0521, 0.0867, 0.0404, 0.0509, 0.0589, 0.0576, 0.0566], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0424, 0.0372, 0.0282, 0.0273, 0.0281, 0.0347, 0.0301], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:01:46,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3707, 4.0612, 3.3138, 4.4552, 4.3693, 4.2322, 4.5088, 4.3667], device='cuda:2'), covar=tensor([0.0784, 0.1013, 0.3428, 0.0878, 0.1129, 0.0898, 0.0974, 0.1261], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0412, 0.0529, 0.0413, 0.0312, 0.0307, 0.0339, 0.0344], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:01:56,564 INFO [optim.py:368] (2/8) 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 04:02:01,479 INFO [train.py:904] (2/8) Epoch 4, batch 6600, loss[loss=0.2447, simple_loss=0.3277, pruned_loss=0.08089, over 16705.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3391, pruned_loss=0.09658, over 3088821.14 frames. ], batch size: 57, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:02:09,068 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:02:22,622 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:59,251 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5890, 3.5303, 2.8479, 2.3007, 2.5068, 2.1289, 3.7224, 3.8579], device='cuda:2'), covar=tensor([0.1926, 0.0614, 0.1129, 0.1322, 0.1974, 0.1360, 0.0320, 0.0419], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0244, 0.0259, 0.0231, 0.0304, 0.0196, 0.0228, 0.0235], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:03:11,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8516, 3.9880, 1.9071, 4.3104, 2.6196, 4.2518, 2.1370, 2.9196], device='cuda:2'), covar=tensor([0.0099, 0.0191, 0.1416, 0.0027, 0.0590, 0.0227, 0.1273, 0.0510], device='cuda:2'), in_proj_covar=tensor([0.0108, 0.0146, 0.0175, 0.0076, 0.0157, 0.0174, 0.0184, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 04:03:19,954 INFO [train.py:904] (2/8) Epoch 4, batch 6650, loss[loss=0.3381, simple_loss=0.3781, pruned_loss=0.1491, over 11739.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3401, pruned_loss=0.09782, over 3099674.73 frames. ], batch size: 247, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:04:17,487 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6994, 4.9235, 4.9933, 5.0138, 4.9317, 5.5367, 5.0472, 4.8327], device='cuda:2'), covar=tensor([0.0690, 0.1213, 0.1126, 0.1356, 0.1816, 0.0654, 0.1044, 0.1917], device='cuda:2'), in_proj_covar=tensor([0.0254, 0.0351, 0.0338, 0.0311, 0.0403, 0.0366, 0.0283, 0.0412], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 04:04:33,021 INFO [optim.py:368] (2/8) 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 04:04:37,775 INFO [train.py:904] (2/8) Epoch 4, batch 6700, loss[loss=0.2585, simple_loss=0.3382, pruned_loss=0.08938, over 16802.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3378, pruned_loss=0.09673, over 3107772.00 frames. ], batch size: 83, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:04:53,114 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:05:45,343 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9870, 2.3230, 1.6948, 1.9875, 2.5862, 2.4003, 3.0977, 2.9008], device='cuda:2'), covar=tensor([0.0025, 0.0153, 0.0220, 0.0175, 0.0084, 0.0148, 0.0042, 0.0082], device='cuda:2'), in_proj_covar=tensor([0.0068, 0.0138, 0.0144, 0.0138, 0.0132, 0.0143, 0.0103, 0.0122], device='cuda:2'), out_proj_covar=tensor([8.7405e-05, 1.7977e-04, 1.8359e-04, 1.7597e-04, 1.7219e-04, 1.8750e-04, 1.3109e-04, 1.6034e-04], device='cuda:2') 2023-04-28 04:05:53,495 INFO [train.py:904] (2/8) Epoch 4, batch 6750, loss[loss=0.2464, simple_loss=0.3169, pruned_loss=0.08798, over 16800.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3371, pruned_loss=0.09698, over 3112674.20 frames. ], batch size: 39, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:06:18,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3539, 2.8718, 2.4774, 2.3013, 2.1805, 1.9350, 2.8469, 3.0136], device='cuda:2'), covar=tensor([0.1780, 0.0641, 0.1051, 0.1201, 0.2118, 0.1457, 0.0469, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0281, 0.0248, 0.0262, 0.0233, 0.0312, 0.0198, 0.0233, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:06:40,932 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1351, 3.9908, 3.9611, 3.2798, 3.9080, 1.5872, 3.7306, 3.8330], device='cuda:2'), covar=tensor([0.0058, 0.0056, 0.0087, 0.0299, 0.0064, 0.1691, 0.0087, 0.0124], device='cuda:2'), in_proj_covar=tensor([0.0082, 0.0071, 0.0109, 0.0119, 0.0080, 0.0132, 0.0096, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:07:01,286 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4859, 3.4162, 2.6698, 2.2088, 2.5340, 1.9241, 3.4993, 3.6666], device='cuda:2'), covar=tensor([0.2017, 0.0644, 0.1199, 0.1356, 0.1689, 0.1452, 0.0449, 0.0507], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0246, 0.0260, 0.0231, 0.0309, 0.0197, 0.0230, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:07:06,103 INFO [optim.py:368] (2/8) 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 04:07:10,550 INFO [train.py:904] (2/8) Epoch 4, batch 6800, loss[loss=0.3024, simple_loss=0.3538, pruned_loss=0.1255, over 11353.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3374, pruned_loss=0.09719, over 3101998.86 frames. ], batch size: 247, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:07:18,694 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8049, 4.6796, 4.5135, 3.8010, 4.5080, 1.7053, 4.3799, 4.4764], device='cuda:2'), covar=tensor([0.0049, 0.0042, 0.0070, 0.0308, 0.0057, 0.1592, 0.0066, 0.0105], device='cuda:2'), in_proj_covar=tensor([0.0083, 0.0071, 0.0109, 0.0119, 0.0080, 0.0132, 0.0096, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:07:27,171 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:31,467 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:35,279 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:37,211 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:37,278 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:27,076 INFO [train.py:904] (2/8) Epoch 4, batch 6850, loss[loss=0.2645, simple_loss=0.3429, pruned_loss=0.09303, over 17216.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3385, pruned_loss=0.09749, over 3096436.57 frames. ], batch size: 44, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:08:38,764 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:46,842 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:49,157 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:08,807 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:36,160 INFO [optim.py:368] (2/8) 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 04:09:39,829 INFO [train.py:904] (2/8) Epoch 4, batch 6900, loss[loss=0.2867, simple_loss=0.3568, pruned_loss=0.1083, over 16898.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3407, pruned_loss=0.09668, over 3110195.20 frames. ], batch size: 109, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:09:43,034 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 04:09:48,108 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 04:09:52,906 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:59,876 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:10:55,353 INFO [train.py:904] (2/8) Epoch 4, batch 6950, loss[loss=0.293, simple_loss=0.3583, pruned_loss=0.1139, over 16739.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3431, pruned_loss=0.09906, over 3111092.42 frames. ], batch size: 57, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:10:57,830 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 04:11:00,664 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:11:32,952 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:11:50,444 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-04-28 04:11:54,259 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8652, 2.9513, 2.7958, 2.0114, 2.5807, 1.9987, 2.5824, 2.8633], device='cuda:2'), covar=tensor([0.0264, 0.0409, 0.0430, 0.1398, 0.0654, 0.0858, 0.0608, 0.0591], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0119, 0.0150, 0.0139, 0.0134, 0.0124, 0.0141, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 04:12:10,356 INFO [optim.py:368] (2/8) 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 04:12:11,785 INFO [train.py:904] (2/8) Epoch 4, batch 7000, loss[loss=0.2355, simple_loss=0.3282, pruned_loss=0.07144, over 16877.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3438, pruned_loss=0.09865, over 3108871.97 frames. ], batch size: 90, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:12:28,099 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 04:13:29,622 INFO [train.py:904] (2/8) Epoch 4, batch 7050, loss[loss=0.2601, simple_loss=0.3374, pruned_loss=0.09144, over 16726.00 frames. ], tot_loss[loss=0.2698, simple_loss=0.3439, pruned_loss=0.09789, over 3109809.48 frames. ], batch size: 124, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:13:43,149 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:13:45,799 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3237, 2.8201, 2.5475, 2.3198, 2.2708, 1.9720, 2.8547, 3.0035], device='cuda:2'), covar=tensor([0.1553, 0.0614, 0.0960, 0.1041, 0.1722, 0.1279, 0.0443, 0.0513], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0245, 0.0261, 0.0232, 0.0305, 0.0195, 0.0231, 0.0236], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:14:45,394 INFO [optim.py:368] (2/8) 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 04:14:46,710 INFO [train.py:904] (2/8) Epoch 4, batch 7100, loss[loss=0.2487, simple_loss=0.3265, pruned_loss=0.08545, over 16789.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3428, pruned_loss=0.09793, over 3101035.22 frames. ], batch size: 83, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:15:13,094 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:02,653 INFO [train.py:904] (2/8) Epoch 4, batch 7150, loss[loss=0.2468, simple_loss=0.3229, pruned_loss=0.08539, over 17058.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3392, pruned_loss=0.09621, over 3118508.06 frames. ], batch size: 55, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:16:03,260 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1598, 3.8339, 3.4830, 1.9775, 2.8312, 2.3927, 3.5243, 3.5953], device='cuda:2'), covar=tensor([0.0292, 0.0618, 0.0487, 0.1686, 0.0846, 0.0952, 0.0690, 0.0768], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0124, 0.0155, 0.0143, 0.0137, 0.0128, 0.0144, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 04:16:23,199 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:24,834 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:38,272 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:17,120 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 4, batch 7200, loss[loss=0.2795, simple_loss=0.3423, pruned_loss=0.1084, over 11896.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3377, pruned_loss=0.09559, over 3076518.24 frames. ], batch size: 250, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:17:33,228 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:35,455 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:18:40,008 INFO [train.py:904] (2/8) Epoch 4, batch 7250, loss[loss=0.2332, simple_loss=0.3088, pruned_loss=0.07884, over 16901.00 frames. ], tot_loss[loss=0.2609, simple_loss=0.3345, pruned_loss=0.09368, over 3086117.73 frames. ], batch size: 90, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:18:52,049 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:10,335 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:12,288 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5722, 1.4612, 2.0202, 2.3700, 2.5058, 2.6454, 1.4794, 2.7505], device='cuda:2'), covar=tensor([0.0057, 0.0208, 0.0131, 0.0101, 0.0097, 0.0071, 0.0208, 0.0041], device='cuda:2'), in_proj_covar=tensor([0.0101, 0.0131, 0.0118, 0.0110, 0.0118, 0.0082, 0.0131, 0.0073], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 04:19:36,100 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3464, 4.2398, 3.9045, 1.9626, 2.9506, 2.8124, 3.4157, 3.8945], device='cuda:2'), covar=tensor([0.0349, 0.0409, 0.0400, 0.1660, 0.0796, 0.0775, 0.0774, 0.0817], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0122, 0.0154, 0.0142, 0.0137, 0.0128, 0.0143, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 04:19:55,472 INFO [optim.py:368] (2/8) 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 04:19:57,416 INFO [train.py:904] (2/8) Epoch 4, batch 7300, loss[loss=0.2358, simple_loss=0.321, pruned_loss=0.07537, over 16571.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.333, pruned_loss=0.09244, over 3097732.62 frames. ], batch size: 62, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:19:57,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8602, 4.0806, 3.8209, 3.8969, 3.6154, 3.6922, 3.8565, 4.0117], device='cuda:2'), covar=tensor([0.0631, 0.0639, 0.0873, 0.0439, 0.0528, 0.1041, 0.0516, 0.0799], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0422, 0.0372, 0.0282, 0.0272, 0.0285, 0.0347, 0.0302], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:20:42,672 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8170, 3.9448, 4.3666, 4.3160, 4.2669, 3.8279, 4.0214, 3.8632], device='cuda:2'), covar=tensor([0.0281, 0.0275, 0.0254, 0.0335, 0.0393, 0.0336, 0.0661, 0.0421], device='cuda:2'), in_proj_covar=tensor([0.0207, 0.0198, 0.0204, 0.0212, 0.0249, 0.0216, 0.0315, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 04:21:14,972 INFO [train.py:904] (2/8) Epoch 4, batch 7350, loss[loss=0.2266, simple_loss=0.3098, pruned_loss=0.07177, over 16408.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.3328, pruned_loss=0.09251, over 3075770.18 frames. ], batch size: 68, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:22:21,597 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 04:22:29,890 INFO [optim.py:368] (2/8) 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 04:22:31,807 INFO [train.py:904] (2/8) Epoch 4, batch 7400, loss[loss=0.2894, simple_loss=0.3697, pruned_loss=0.1046, over 16567.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3343, pruned_loss=0.09358, over 3081276.85 frames. ], batch size: 68, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:23:40,632 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7557, 3.3830, 3.4298, 2.2007, 3.2206, 3.3480, 3.3702, 1.5621], device='cuda:2'), covar=tensor([0.0336, 0.0021, 0.0025, 0.0227, 0.0034, 0.0052, 0.0028, 0.0310], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0051, 0.0056, 0.0110, 0.0058, 0.0064, 0.0059, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 04:23:49,015 INFO [train.py:904] (2/8) Epoch 4, batch 7450, loss[loss=0.2955, simple_loss=0.3744, pruned_loss=0.1083, over 15305.00 frames. ], tot_loss[loss=0.2634, simple_loss=0.3364, pruned_loss=0.09517, over 3081121.45 frames. ], batch size: 191, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:24:26,714 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:06,399 INFO [optim.py:368] (2/8) 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 04:25:07,721 INFO [train.py:904] (2/8) Epoch 4, batch 7500, loss[loss=0.2868, simple_loss=0.3565, pruned_loss=0.1086, over 15239.00 frames. ], tot_loss[loss=0.2648, simple_loss=0.3381, pruned_loss=0.09572, over 3083779.28 frames. ], batch size: 190, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:25:40,562 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:50,222 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:27,336 INFO [train.py:904] (2/8) Epoch 4, batch 7550, loss[loss=0.257, simple_loss=0.3261, pruned_loss=0.09399, over 16190.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.337, pruned_loss=0.09596, over 3080405.74 frames. ], batch size: 165, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:26:54,332 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:07,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5026, 4.4086, 4.3948, 2.9586, 4.0543, 4.3134, 4.1500, 2.3160], device='cuda:2'), covar=tensor([0.0294, 0.0015, 0.0023, 0.0206, 0.0030, 0.0048, 0.0031, 0.0266], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0051, 0.0056, 0.0111, 0.0058, 0.0065, 0.0061, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 04:27:11,973 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8806, 2.8322, 2.2782, 3.7577, 3.4168, 3.6179, 1.5530, 2.5480], device='cuda:2'), covar=tensor([0.1113, 0.0410, 0.1068, 0.0066, 0.0218, 0.0283, 0.1190, 0.0748], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0138, 0.0164, 0.0076, 0.0157, 0.0161, 0.0156, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 04:27:24,024 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:25,848 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1007, 4.0383, 3.9667, 3.2604, 3.9560, 1.5608, 3.8160, 3.8823], device='cuda:2'), covar=tensor([0.0075, 0.0057, 0.0093, 0.0313, 0.0061, 0.1717, 0.0084, 0.0127], device='cuda:2'), in_proj_covar=tensor([0.0082, 0.0070, 0.0107, 0.0116, 0.0080, 0.0131, 0.0095, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:27:38,137 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 4, batch 7600, loss[loss=0.2267, simple_loss=0.3032, pruned_loss=0.07504, over 17253.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.336, pruned_loss=0.09617, over 3088137.20 frames. ], batch size: 52, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:28:05,545 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:28:55,058 INFO [train.py:904] (2/8) Epoch 4, batch 7650, loss[loss=0.2747, simple_loss=0.3496, pruned_loss=0.09988, over 16314.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3372, pruned_loss=0.09749, over 3093652.82 frames. ], batch size: 146, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:30:08,813 INFO [optim.py:368] (2/8) 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 04:30:09,972 INFO [train.py:904] (2/8) Epoch 4, batch 7700, loss[loss=0.2662, simple_loss=0.3348, pruned_loss=0.09878, over 17041.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3366, pruned_loss=0.0972, over 3110229.67 frames. ], batch size: 55, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:30:24,869 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 04:30:38,039 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7355, 3.4743, 2.9277, 1.8599, 2.5472, 2.1590, 3.1665, 3.3389], device='cuda:2'), covar=tensor([0.0244, 0.0428, 0.0566, 0.1493, 0.0749, 0.0938, 0.0525, 0.0544], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0123, 0.0155, 0.0141, 0.0134, 0.0128, 0.0142, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 04:31:12,828 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8766, 4.7414, 4.6658, 4.6052, 4.1455, 4.7059, 4.6737, 4.3472], device='cuda:2'), covar=tensor([0.0401, 0.0303, 0.0182, 0.0128, 0.0822, 0.0324, 0.0227, 0.0398], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0157, 0.0191, 0.0157, 0.0221, 0.0186, 0.0144, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:31:26,736 INFO [train.py:904] (2/8) Epoch 4, batch 7750, loss[loss=0.2778, simple_loss=0.3377, pruned_loss=0.109, over 11727.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3362, pruned_loss=0.09624, over 3105912.72 frames. ], batch size: 248, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:31:40,463 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5032, 4.2523, 4.4358, 4.7580, 4.8307, 4.3118, 4.8406, 4.7661], device='cuda:2'), covar=tensor([0.0936, 0.0702, 0.1264, 0.0414, 0.0407, 0.0614, 0.0403, 0.0409], device='cuda:2'), in_proj_covar=tensor([0.0342, 0.0413, 0.0539, 0.0428, 0.0320, 0.0312, 0.0347, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:32:10,985 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 04:32:26,981 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4638, 2.9159, 2.6168, 2.4406, 2.3593, 2.0628, 2.9569, 3.0645], device='cuda:2'), covar=tensor([0.1465, 0.0515, 0.0885, 0.0967, 0.1549, 0.1208, 0.0323, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0248, 0.0264, 0.0235, 0.0307, 0.0197, 0.0233, 0.0236], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:32:40,400 INFO [optim.py:368] (2/8) 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 04:32:42,173 INFO [train.py:904] (2/8) Epoch 4, batch 7800, loss[loss=0.3715, simple_loss=0.3975, pruned_loss=0.1727, over 11527.00 frames. ], tot_loss[loss=0.2654, simple_loss=0.3369, pruned_loss=0.097, over 3099040.24 frames. ], batch size: 248, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:33:58,979 INFO [train.py:904] (2/8) Epoch 4, batch 7850, loss[loss=0.2817, simple_loss=0.3391, pruned_loss=0.1122, over 11407.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3379, pruned_loss=0.0971, over 3094983.68 frames. ], batch size: 247, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:34:38,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1752, 3.8968, 3.8061, 2.5973, 3.6564, 3.6954, 3.6889, 2.0551], device='cuda:2'), covar=tensor([0.0299, 0.0015, 0.0024, 0.0196, 0.0030, 0.0062, 0.0026, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0049, 0.0057, 0.0110, 0.0058, 0.0066, 0.0060, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 04:34:50,566 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:05,029 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3571, 1.2726, 1.7067, 2.1581, 2.2566, 2.4672, 1.4638, 2.3337], device='cuda:2'), covar=tensor([0.0059, 0.0233, 0.0149, 0.0115, 0.0096, 0.0069, 0.0199, 0.0056], device='cuda:2'), in_proj_covar=tensor([0.0100, 0.0131, 0.0118, 0.0111, 0.0116, 0.0081, 0.0129, 0.0074], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 04:35:12,442 INFO [optim.py:368] (2/8) 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 04:35:14,729 INFO [train.py:904] (2/8) Epoch 4, batch 7900, loss[loss=0.2708, simple_loss=0.3423, pruned_loss=0.09969, over 16607.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.337, pruned_loss=0.09623, over 3110719.07 frames. ], batch size: 57, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:35:28,608 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:36:22,411 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2951, 4.6322, 4.3399, 4.3884, 4.0306, 3.9746, 4.1974, 4.5996], device='cuda:2'), covar=tensor([0.0618, 0.0562, 0.0793, 0.0421, 0.0579, 0.0991, 0.0591, 0.0614], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0430, 0.0375, 0.0278, 0.0276, 0.0293, 0.0353, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:36:34,670 INFO [train.py:904] (2/8) Epoch 4, batch 7950, loss[loss=0.2664, simple_loss=0.3349, pruned_loss=0.09894, over 16355.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3376, pruned_loss=0.09682, over 3114998.65 frames. ], batch size: 35, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:37:05,117 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:37:06,142 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9931, 3.7258, 3.6740, 2.2888, 3.4276, 3.6013, 3.5101, 2.0154], device='cuda:2'), covar=tensor([0.0346, 0.0022, 0.0032, 0.0239, 0.0045, 0.0066, 0.0036, 0.0288], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0049, 0.0056, 0.0110, 0.0058, 0.0066, 0.0060, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 04:37:12,782 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6575, 1.5506, 1.9153, 2.4737, 2.4072, 2.8979, 1.5232, 2.7572], device='cuda:2'), covar=tensor([0.0065, 0.0194, 0.0165, 0.0124, 0.0092, 0.0057, 0.0210, 0.0049], device='cuda:2'), in_proj_covar=tensor([0.0101, 0.0130, 0.0117, 0.0110, 0.0115, 0.0080, 0.0127, 0.0073], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 04:37:30,629 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1111, 3.5926, 3.6053, 1.4846, 3.7370, 3.7642, 2.9175, 2.8259], device='cuda:2'), covar=tensor([0.0857, 0.0097, 0.0143, 0.1326, 0.0070, 0.0055, 0.0315, 0.0407], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0084, 0.0084, 0.0144, 0.0072, 0.0075, 0.0116, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 04:37:49,210 INFO [optim.py:368] (2/8) 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 04:37:50,967 INFO [train.py:904] (2/8) Epoch 4, batch 8000, loss[loss=0.2591, simple_loss=0.3333, pruned_loss=0.09242, over 16876.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.338, pruned_loss=0.09743, over 3113705.96 frames. ], batch size: 116, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:38:41,398 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 04:39:04,864 INFO [train.py:904] (2/8) Epoch 4, batch 8050, loss[loss=0.3394, simple_loss=0.3847, pruned_loss=0.147, over 11582.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3375, pruned_loss=0.09681, over 3108311.46 frames. ], batch size: 246, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:39:37,481 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 04:39:37,554 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-04-28 04:40:21,961 INFO [optim.py:368] (2/8) 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 04:40:23,265 INFO [train.py:904] (2/8) Epoch 4, batch 8100, loss[loss=0.26, simple_loss=0.3281, pruned_loss=0.09599, over 16885.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.337, pruned_loss=0.09607, over 3114268.07 frames. ], batch size: 109, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:41:31,239 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 04:41:41,522 INFO [train.py:904] (2/8) Epoch 4, batch 8150, loss[loss=0.2429, simple_loss=0.3103, pruned_loss=0.08777, over 16896.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3351, pruned_loss=0.09538, over 3104617.91 frames. ], batch size: 96, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:42:02,957 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5635, 4.8178, 4.4747, 4.5689, 4.2147, 4.0857, 4.2841, 4.7909], device='cuda:2'), covar=tensor([0.0595, 0.0608, 0.0818, 0.0425, 0.0591, 0.0970, 0.0558, 0.0680], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0445, 0.0388, 0.0284, 0.0283, 0.0299, 0.0361, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:42:34,014 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:42:40,344 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 04:42:57,102 INFO [optim.py:368] (2/8) 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 04:42:59,074 INFO [train.py:904] (2/8) Epoch 4, batch 8200, loss[loss=0.2372, simple_loss=0.3024, pruned_loss=0.086, over 16430.00 frames. ], tot_loss[loss=0.259, simple_loss=0.3312, pruned_loss=0.0934, over 3119428.33 frames. ], batch size: 35, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:43:00,333 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0905, 3.4266, 3.3350, 2.4902, 3.4060, 3.5326, 3.2005, 1.3989], device='cuda:2'), covar=tensor([0.0356, 0.0035, 0.0055, 0.0223, 0.0047, 0.0068, 0.0082, 0.0433], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0050, 0.0058, 0.0113, 0.0057, 0.0067, 0.0062, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 04:43:45,188 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 04:43:53,410 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:43:56,318 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-28 04:44:23,327 INFO [train.py:904] (2/8) Epoch 4, batch 8250, loss[loss=0.2229, simple_loss=0.3112, pruned_loss=0.06727, over 16701.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3295, pruned_loss=0.09095, over 3095661.25 frames. ], batch size: 62, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:44:49,018 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:45:43,965 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 4, batch 8300, loss[loss=0.2265, simple_loss=0.3142, pruned_loss=0.0694, over 16742.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3261, pruned_loss=0.08735, over 3082972.35 frames. ], batch size: 124, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:46:00,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3305, 2.0190, 2.0896, 3.7515, 1.7152, 3.0828, 2.1344, 1.9288], device='cuda:2'), covar=tensor([0.0469, 0.1661, 0.0882, 0.0266, 0.2778, 0.0621, 0.1677, 0.2266], device='cuda:2'), in_proj_covar=tensor([0.0290, 0.0281, 0.0232, 0.0287, 0.0349, 0.0255, 0.0259, 0.0339], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:47:07,346 INFO [train.py:904] (2/8) Epoch 4, batch 8350, loss[loss=0.2537, simple_loss=0.3163, pruned_loss=0.09555, over 11758.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3243, pruned_loss=0.08455, over 3058857.94 frames. ], batch size: 247, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:47:13,323 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7670, 3.5880, 3.7889, 3.9495, 4.0146, 3.5653, 3.9979, 4.0344], device='cuda:2'), covar=tensor([0.0770, 0.0675, 0.0889, 0.0394, 0.0364, 0.1192, 0.0425, 0.0370], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0407, 0.0519, 0.0417, 0.0310, 0.0299, 0.0335, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:47:44,459 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:48:26,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9859, 2.1047, 2.2542, 3.2432, 2.0168, 2.7706, 2.2822, 1.9703], device='cuda:2'), covar=tensor([0.0444, 0.1564, 0.0748, 0.0320, 0.2516, 0.0693, 0.1492, 0.2310], device='cuda:2'), in_proj_covar=tensor([0.0287, 0.0280, 0.0232, 0.0284, 0.0348, 0.0256, 0.0258, 0.0337], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:48:29,398 INFO [optim.py:368] (2/8) 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 04:48:30,621 INFO [train.py:904] (2/8) Epoch 4, batch 8400, loss[loss=0.2199, simple_loss=0.3051, pruned_loss=0.06734, over 16265.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3207, pruned_loss=0.08148, over 3066195.23 frames. ], batch size: 165, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:49:26,870 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:49:53,141 INFO [train.py:904] (2/8) Epoch 4, batch 8450, loss[loss=0.2316, simple_loss=0.3031, pruned_loss=0.08007, over 12693.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3181, pruned_loss=0.07929, over 3071794.50 frames. ], batch size: 250, lr: 1.56e-02, grad_scale: 4.0 2023-04-28 04:51:13,821 INFO [optim.py:368] (2/8) 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 04:51:13,836 INFO [train.py:904] (2/8) Epoch 4, batch 8500, loss[loss=0.1966, simple_loss=0.2747, pruned_loss=0.05923, over 11927.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3134, pruned_loss=0.07569, over 3064905.10 frames. ], batch size: 248, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:52:39,866 INFO [train.py:904] (2/8) Epoch 4, batch 8550, loss[loss=0.2596, simple_loss=0.3438, pruned_loss=0.08775, over 16633.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3103, pruned_loss=0.07411, over 3033761.60 frames. ], batch size: 134, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:53:09,885 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:54:21,289 INFO [optim.py:368] (2/8) 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 04:54:21,304 INFO [train.py:904] (2/8) Epoch 4, batch 8600, loss[loss=0.2247, simple_loss=0.3102, pruned_loss=0.06961, over 15374.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3112, pruned_loss=0.0734, over 3036276.35 frames. ], batch size: 191, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:54:50,960 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:22,090 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 04:55:58,610 INFO [train.py:904] (2/8) Epoch 4, batch 8650, loss[loss=0.1824, simple_loss=0.2753, pruned_loss=0.04473, over 16179.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3092, pruned_loss=0.07109, over 3048241.24 frames. ], batch size: 165, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:57:07,877 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-04-28 04:57:44,934 INFO [optim.py:368] (2/8) 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 04:57:44,950 INFO [train.py:904] (2/8) Epoch 4, batch 8700, loss[loss=0.181, simple_loss=0.2723, pruned_loss=0.04484, over 16738.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3051, pruned_loss=0.06904, over 3055783.52 frames. ], batch size: 83, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:57:55,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6264, 4.7216, 4.7740, 4.7319, 4.7655, 5.2623, 4.9466, 4.5794], device='cuda:2'), covar=tensor([0.0658, 0.1365, 0.1279, 0.1400, 0.2135, 0.0746, 0.0848, 0.1773], device='cuda:2'), in_proj_covar=tensor([0.0237, 0.0330, 0.0324, 0.0288, 0.0379, 0.0350, 0.0274, 0.0386], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 04:58:36,852 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:49,689 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:59:20,254 INFO [train.py:904] (2/8) Epoch 4, batch 8750, loss[loss=0.2212, simple_loss=0.3097, pruned_loss=0.06631, over 16153.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.305, pruned_loss=0.06861, over 3056098.16 frames. ], batch size: 165, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 04:59:49,907 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6491, 2.7564, 2.4395, 4.0445, 3.6084, 3.9212, 1.4253, 3.0943], device='cuda:2'), covar=tensor([0.1640, 0.0663, 0.1256, 0.0096, 0.0255, 0.0317, 0.1670, 0.0683], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0137, 0.0166, 0.0076, 0.0148, 0.0162, 0.0157, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 05:01:04,982 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:01:14,441 INFO [train.py:904] (2/8) Epoch 4, batch 8800, loss[loss=0.1925, simple_loss=0.2912, pruned_loss=0.04688, over 16844.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3035, pruned_loss=0.06723, over 3083644.73 frames. ], batch size: 90, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:01:15,986 INFO [optim.py:368] (2/8) 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 05:02:23,614 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:02:58,288 INFO [train.py:904] (2/8) Epoch 4, batch 8850, loss[loss=0.225, simple_loss=0.2995, pruned_loss=0.07531, over 12324.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3052, pruned_loss=0.06648, over 3070882.94 frames. ], batch size: 248, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:03:01,235 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 05:04:13,525 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 05:04:32,371 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:44,180 INFO [train.py:904] (2/8) Epoch 4, batch 8900, loss[loss=0.2311, simple_loss=0.3191, pruned_loss=0.07154, over 16539.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3053, pruned_loss=0.06543, over 3080646.98 frames. ], batch size: 62, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:04:49,519 INFO [optim.py:368] (2/8) 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 05:05:10,979 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5058, 2.0286, 1.6531, 1.9148, 2.3950, 2.2472, 2.5586, 2.5698], device='cuda:2'), covar=tensor([0.0024, 0.0175, 0.0222, 0.0216, 0.0105, 0.0155, 0.0068, 0.0098], device='cuda:2'), in_proj_covar=tensor([0.0066, 0.0141, 0.0143, 0.0140, 0.0133, 0.0143, 0.0099, 0.0118], device='cuda:2'), out_proj_covar=tensor([8.3463e-05, 1.8128e-04, 1.7889e-04, 1.7528e-04, 1.7147e-04, 1.8426e-04, 1.2213e-04, 1.5201e-04], device='cuda:2') 2023-04-28 05:06:47,799 INFO [train.py:904] (2/8) Epoch 4, batch 8950, loss[loss=0.2241, simple_loss=0.3001, pruned_loss=0.07405, over 12786.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3056, pruned_loss=0.06621, over 3089351.70 frames. ], batch size: 248, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:08:32,475 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-28 05:08:35,742 INFO [train.py:904] (2/8) Epoch 4, batch 9000, loss[loss=0.1868, simple_loss=0.2683, pruned_loss=0.0527, over 16507.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.302, pruned_loss=0.06438, over 3090532.70 frames. ], batch size: 68, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:08:35,742 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 05:08:45,780 INFO [train.py:938] (2/8) Epoch 4, validation: loss=0.1802, simple_loss=0.283, pruned_loss=0.0387, over 944034.00 frames. 2023-04-28 05:08:45,782 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 05:08:49,857 INFO [optim.py:368] (2/8) 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 05:09:11,600 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:09:43,584 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3178, 3.2179, 3.3251, 3.4944, 3.4870, 3.1854, 3.4688, 3.5314], device='cuda:2'), covar=tensor([0.0719, 0.0600, 0.1010, 0.0461, 0.0532, 0.1972, 0.0658, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0394, 0.0497, 0.0397, 0.0301, 0.0292, 0.0325, 0.0325], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:09:43,602 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:10:29,876 INFO [train.py:904] (2/8) Epoch 4, batch 9050, loss[loss=0.2545, simple_loss=0.3202, pruned_loss=0.09443, over 12421.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3031, pruned_loss=0.06546, over 3082397.92 frames. ], batch size: 246, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:10:30,866 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5800, 4.1485, 4.4157, 1.6127, 4.6351, 4.6728, 3.4886, 3.6489], device='cuda:2'), covar=tensor([0.0743, 0.0096, 0.0093, 0.1267, 0.0028, 0.0031, 0.0234, 0.0299], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0084, 0.0078, 0.0141, 0.0068, 0.0072, 0.0111, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 05:10:58,495 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4675, 3.8852, 4.0355, 1.7274, 4.2520, 4.3010, 3.3304, 3.3173], device='cuda:2'), covar=tensor([0.0694, 0.0087, 0.0106, 0.1113, 0.0033, 0.0028, 0.0219, 0.0297], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0084, 0.0078, 0.0140, 0.0068, 0.0072, 0.0110, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 05:11:06,790 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5316, 3.5221, 4.0270, 4.0036, 4.0165, 3.6584, 3.7557, 3.6663], device='cuda:2'), covar=tensor([0.0283, 0.0471, 0.0324, 0.0378, 0.0365, 0.0304, 0.0693, 0.0358], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0197, 0.0204, 0.0203, 0.0241, 0.0217, 0.0300, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 05:11:17,216 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:21,900 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:43,388 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-04-28 05:11:51,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6589, 4.6221, 4.4365, 4.2886, 4.0288, 4.4964, 4.4662, 4.2096], device='cuda:2'), covar=tensor([0.0423, 0.0227, 0.0203, 0.0169, 0.0692, 0.0249, 0.0223, 0.0392], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0148, 0.0186, 0.0152, 0.0202, 0.0177, 0.0136, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:11:54,713 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:12:03,992 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2277, 4.3756, 4.4454, 4.3958, 4.4189, 4.8937, 4.5539, 4.3014], device='cuda:2'), covar=tensor([0.1084, 0.1382, 0.1266, 0.1698, 0.2161, 0.0959, 0.1274, 0.2345], device='cuda:2'), in_proj_covar=tensor([0.0234, 0.0335, 0.0325, 0.0293, 0.0375, 0.0351, 0.0277, 0.0389], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:12:08,062 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 05:12:14,618 INFO [train.py:904] (2/8) Epoch 4, batch 9100, loss[loss=0.2296, simple_loss=0.3169, pruned_loss=0.07117, over 15499.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3024, pruned_loss=0.06578, over 3086235.61 frames. ], batch size: 192, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:12:18,749 INFO [optim.py:368] (2/8) 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 05:12:22,192 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2231, 4.0181, 4.2653, 4.4949, 4.5932, 4.1391, 4.6306, 4.5779], device='cuda:2'), covar=tensor([0.0766, 0.0671, 0.1075, 0.0452, 0.0389, 0.0570, 0.0371, 0.0333], device='cuda:2'), in_proj_covar=tensor([0.0321, 0.0389, 0.0493, 0.0397, 0.0299, 0.0287, 0.0321, 0.0321], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:14:15,386 INFO [train.py:904] (2/8) Epoch 4, batch 9150, loss[loss=0.1844, simple_loss=0.2753, pruned_loss=0.04677, over 17044.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.3025, pruned_loss=0.06499, over 3088031.43 frames. ], batch size: 97, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:14:37,508 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5616, 1.8163, 1.5551, 1.6716, 2.2949, 2.0626, 2.4085, 2.5071], device='cuda:2'), covar=tensor([0.0018, 0.0185, 0.0208, 0.0208, 0.0086, 0.0159, 0.0061, 0.0077], device='cuda:2'), in_proj_covar=tensor([0.0065, 0.0143, 0.0143, 0.0141, 0.0134, 0.0144, 0.0099, 0.0116], device='cuda:2'), out_proj_covar=tensor([8.0708e-05, 1.8394e-04, 1.7875e-04, 1.7657e-04, 1.7296e-04, 1.8599e-04, 1.2122e-04, 1.4950e-04], device='cuda:2') 2023-04-28 05:15:41,239 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:00,933 INFO [train.py:904] (2/8) Epoch 4, batch 9200, loss[loss=0.2033, simple_loss=0.2892, pruned_loss=0.05871, over 16591.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2974, pruned_loss=0.06363, over 3093744.92 frames. ], batch size: 75, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:16:04,310 INFO [optim.py:368] (2/8) 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 05:16:17,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9304, 1.2791, 1.6657, 1.7158, 1.8674, 1.8561, 1.3777, 1.8684], device='cuda:2'), covar=tensor([0.0082, 0.0150, 0.0085, 0.0100, 0.0095, 0.0060, 0.0155, 0.0051], device='cuda:2'), in_proj_covar=tensor([0.0101, 0.0130, 0.0116, 0.0111, 0.0115, 0.0080, 0.0130, 0.0072], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 05:17:34,404 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:17:35,637 INFO [train.py:904] (2/8) Epoch 4, batch 9250, loss[loss=0.1991, simple_loss=0.2876, pruned_loss=0.05526, over 16826.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2976, pruned_loss=0.06397, over 3091836.87 frames. ], batch size: 124, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:19:05,738 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 05:19:12,719 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6406, 4.9251, 4.6589, 4.7323, 4.3032, 4.3291, 4.4278, 4.9274], device='cuda:2'), covar=tensor([0.0518, 0.0612, 0.0847, 0.0357, 0.0537, 0.0794, 0.0618, 0.0655], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0411, 0.0343, 0.0260, 0.0261, 0.0276, 0.0327, 0.0296], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:19:14,440 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1501, 4.2032, 4.2949, 4.2563, 4.3211, 4.7359, 4.3862, 4.0133], device='cuda:2'), covar=tensor([0.1070, 0.1417, 0.1031, 0.1419, 0.2124, 0.0826, 0.1027, 0.2119], device='cuda:2'), in_proj_covar=tensor([0.0235, 0.0335, 0.0326, 0.0292, 0.0378, 0.0352, 0.0276, 0.0388], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:19:26,128 INFO [train.py:904] (2/8) Epoch 4, batch 9300, loss[loss=0.2013, simple_loss=0.2762, pruned_loss=0.06314, over 12403.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2953, pruned_loss=0.0629, over 3074180.56 frames. ], batch size: 246, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:19:30,018 INFO [optim.py:368] (2/8) 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 05:19:42,764 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-28 05:19:48,962 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 05:20:06,097 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:11,784 INFO [train.py:904] (2/8) Epoch 4, batch 9350, loss[loss=0.2134, simple_loss=0.2959, pruned_loss=0.0655, over 16922.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2948, pruned_loss=0.0626, over 3079088.00 frames. ], batch size: 116, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:21:41,272 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:46,605 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7372, 3.7825, 3.8944, 3.8066, 3.8446, 4.2417, 3.9671, 3.7084], device='cuda:2'), covar=tensor([0.2058, 0.2015, 0.1415, 0.2113, 0.2588, 0.1352, 0.1240, 0.2439], device='cuda:2'), in_proj_covar=tensor([0.0234, 0.0333, 0.0324, 0.0291, 0.0378, 0.0349, 0.0276, 0.0382], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:21:48,576 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:05,667 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:30,159 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:51,014 INFO [train.py:904] (2/8) Epoch 4, batch 9400, loss[loss=0.1897, simple_loss=0.2686, pruned_loss=0.05538, over 12422.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2954, pruned_loss=0.06263, over 3063157.56 frames. ], batch size: 247, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:22:57,589 INFO [optim.py:368] (2/8) 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 05:23:18,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8395, 3.1208, 2.4719, 4.4835, 4.0487, 4.2380, 1.3744, 3.0125], device='cuda:2'), covar=tensor([0.1180, 0.0470, 0.1078, 0.0053, 0.0149, 0.0232, 0.1339, 0.0632], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0135, 0.0164, 0.0075, 0.0141, 0.0159, 0.0158, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 05:23:31,076 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-04-28 05:23:41,189 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:09,631 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:32,546 INFO [train.py:904] (2/8) Epoch 4, batch 9450, loss[loss=0.1926, simple_loss=0.2794, pruned_loss=0.05289, over 16408.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2972, pruned_loss=0.06296, over 3064618.24 frames. ], batch size: 68, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:25:54,139 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:26:06,004 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2790, 4.1629, 4.1174, 3.5648, 4.0911, 1.4794, 3.8717, 4.0229], device='cuda:2'), covar=tensor([0.0052, 0.0053, 0.0073, 0.0207, 0.0053, 0.1651, 0.0088, 0.0128], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0067, 0.0104, 0.0103, 0.0077, 0.0129, 0.0092, 0.0099], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:26:13,854 INFO [train.py:904] (2/8) Epoch 4, batch 9500, loss[loss=0.1705, simple_loss=0.252, pruned_loss=0.0445, over 12857.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2956, pruned_loss=0.06229, over 3047088.54 frames. ], batch size: 247, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:26:21,218 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:27:48,798 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:27:59,647 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-28 05:28:03,838 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0261, 2.7706, 2.6871, 1.7235, 2.9127, 2.9012, 2.5879, 2.3855], device='cuda:2'), covar=tensor([0.0622, 0.0118, 0.0123, 0.0905, 0.0068, 0.0083, 0.0266, 0.0362], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0081, 0.0074, 0.0137, 0.0067, 0.0072, 0.0108, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 05:28:04,403 INFO [train.py:904] (2/8) Epoch 4, batch 9550, loss[loss=0.2227, simple_loss=0.2982, pruned_loss=0.07361, over 12462.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2952, pruned_loss=0.0624, over 3066525.69 frames. ], batch size: 247, lr: 1.53e-02, grad_scale: 2.0 2023-04-28 05:29:46,555 INFO [train.py:904] (2/8) Epoch 4, batch 9600, loss[loss=0.2308, simple_loss=0.2966, pruned_loss=0.08245, over 12376.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2973, pruned_loss=0.0639, over 3048751.28 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:29:52,055 INFO [optim.py:368] (2/8) 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 05:31:32,617 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 05:31:33,068 INFO [train.py:904] (2/8) Epoch 4, batch 9650, loss[loss=0.1947, simple_loss=0.2831, pruned_loss=0.05312, over 16532.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.3001, pruned_loss=0.06449, over 3063835.76 frames. ], batch size: 62, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:32:17,612 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:25,197 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:33:21,181 INFO [train.py:904] (2/8) Epoch 4, batch 9700, loss[loss=0.2118, simple_loss=0.2915, pruned_loss=0.06604, over 16711.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2988, pruned_loss=0.06424, over 3053612.03 frames. ], batch size: 134, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:33:26,546 INFO [optim.py:368] (2/8) 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 05:33:27,497 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9548, 5.2493, 5.0756, 5.0032, 4.5903, 4.5541, 4.6683, 5.3241], device='cuda:2'), covar=tensor([0.0527, 0.0673, 0.0766, 0.0433, 0.0553, 0.0650, 0.0605, 0.0601], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0421, 0.0355, 0.0269, 0.0271, 0.0279, 0.0335, 0.0302], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:33:41,579 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9017, 2.0317, 2.2196, 3.1637, 1.9961, 2.6647, 2.2738, 1.8936], device='cuda:2'), covar=tensor([0.0509, 0.1663, 0.0809, 0.0317, 0.2499, 0.0746, 0.1397, 0.2253], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0281, 0.0230, 0.0283, 0.0341, 0.0256, 0.0254, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:33:53,099 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:01,385 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:54,816 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:35:03,565 INFO [train.py:904] (2/8) Epoch 4, batch 9750, loss[loss=0.1902, simple_loss=0.2811, pruned_loss=0.04962, over 16540.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2967, pruned_loss=0.06334, over 3043596.55 frames. ], batch size: 68, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:35:20,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8818, 4.6259, 4.7451, 5.0484, 5.1500, 4.4959, 5.1439, 5.1132], device='cuda:2'), covar=tensor([0.0844, 0.0586, 0.1069, 0.0401, 0.0445, 0.0508, 0.0382, 0.0340], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0401, 0.0502, 0.0412, 0.0304, 0.0295, 0.0330, 0.0336], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:35:36,912 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4130, 2.0275, 2.0091, 3.8449, 1.7564, 2.9819, 2.2424, 1.9293], device='cuda:2'), covar=tensor([0.0451, 0.1677, 0.0875, 0.0256, 0.2767, 0.0764, 0.1479, 0.2333], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0282, 0.0231, 0.0284, 0.0341, 0.0257, 0.0254, 0.0327], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:35:38,839 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:36:03,509 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4003, 4.3656, 4.0991, 3.7157, 4.2095, 1.5040, 4.0037, 4.0895], device='cuda:2'), covar=tensor([0.0049, 0.0038, 0.0077, 0.0185, 0.0050, 0.1666, 0.0072, 0.0108], device='cuda:2'), in_proj_covar=tensor([0.0078, 0.0065, 0.0102, 0.0101, 0.0076, 0.0130, 0.0091, 0.0097], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:36:45,109 INFO [train.py:904] (2/8) Epoch 4, batch 9800, loss[loss=0.2112, simple_loss=0.3098, pruned_loss=0.05628, over 16674.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2972, pruned_loss=0.06213, over 3063463.87 frames. ], batch size: 134, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:36:51,067 INFO [optim.py:368] (2/8) 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 05:36:56,635 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:24,847 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:37,805 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:17,359 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:38:29,681 INFO [train.py:904] (2/8) Epoch 4, batch 9850, loss[loss=0.2185, simple_loss=0.2998, pruned_loss=0.06864, over 15304.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2976, pruned_loss=0.06117, over 3085792.70 frames. ], batch size: 190, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:38:57,242 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 05:39:09,978 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1936, 4.2501, 4.3482, 4.3441, 4.4040, 4.7510, 4.3722, 4.0968], device='cuda:2'), covar=tensor([0.1143, 0.1396, 0.1111, 0.1617, 0.1861, 0.0804, 0.1018, 0.2006], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0338, 0.0327, 0.0294, 0.0389, 0.0355, 0.0274, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:39:30,963 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:39,127 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:05,815 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:40:21,695 INFO [train.py:904] (2/8) Epoch 4, batch 9900, loss[loss=0.2305, simple_loss=0.3275, pruned_loss=0.06676, over 16958.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2985, pruned_loss=0.06137, over 3068379.67 frames. ], batch size: 125, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:40:27,910 INFO [optim.py:368] (2/8) 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 05:41:56,787 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:42:18,080 INFO [train.py:904] (2/8) Epoch 4, batch 9950, loss[loss=0.2204, simple_loss=0.3046, pruned_loss=0.06808, over 16305.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.3006, pruned_loss=0.06237, over 3058889.19 frames. ], batch size: 146, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:42:50,109 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7572, 3.9119, 3.2659, 2.4882, 2.8585, 2.2908, 4.1163, 3.9815], device='cuda:2'), covar=tensor([0.1838, 0.0531, 0.1017, 0.1368, 0.1381, 0.1230, 0.0305, 0.0418], device='cuda:2'), in_proj_covar=tensor([0.0262, 0.0237, 0.0252, 0.0228, 0.0235, 0.0189, 0.0221, 0.0220], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:43:13,909 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:20,283 INFO [train.py:904] (2/8) Epoch 4, batch 10000, loss[loss=0.1979, simple_loss=0.2928, pruned_loss=0.05155, over 15562.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2989, pruned_loss=0.06187, over 3069602.82 frames. ], batch size: 193, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:44:26,639 INFO [optim.py:368] (2/8) 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 05:44:55,733 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1621, 5.5159, 5.2508, 5.3652, 4.9508, 4.7634, 4.9697, 5.5211], device='cuda:2'), covar=tensor([0.0697, 0.0659, 0.0722, 0.0335, 0.0522, 0.0543, 0.0503, 0.0702], device='cuda:2'), in_proj_covar=tensor([0.0310, 0.0421, 0.0352, 0.0271, 0.0271, 0.0281, 0.0339, 0.0304], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:45:01,305 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:01,396 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:41,109 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 05:45:46,844 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:46:02,920 INFO [train.py:904] (2/8) Epoch 4, batch 10050, loss[loss=0.2113, simple_loss=0.2987, pruned_loss=0.06196, over 12068.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.299, pruned_loss=0.06183, over 3072762.50 frames. ], batch size: 247, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:46:32,919 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7493, 4.0769, 1.8973, 4.3588, 2.7001, 4.2315, 2.2077, 3.0995], device='cuda:2'), covar=tensor([0.0095, 0.0152, 0.1449, 0.0030, 0.0709, 0.0402, 0.1206, 0.0509], device='cuda:2'), in_proj_covar=tensor([0.0108, 0.0134, 0.0170, 0.0077, 0.0153, 0.0162, 0.0179, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 05:46:38,831 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:38,803 INFO [train.py:904] (2/8) Epoch 4, batch 10100, loss[loss=0.185, simple_loss=0.2727, pruned_loss=0.0487, over 16868.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2992, pruned_loss=0.06212, over 3055124.44 frames. ], batch size: 102, lr: 1.52e-02, grad_scale: 8.0 2023-04-28 05:47:39,293 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,162 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,813 INFO [optim.py:368] (2/8) 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 05:48:27,648 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:49:25,172 INFO [train.py:904] (2/8) Epoch 5, batch 0, loss[loss=0.3684, simple_loss=0.3898, pruned_loss=0.1735, over 16502.00 frames. ], tot_loss[loss=0.3684, simple_loss=0.3898, pruned_loss=0.1735, over 16502.00 frames. ], batch size: 146, lr: 1.42e-02, grad_scale: 8.0 2023-04-28 05:49:25,172 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 05:49:32,545 INFO [train.py:938] (2/8) Epoch 5, validation: loss=0.1789, simple_loss=0.2817, pruned_loss=0.03802, over 944034.00 frames. 2023-04-28 05:49:32,546 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 05:49:38,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7725, 4.5437, 4.1993, 2.1427, 3.2830, 2.7570, 3.9431, 4.1252], device='cuda:2'), covar=tensor([0.0208, 0.0437, 0.0372, 0.1402, 0.0589, 0.0819, 0.0568, 0.0729], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0111, 0.0154, 0.0142, 0.0132, 0.0126, 0.0136, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 05:49:53,807 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 05:50:11,886 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:42,808 INFO [train.py:904] (2/8) Epoch 5, batch 50, loss[loss=0.2234, simple_loss=0.3084, pruned_loss=0.06922, over 17112.00 frames. ], tot_loss[loss=0.2601, simple_loss=0.326, pruned_loss=0.09708, over 757209.96 frames. ], batch size: 55, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:50:43,171 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:49,836 INFO [optim.py:368] (2/8) 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 05:50:52,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8471, 2.3218, 2.4015, 4.1369, 1.9390, 3.0292, 2.1640, 2.2013], device='cuda:2'), covar=tensor([0.0490, 0.1824, 0.0887, 0.0316, 0.3018, 0.0858, 0.1856, 0.2323], device='cuda:2'), in_proj_covar=tensor([0.0291, 0.0290, 0.0234, 0.0292, 0.0348, 0.0260, 0.0260, 0.0338], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:51:29,204 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:51:50,913 INFO [train.py:904] (2/8) Epoch 5, batch 100, loss[loss=0.2902, simple_loss=0.3561, pruned_loss=0.1122, over 12239.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3167, pruned_loss=0.08838, over 1320131.67 frames. ], batch size: 247, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:52:07,450 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:52:59,302 INFO [train.py:904] (2/8) Epoch 5, batch 150, loss[loss=0.2423, simple_loss=0.3177, pruned_loss=0.08346, over 16727.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3128, pruned_loss=0.0853, over 1772037.93 frames. ], batch size: 62, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:53:08,165 INFO [optim.py:368] (2/8) 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 05:53:14,293 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 05:54:09,222 INFO [train.py:904] (2/8) Epoch 5, batch 200, loss[loss=0.2478, simple_loss=0.3148, pruned_loss=0.09036, over 16470.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3112, pruned_loss=0.0837, over 2120456.92 frames. ], batch size: 68, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:54:52,248 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0291, 5.3759, 5.0674, 5.1437, 4.7693, 4.6158, 4.9045, 5.4254], device='cuda:2'), covar=tensor([0.0738, 0.0672, 0.0910, 0.0477, 0.0765, 0.0728, 0.0678, 0.0771], device='cuda:2'), in_proj_covar=tensor([0.0342, 0.0459, 0.0396, 0.0302, 0.0298, 0.0306, 0.0372, 0.0329], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 05:55:13,917 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:17,643 INFO [train.py:904] (2/8) Epoch 5, batch 250, loss[loss=0.2295, simple_loss=0.3168, pruned_loss=0.07109, over 17037.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3089, pruned_loss=0.08318, over 2388839.54 frames. ], batch size: 53, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:55:18,005 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:25,559 INFO [optim.py:368] (2/8) 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 05:55:37,342 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:50,391 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:51,755 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2912, 4.5818, 2.4407, 4.9621, 3.1536, 4.8298, 2.5777, 3.3777], device='cuda:2'), covar=tensor([0.0096, 0.0194, 0.1349, 0.0027, 0.0666, 0.0267, 0.1191, 0.0513], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0147, 0.0176, 0.0082, 0.0159, 0.0176, 0.0185, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 05:56:24,970 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:26,414 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:27,683 INFO [train.py:904] (2/8) Epoch 5, batch 300, loss[loss=0.2041, simple_loss=0.2883, pruned_loss=0.05994, over 17204.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3055, pruned_loss=0.08086, over 2589656.78 frames. ], batch size: 44, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:56:58,128 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:04,452 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:07,258 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:39,677 INFO [train.py:904] (2/8) Epoch 5, batch 350, loss[loss=0.2385, simple_loss=0.2937, pruned_loss=0.09166, over 15403.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.301, pruned_loss=0.07877, over 2757520.20 frames. ], batch size: 190, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:57:48,102 INFO [optim.py:368] (2/8) 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 05:57:54,074 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:58:14,880 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:22,909 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4639, 4.9187, 5.3892, 5.2963, 5.2853, 4.8357, 4.4070, 4.5688], device='cuda:2'), covar=tensor([0.0549, 0.0570, 0.0476, 0.0837, 0.0834, 0.0576, 0.1399, 0.0479], device='cuda:2'), in_proj_covar=tensor([0.0228, 0.0220, 0.0224, 0.0230, 0.0266, 0.0238, 0.0338, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 05:58:28,023 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:44,155 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:48,861 INFO [train.py:904] (2/8) Epoch 5, batch 400, loss[loss=0.206, simple_loss=0.2874, pruned_loss=0.06229, over 17035.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.2988, pruned_loss=0.07818, over 2885023.20 frames. ], batch size: 50, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 05:58:59,074 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:24,238 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8077, 4.8992, 5.1092, 5.0537, 4.9615, 5.5214, 5.2417, 4.9498], device='cuda:2'), covar=tensor([0.0871, 0.1882, 0.1342, 0.1972, 0.2860, 0.1105, 0.1134, 0.2487], device='cuda:2'), in_proj_covar=tensor([0.0263, 0.0393, 0.0371, 0.0337, 0.0450, 0.0401, 0.0306, 0.0446], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 05:59:36,590 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:51,091 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:00,530 INFO [train.py:904] (2/8) Epoch 5, batch 450, loss[loss=0.1777, simple_loss=0.2627, pruned_loss=0.04637, over 16981.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.2963, pruned_loss=0.07658, over 2983344.67 frames. ], batch size: 41, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:00:09,409 INFO [optim.py:368] (2/8) 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 06:00:11,806 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:10,606 INFO [train.py:904] (2/8) Epoch 5, batch 500, loss[loss=0.2443, simple_loss=0.3292, pruned_loss=0.07971, over 16747.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2946, pruned_loss=0.07493, over 3042795.68 frames. ], batch size: 57, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:01:12,246 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:16,704 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:45,804 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6526, 4.8083, 5.2896, 5.3307, 5.3348, 4.8472, 4.8531, 4.5204], device='cuda:2'), covar=tensor([0.0204, 0.0280, 0.0334, 0.0376, 0.0294, 0.0279, 0.0612, 0.0353], device='cuda:2'), in_proj_covar=tensor([0.0228, 0.0222, 0.0230, 0.0233, 0.0267, 0.0240, 0.0341, 0.0206], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 06:02:14,529 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:02:19,017 INFO [train.py:904] (2/8) Epoch 5, batch 550, loss[loss=0.192, simple_loss=0.2652, pruned_loss=0.05944, over 16781.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2942, pruned_loss=0.074, over 3110939.11 frames. ], batch size: 39, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:02:27,246 INFO [optim.py:368] (2/8) 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 06:02:35,961 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:21,440 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:27,688 INFO [train.py:904] (2/8) Epoch 5, batch 600, loss[loss=0.207, simple_loss=0.2851, pruned_loss=0.06443, over 17134.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2932, pruned_loss=0.07439, over 3161035.52 frames. ], batch size: 46, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:03:54,837 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:04:33,941 INFO [train.py:904] (2/8) Epoch 5, batch 650, loss[loss=0.2102, simple_loss=0.2777, pruned_loss=0.07135, over 12600.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2905, pruned_loss=0.07289, over 3197785.36 frames. ], batch size: 249, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:04:40,902 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:04:42,445 INFO [optim.py:368] (2/8) 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 06:05:39,909 INFO [train.py:904] (2/8) Epoch 5, batch 700, loss[loss=0.1882, simple_loss=0.2753, pruned_loss=0.05055, over 17173.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2896, pruned_loss=0.07216, over 3235776.51 frames. ], batch size: 46, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:05:48,989 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:11,546 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 06:06:49,378 INFO [train.py:904] (2/8) Epoch 5, batch 750, loss[loss=0.2077, simple_loss=0.2965, pruned_loss=0.05945, over 17267.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2901, pruned_loss=0.07152, over 3262409.14 frames. ], batch size: 52, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:06:52,746 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) 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 06:07:58,380 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 06:07:58,822 INFO [train.py:904] (2/8) Epoch 5, batch 800, loss[loss=0.1852, simple_loss=0.2744, pruned_loss=0.04799, over 17113.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2901, pruned_loss=0.07142, over 3268735.61 frames. ], batch size: 47, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:07:59,097 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:08,623 INFO [train.py:904] (2/8) Epoch 5, batch 850, loss[loss=0.201, simple_loss=0.2718, pruned_loss=0.06507, over 16829.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2894, pruned_loss=0.07108, over 3273950.25 frames. ], batch size: 83, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:09:16,401 INFO [optim.py:368] (2/8) 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 06:09:17,866 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:40,941 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0772, 3.0412, 3.5361, 2.2867, 3.2806, 3.5724, 3.4224, 1.9792], device='cuda:2'), covar=tensor([0.0276, 0.0101, 0.0024, 0.0216, 0.0042, 0.0032, 0.0035, 0.0242], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0059, 0.0060, 0.0113, 0.0059, 0.0066, 0.0062, 0.0103], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 06:10:16,013 INFO [train.py:904] (2/8) Epoch 5, batch 900, loss[loss=0.2482, simple_loss=0.3135, pruned_loss=0.09147, over 15600.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2885, pruned_loss=0.07099, over 3280827.01 frames. ], batch size: 191, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:10:45,064 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:22,430 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9151, 4.2629, 1.8096, 4.5519, 2.6765, 4.5032, 2.0954, 2.9605], device='cuda:2'), covar=tensor([0.0114, 0.0192, 0.1563, 0.0036, 0.0801, 0.0208, 0.1333, 0.0579], device='cuda:2'), in_proj_covar=tensor([0.0116, 0.0154, 0.0177, 0.0086, 0.0160, 0.0183, 0.0186, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 06:11:27,561 INFO [train.py:904] (2/8) Epoch 5, batch 950, loss[loss=0.2088, simple_loss=0.2777, pruned_loss=0.06995, over 16883.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2888, pruned_loss=0.07148, over 3285109.99 frames. ], batch size: 96, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:11:34,464 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:11:35,282 INFO [optim.py:368] (2/8) 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 06:11:53,683 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:04,697 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9602, 4.2572, 3.3590, 2.4621, 3.1493, 2.3636, 4.6086, 4.4353], device='cuda:2'), covar=tensor([0.2007, 0.0546, 0.1087, 0.1413, 0.2141, 0.1471, 0.0284, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0253, 0.0268, 0.0239, 0.0294, 0.0202, 0.0237, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:12:37,216 INFO [train.py:904] (2/8) Epoch 5, batch 1000, loss[loss=0.1805, simple_loss=0.2619, pruned_loss=0.04956, over 16866.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2872, pruned_loss=0.07099, over 3294296.73 frames. ], batch size: 42, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:12:41,514 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:27,154 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 06:13:45,696 INFO [train.py:904] (2/8) Epoch 5, batch 1050, loss[loss=0.2625, simple_loss=0.3142, pruned_loss=0.1053, over 16725.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2863, pruned_loss=0.07046, over 3300007.71 frames. ], batch size: 124, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:13:49,039 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:50,359 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:54,767 INFO [optim.py:368] (2/8) 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 06:14:00,468 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2085, 3.3957, 1.6549, 3.3037, 2.4970, 3.4316, 1.6955, 2.6243], device='cuda:2'), covar=tensor([0.0118, 0.0203, 0.1476, 0.0135, 0.0626, 0.0420, 0.1398, 0.0542], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0151, 0.0174, 0.0086, 0.0159, 0.0184, 0.0184, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 06:14:56,088 INFO [train.py:904] (2/8) Epoch 5, batch 1100, loss[loss=0.1867, simple_loss=0.2636, pruned_loss=0.05493, over 16676.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2852, pruned_loss=0.07009, over 3303870.02 frames. ], batch size: 76, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:14:56,429 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:56,492 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:15:01,120 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-28 06:15:15,784 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:15:36,006 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 06:16:02,694 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:05,298 INFO [train.py:904] (2/8) Epoch 5, batch 1150, loss[loss=0.2348, simple_loss=0.3004, pruned_loss=0.08462, over 16875.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2847, pruned_loss=0.06876, over 3319092.81 frames. ], batch size: 109, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:16:12,831 INFO [optim.py:368] (2/8) 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 06:16:14,941 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:59,603 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3653, 3.9542, 3.4165, 1.9981, 2.8520, 2.4103, 3.6386, 3.7386], device='cuda:2'), covar=tensor([0.0211, 0.0497, 0.0495, 0.1472, 0.0683, 0.0891, 0.0455, 0.0692], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0133, 0.0156, 0.0144, 0.0136, 0.0127, 0.0143, 0.0139], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 06:17:14,374 INFO [train.py:904] (2/8) Epoch 5, batch 1200, loss[loss=0.2101, simple_loss=0.2796, pruned_loss=0.07028, over 17025.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.284, pruned_loss=0.06843, over 3311497.66 frames. ], batch size: 41, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:17:21,141 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:18:23,576 INFO [train.py:904] (2/8) Epoch 5, batch 1250, loss[loss=0.1892, simple_loss=0.2748, pruned_loss=0.05176, over 16835.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2852, pruned_loss=0.0685, over 3310679.95 frames. ], batch size: 42, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:18:31,523 INFO [optim.py:368] (2/8) 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 06:18:40,070 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 06:18:40,320 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 06:19:30,797 INFO [train.py:904] (2/8) Epoch 5, batch 1300, loss[loss=0.2106, simple_loss=0.296, pruned_loss=0.06261, over 17188.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2849, pruned_loss=0.06842, over 3313915.87 frames. ], batch size: 46, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:19:38,948 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:20:38,008 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:20:42,275 INFO [train.py:904] (2/8) Epoch 5, batch 1350, loss[loss=0.2337, simple_loss=0.3065, pruned_loss=0.0805, over 15532.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2854, pruned_loss=0.06813, over 3310303.82 frames. ], batch size: 190, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:20:51,205 INFO [optim.py:368] (2/8) 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 06:20:53,341 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 06:21:05,591 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:55,630 INFO [train.py:904] (2/8) Epoch 5, batch 1400, loss[loss=0.2374, simple_loss=0.293, pruned_loss=0.09093, over 16929.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2853, pruned_loss=0.0678, over 3306858.77 frames. ], batch size: 90, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:22:08,260 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:22:09,373 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:23:05,348 INFO [train.py:904] (2/8) Epoch 5, batch 1450, loss[loss=0.1722, simple_loss=0.2499, pruned_loss=0.04722, over 16738.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2839, pruned_loss=0.06752, over 3307757.17 frames. ], batch size: 39, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:23:15,558 INFO [optim.py:368] (2/8) 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 06:24:14,180 INFO [train.py:904] (2/8) Epoch 5, batch 1500, loss[loss=0.2406, simple_loss=0.3016, pruned_loss=0.08985, over 12219.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2829, pruned_loss=0.06636, over 3312540.83 frames. ], batch size: 247, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:24:41,104 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4526, 3.3053, 2.7608, 2.2136, 2.3889, 2.1128, 3.2354, 3.2863], device='cuda:2'), covar=tensor([0.1838, 0.0565, 0.0952, 0.1336, 0.1915, 0.1324, 0.0396, 0.0751], device='cuda:2'), in_proj_covar=tensor([0.0274, 0.0251, 0.0268, 0.0241, 0.0297, 0.0201, 0.0238, 0.0260], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:25:21,155 INFO [train.py:904] (2/8) Epoch 5, batch 1550, loss[loss=0.2687, simple_loss=0.3211, pruned_loss=0.1082, over 16732.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2856, pruned_loss=0.06884, over 3318440.97 frames. ], batch size: 76, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:25:22,034 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 06:25:32,954 INFO [optim.py:368] (2/8) 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 06:25:51,290 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 06:26:11,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2036, 4.8874, 5.0889, 5.4292, 5.5303, 4.7872, 5.4913, 5.4630], device='cuda:2'), covar=tensor([0.0748, 0.0742, 0.1305, 0.0415, 0.0359, 0.0515, 0.0348, 0.0356], device='cuda:2'), in_proj_covar=tensor([0.0406, 0.0502, 0.0639, 0.0510, 0.0383, 0.0371, 0.0400, 0.0421], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:26:32,627 INFO [train.py:904] (2/8) Epoch 5, batch 1600, loss[loss=0.2098, simple_loss=0.276, pruned_loss=0.07184, over 16905.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2885, pruned_loss=0.07049, over 3304971.32 frames. ], batch size: 90, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:27:39,756 INFO [train.py:904] (2/8) Epoch 5, batch 1650, loss[loss=0.2265, simple_loss=0.2916, pruned_loss=0.08069, over 16944.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2883, pruned_loss=0.06962, over 3315783.10 frames. ], batch size: 96, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:27:49,823 INFO [optim.py:368] (2/8) 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 06:27:56,371 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:49,288 INFO [train.py:904] (2/8) Epoch 5, batch 1700, loss[loss=0.2081, simple_loss=0.2981, pruned_loss=0.05902, over 16719.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2899, pruned_loss=0.07003, over 3320530.09 frames. ], batch size: 57, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:28:54,202 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:29:01,194 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:29:56,499 INFO [train.py:904] (2/8) Epoch 5, batch 1750, loss[loss=0.2168, simple_loss=0.2951, pruned_loss=0.06924, over 17200.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.291, pruned_loss=0.06988, over 3325851.47 frames. ], batch size: 44, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:30:05,725 INFO [optim.py:368] (2/8) 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 06:30:05,953 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:30:07,405 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3738, 4.3176, 4.0045, 1.8381, 2.9365, 2.6733, 3.6629, 3.9350], device='cuda:2'), covar=tensor([0.0271, 0.0474, 0.0396, 0.1538, 0.0735, 0.0839, 0.0630, 0.0777], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0131, 0.0150, 0.0139, 0.0132, 0.0123, 0.0138, 0.0137], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 06:30:35,428 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:31:06,099 INFO [train.py:904] (2/8) Epoch 5, batch 1800, loss[loss=0.1887, simple_loss=0.2759, pruned_loss=0.05073, over 17204.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.293, pruned_loss=0.07072, over 3320086.02 frames. ], batch size: 44, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:31:26,121 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7856, 4.6577, 4.6276, 4.4929, 4.1590, 4.6197, 4.6768, 4.3246], device='cuda:2'), covar=tensor([0.0429, 0.0311, 0.0201, 0.0194, 0.0856, 0.0306, 0.0305, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0191, 0.0231, 0.0200, 0.0268, 0.0223, 0.0168, 0.0253], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 06:32:01,376 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 5, batch 1850, loss[loss=0.2683, simple_loss=0.3306, pruned_loss=0.103, over 15478.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2952, pruned_loss=0.07161, over 3305615.79 frames. ], batch size: 190, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:26,219 INFO [optim.py:368] (2/8) 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 06:33:25,574 INFO [train.py:904] (2/8) Epoch 5, batch 1900, loss[loss=0.214, simple_loss=0.2873, pruned_loss=0.07032, over 16485.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2944, pruned_loss=0.07132, over 3306728.18 frames. ], batch size: 68, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:33:32,654 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:16,445 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 06:34:35,496 INFO [train.py:904] (2/8) Epoch 5, batch 1950, loss[loss=0.1913, simple_loss=0.2711, pruned_loss=0.05578, over 16667.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2938, pruned_loss=0.07043, over 3300092.64 frames. ], batch size: 89, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:34:47,179 INFO [optim.py:368] (2/8) 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 06:34:52,209 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:59,454 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:35:36,276 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6546, 4.5097, 4.5751, 4.6712, 4.5719, 5.1734, 4.8217, 4.5085], device='cuda:2'), covar=tensor([0.1040, 0.1655, 0.1380, 0.1714, 0.2602, 0.0971, 0.1109, 0.2458], device='cuda:2'), in_proj_covar=tensor([0.0276, 0.0406, 0.0383, 0.0344, 0.0465, 0.0413, 0.0317, 0.0459], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 06:35:48,039 INFO [train.py:904] (2/8) Epoch 5, batch 2000, loss[loss=0.1825, simple_loss=0.2565, pruned_loss=0.05426, over 16838.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2937, pruned_loss=0.07058, over 3301326.80 frames. ], batch size: 39, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:35:51,692 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:36:00,883 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:20,903 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7330, 3.8662, 3.0268, 2.4229, 2.7771, 2.3215, 3.6934, 3.8092], device='cuda:2'), covar=tensor([0.1799, 0.0487, 0.0988, 0.1427, 0.2123, 0.1373, 0.0402, 0.0683], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0254, 0.0271, 0.0243, 0.0304, 0.0203, 0.0240, 0.0265], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 06:36:47,472 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2124, 4.9247, 5.2166, 5.4641, 5.5936, 4.7057, 5.5186, 5.5147], device='cuda:2'), covar=tensor([0.0861, 0.0743, 0.1326, 0.0436, 0.0390, 0.0505, 0.0383, 0.0370], device='cuda:2'), in_proj_covar=tensor([0.0405, 0.0494, 0.0636, 0.0513, 0.0381, 0.0373, 0.0399, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:36:55,051 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:57,536 INFO [train.py:904] (2/8) Epoch 5, batch 2050, loss[loss=0.2417, simple_loss=0.3012, pruned_loss=0.09116, over 16859.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2937, pruned_loss=0.07128, over 3306994.86 frames. ], batch size: 116, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:36:59,031 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:37:06,886 INFO [optim.py:368] (2/8) 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 06:37:40,910 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-28 06:38:05,313 INFO [train.py:904] (2/8) Epoch 5, batch 2100, loss[loss=0.1834, simple_loss=0.2615, pruned_loss=0.05265, over 16786.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2951, pruned_loss=0.07158, over 3313206.42 frames. ], batch size: 39, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:38:18,145 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:38:35,601 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5520, 3.6288, 2.8284, 2.2003, 2.6290, 2.1057, 3.4511, 3.5442], device='cuda:2'), covar=tensor([0.1912, 0.0512, 0.0986, 0.1503, 0.1912, 0.1425, 0.0420, 0.0725], device='cuda:2'), in_proj_covar=tensor([0.0275, 0.0250, 0.0268, 0.0242, 0.0301, 0.0200, 0.0238, 0.0261], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:38:37,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8981, 5.3046, 4.9835, 5.0020, 4.6706, 4.4389, 4.7490, 5.3013], device='cuda:2'), covar=tensor([0.0762, 0.0619, 0.0920, 0.0454, 0.0700, 0.0911, 0.0687, 0.0700], device='cuda:2'), in_proj_covar=tensor([0.0375, 0.0507, 0.0423, 0.0324, 0.0322, 0.0325, 0.0403, 0.0360], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:38:42,669 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8231, 3.3026, 2.6090, 4.5124, 4.0799, 4.1182, 1.5861, 3.1120], device='cuda:2'), covar=tensor([0.1277, 0.0449, 0.1023, 0.0079, 0.0287, 0.0295, 0.1326, 0.0634], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0138, 0.0164, 0.0084, 0.0174, 0.0168, 0.0155, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 06:38:52,992 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:39:14,979 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 06:39:15,224 INFO [train.py:904] (2/8) Epoch 5, batch 2150, loss[loss=0.267, simple_loss=0.3217, pruned_loss=0.1061, over 15420.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.296, pruned_loss=0.07219, over 3309024.75 frames. ], batch size: 190, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:39:24,098 INFO [optim.py:368] (2/8) 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 06:39:56,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3731, 2.2556, 1.9449, 2.1194, 2.7768, 2.5031, 3.6278, 3.0303], device='cuda:2'), covar=tensor([0.0031, 0.0195, 0.0214, 0.0196, 0.0116, 0.0169, 0.0068, 0.0116], device='cuda:2'), in_proj_covar=tensor([0.0085, 0.0152, 0.0150, 0.0148, 0.0148, 0.0154, 0.0127, 0.0135], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:40:18,219 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 06:40:23,566 INFO [train.py:904] (2/8) Epoch 5, batch 2200, loss[loss=0.1956, simple_loss=0.2703, pruned_loss=0.06041, over 16806.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2954, pruned_loss=0.07195, over 3316147.66 frames. ], batch size: 39, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:40:30,056 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7767, 3.9099, 2.9723, 2.3787, 2.9126, 2.2483, 3.8027, 3.8904], device='cuda:2'), covar=tensor([0.1884, 0.0536, 0.1171, 0.1443, 0.1997, 0.1384, 0.0409, 0.0642], device='cuda:2'), in_proj_covar=tensor([0.0276, 0.0251, 0.0270, 0.0242, 0.0305, 0.0202, 0.0239, 0.0261], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:40:52,626 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 06:41:06,763 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4892, 4.3647, 4.3231, 4.2825, 4.0201, 4.3826, 4.2310, 4.0864], device='cuda:2'), covar=tensor([0.0410, 0.0360, 0.0199, 0.0202, 0.0733, 0.0287, 0.0450, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0193, 0.0231, 0.0202, 0.0266, 0.0221, 0.0169, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 06:41:34,698 INFO [train.py:904] (2/8) Epoch 5, batch 2250, loss[loss=0.3459, simple_loss=0.3993, pruned_loss=0.1462, over 11823.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.2966, pruned_loss=0.07249, over 3314381.72 frames. ], batch size: 246, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:41:43,706 INFO [optim.py:368] (2/8) 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 06:41:46,005 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:49,535 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:52,950 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 06:42:14,355 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:44,587 INFO [train.py:904] (2/8) Epoch 5, batch 2300, loss[loss=0.2497, simple_loss=0.3044, pruned_loss=0.09753, over 16769.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2971, pruned_loss=0.07249, over 3324378.87 frames. ], batch size: 134, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:43:11,623 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:38,227 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:42,456 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1916, 4.6607, 3.6750, 2.5932, 3.4557, 2.4572, 4.7456, 4.5851], device='cuda:2'), covar=tensor([0.1918, 0.0417, 0.0992, 0.1512, 0.2311, 0.1436, 0.0314, 0.0477], device='cuda:2'), in_proj_covar=tensor([0.0275, 0.0251, 0.0268, 0.0242, 0.0304, 0.0202, 0.0239, 0.0261], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:43:53,244 INFO [train.py:904] (2/8) Epoch 5, batch 2350, loss[loss=0.2393, simple_loss=0.3048, pruned_loss=0.08686, over 16679.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.2977, pruned_loss=0.07358, over 3321454.60 frames. ], batch size: 134, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:44:03,299 INFO [optim.py:368] (2/8) 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 06:44:14,055 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7910, 1.2733, 1.5129, 1.7143, 1.8455, 1.9424, 1.3919, 1.7007], device='cuda:2'), covar=tensor([0.0084, 0.0192, 0.0100, 0.0137, 0.0098, 0.0067, 0.0172, 0.0040], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0142, 0.0126, 0.0128, 0.0125, 0.0091, 0.0137, 0.0080], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 06:44:25,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3022, 5.3692, 5.2275, 4.5923, 5.1591, 2.2822, 4.9379, 5.2240], device='cuda:2'), covar=tensor([0.0053, 0.0035, 0.0066, 0.0269, 0.0051, 0.1225, 0.0065, 0.0088], device='cuda:2'), in_proj_covar=tensor([0.0097, 0.0083, 0.0128, 0.0134, 0.0097, 0.0138, 0.0111, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 06:44:26,865 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4401, 3.9408, 4.0611, 1.7790, 4.1419, 4.1707, 3.3641, 3.0034], device='cuda:2'), covar=tensor([0.0720, 0.0087, 0.0119, 0.1170, 0.0061, 0.0087, 0.0242, 0.0393], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0085, 0.0085, 0.0139, 0.0071, 0.0079, 0.0113, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 06:45:02,226 INFO [train.py:904] (2/8) Epoch 5, batch 2400, loss[loss=0.2781, simple_loss=0.3296, pruned_loss=0.1133, over 16713.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.2987, pruned_loss=0.07435, over 3311682.55 frames. ], batch size: 124, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:45:08,767 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:30,666 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 06:45:52,166 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:46:13,639 INFO [train.py:904] (2/8) Epoch 5, batch 2450, loss[loss=0.2223, simple_loss=0.3056, pruned_loss=0.06951, over 16610.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.299, pruned_loss=0.07394, over 3311571.46 frames. ], batch size: 62, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:46:26,011 INFO [optim.py:368] (2/8) 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 06:46:57,983 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:47:08,058 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4479, 4.2695, 3.9516, 1.8011, 3.2665, 2.4561, 3.7867, 3.8472], device='cuda:2'), covar=tensor([0.0242, 0.0451, 0.0397, 0.1568, 0.0625, 0.0905, 0.0582, 0.0803], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0132, 0.0150, 0.0139, 0.0131, 0.0124, 0.0138, 0.0137], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 06:47:23,966 INFO [train.py:904] (2/8) Epoch 5, batch 2500, loss[loss=0.2418, simple_loss=0.3407, pruned_loss=0.07143, over 16730.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.2991, pruned_loss=0.07356, over 3310253.54 frames. ], batch size: 62, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:02,109 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7444, 4.0665, 4.1888, 1.8351, 4.3809, 4.4715, 3.3152, 3.3750], device='cuda:2'), covar=tensor([0.0583, 0.0092, 0.0113, 0.1068, 0.0044, 0.0048, 0.0264, 0.0316], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0085, 0.0086, 0.0142, 0.0072, 0.0079, 0.0116, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 06:48:33,490 INFO [train.py:904] (2/8) Epoch 5, batch 2550, loss[loss=0.2338, simple_loss=0.3056, pruned_loss=0.08096, over 16696.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2986, pruned_loss=0.07336, over 3318192.01 frames. ], batch size: 134, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:45,558 INFO [optim.py:368] (2/8) 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 06:48:48,802 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:04,311 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7509, 4.6523, 4.5285, 3.6393, 4.6625, 1.7314, 4.3881, 4.5974], device='cuda:2'), covar=tensor([0.0075, 0.0071, 0.0130, 0.0448, 0.0075, 0.1971, 0.0119, 0.0170], device='cuda:2'), in_proj_covar=tensor([0.0096, 0.0083, 0.0128, 0.0136, 0.0097, 0.0140, 0.0113, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:49:43,647 INFO [train.py:904] (2/8) Epoch 5, batch 2600, loss[loss=0.1978, simple_loss=0.2862, pruned_loss=0.05469, over 17222.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2977, pruned_loss=0.07214, over 3319517.95 frames. ], batch size: 45, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:49:55,550 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:01,747 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:24,436 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2531, 5.2138, 5.0979, 4.5244, 5.1112, 1.9765, 4.7894, 5.2042], device='cuda:2'), covar=tensor([0.0051, 0.0054, 0.0075, 0.0302, 0.0054, 0.1487, 0.0089, 0.0104], device='cuda:2'), in_proj_covar=tensor([0.0095, 0.0081, 0.0125, 0.0133, 0.0095, 0.0137, 0.0111, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:50:31,265 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:37,798 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.49 vs. limit=5.0 2023-04-28 06:50:53,870 INFO [train.py:904] (2/8) Epoch 5, batch 2650, loss[loss=0.197, simple_loss=0.2817, pruned_loss=0.05609, over 17180.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2968, pruned_loss=0.07081, over 3327505.95 frames. ], batch size: 44, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:51:05,580 INFO [optim.py:368] (2/8) 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 06:51:48,171 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 06:52:02,504 INFO [train.py:904] (2/8) Epoch 5, batch 2700, loss[loss=0.2287, simple_loss=0.3, pruned_loss=0.07864, over 15603.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.297, pruned_loss=0.07033, over 3321969.17 frames. ], batch size: 190, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:52:09,034 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:03,894 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4112, 2.2534, 1.9040, 2.0514, 2.7037, 2.4830, 3.4988, 3.0683], device='cuda:2'), covar=tensor([0.0028, 0.0205, 0.0215, 0.0234, 0.0128, 0.0197, 0.0084, 0.0099], device='cuda:2'), in_proj_covar=tensor([0.0084, 0.0152, 0.0151, 0.0147, 0.0148, 0.0155, 0.0130, 0.0135], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 06:53:12,575 INFO [train.py:904] (2/8) Epoch 5, batch 2750, loss[loss=0.216, simple_loss=0.2861, pruned_loss=0.07299, over 15499.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.297, pruned_loss=0.06985, over 3319882.35 frames. ], batch size: 190, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:53:15,874 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:25,426 INFO [optim.py:368] (2/8) 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 06:54:22,964 INFO [train.py:904] (2/8) Epoch 5, batch 2800, loss[loss=0.2471, simple_loss=0.3108, pruned_loss=0.09176, over 16853.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.297, pruned_loss=0.06929, over 3321799.05 frames. ], batch size: 96, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:33,338 INFO [train.py:904] (2/8) Epoch 5, batch 2850, loss[loss=0.22, simple_loss=0.2919, pruned_loss=0.07404, over 16532.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2955, pruned_loss=0.06865, over 3325043.46 frames. ], batch size: 75, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:34,860 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8867, 4.7344, 4.7176, 4.5904, 4.3207, 4.7536, 4.6984, 4.4756], device='cuda:2'), covar=tensor([0.0388, 0.0332, 0.0178, 0.0165, 0.0747, 0.0296, 0.0256, 0.0460], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0197, 0.0229, 0.0201, 0.0267, 0.0226, 0.0170, 0.0257], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 06:55:45,525 INFO [optim.py:368] (2/8) 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 06:56:41,596 INFO [train.py:904] (2/8) Epoch 5, batch 2900, loss[loss=0.1967, simple_loss=0.2677, pruned_loss=0.0628, over 16867.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2947, pruned_loss=0.06977, over 3322835.17 frames. ], batch size: 96, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:57:00,436 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:01,402 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:28,495 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:49,458 INFO [train.py:904] (2/8) Epoch 5, batch 2950, loss[loss=0.2555, simple_loss=0.309, pruned_loss=0.101, over 16912.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2951, pruned_loss=0.0707, over 3323979.05 frames. ], batch size: 116, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:57:53,167 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-28 06:58:02,027 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:15,897 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6461, 3.3257, 2.8242, 1.8570, 2.5683, 2.0861, 3.2122, 3.1752], device='cuda:2'), covar=tensor([0.0259, 0.0503, 0.0544, 0.1499, 0.0723, 0.0936, 0.0497, 0.0646], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0133, 0.0155, 0.0141, 0.0133, 0.0126, 0.0140, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 06:58:24,093 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:34,165 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:59,870 INFO [train.py:904] (2/8) Epoch 5, batch 3000, loss[loss=0.2114, simple_loss=0.28, pruned_loss=0.07139, over 16877.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2957, pruned_loss=0.07143, over 3321273.61 frames. ], batch size: 116, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:59,871 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 06:59:08,826 INFO [train.py:938] (2/8) Epoch 5, validation: loss=0.1564, simple_loss=0.2632, pruned_loss=0.02477, over 944034.00 frames. 2023-04-28 06:59:08,826 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 06:59:30,951 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-28 06:59:50,975 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6987, 1.4541, 2.1288, 2.4392, 2.5399, 2.5031, 1.4803, 2.7349], device='cuda:2'), covar=tensor([0.0054, 0.0205, 0.0126, 0.0112, 0.0087, 0.0100, 0.0179, 0.0026], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0143, 0.0127, 0.0132, 0.0127, 0.0094, 0.0138, 0.0080], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 07:00:18,525 INFO [train.py:904] (2/8) Epoch 5, batch 3050, loss[loss=0.225, simple_loss=0.2893, pruned_loss=0.08036, over 16939.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2952, pruned_loss=0.07191, over 3313873.46 frames. ], batch size: 109, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:00:30,662 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3078, 5.2781, 5.0979, 4.8821, 4.6691, 5.0797, 5.1810, 4.7652], device='cuda:2'), covar=tensor([0.0478, 0.0237, 0.0195, 0.0175, 0.0911, 0.0265, 0.0185, 0.0518], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0199, 0.0234, 0.0204, 0.0271, 0.0229, 0.0172, 0.0257], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:00:31,425 INFO [optim.py:368] (2/8) 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 07:01:11,974 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9487, 1.6350, 2.2902, 2.7024, 2.6971, 2.5978, 1.6971, 2.8775], device='cuda:2'), covar=tensor([0.0051, 0.0211, 0.0138, 0.0102, 0.0085, 0.0116, 0.0203, 0.0034], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0142, 0.0128, 0.0131, 0.0127, 0.0093, 0.0139, 0.0081], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 07:01:23,287 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 07:01:25,947 INFO [train.py:904] (2/8) Epoch 5, batch 3100, loss[loss=0.2069, simple_loss=0.2937, pruned_loss=0.06006, over 16725.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2948, pruned_loss=0.0717, over 3322194.31 frames. ], batch size: 57, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:01:43,478 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5617, 4.6481, 5.1098, 5.0478, 5.0714, 4.6729, 4.6631, 4.4919], device='cuda:2'), covar=tensor([0.0255, 0.0322, 0.0295, 0.0403, 0.0337, 0.0243, 0.0749, 0.0371], device='cuda:2'), in_proj_covar=tensor([0.0249, 0.0243, 0.0246, 0.0248, 0.0297, 0.0258, 0.0375, 0.0220], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 07:02:00,860 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-04-28 07:02:33,558 INFO [train.py:904] (2/8) Epoch 5, batch 3150, loss[loss=0.2295, simple_loss=0.2987, pruned_loss=0.08012, over 16803.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2932, pruned_loss=0.07024, over 3329088.02 frames. ], batch size: 102, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:02:46,878 INFO [optim.py:368] (2/8) 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 07:03:13,146 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-04-28 07:03:22,238 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:03:31,068 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 07:03:42,788 INFO [train.py:904] (2/8) Epoch 5, batch 3200, loss[loss=0.2236, simple_loss=0.3032, pruned_loss=0.072, over 17259.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2925, pruned_loss=0.0694, over 3321404.52 frames. ], batch size: 52, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:03:43,937 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0263, 4.3428, 4.5035, 3.3472, 4.0282, 4.5788, 4.1639, 3.0053], device='cuda:2'), covar=tensor([0.0206, 0.0017, 0.0023, 0.0151, 0.0031, 0.0023, 0.0024, 0.0197], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0059, 0.0059, 0.0110, 0.0060, 0.0067, 0.0064, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:04:21,312 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3849, 4.2957, 4.2781, 4.2332, 3.9226, 4.3916, 4.1661, 4.0917], device='cuda:2'), covar=tensor([0.0440, 0.0314, 0.0231, 0.0178, 0.0758, 0.0248, 0.0458, 0.0427], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0195, 0.0230, 0.0200, 0.0266, 0.0227, 0.0169, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:04:48,217 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:04:52,114 INFO [train.py:904] (2/8) Epoch 5, batch 3250, loss[loss=0.2439, simple_loss=0.3067, pruned_loss=0.09055, over 16122.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2932, pruned_loss=0.06991, over 3306860.60 frames. ], batch size: 164, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:05:06,669 INFO [optim.py:368] (2/8) 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 07:05:19,444 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:05:29,860 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3012, 2.4796, 2.4020, 4.9595, 2.1102, 3.6817, 2.4790, 2.5453], device='cuda:2'), covar=tensor([0.0460, 0.2109, 0.1046, 0.0190, 0.2941, 0.0864, 0.1813, 0.2690], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0322, 0.0256, 0.0312, 0.0364, 0.0309, 0.0285, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:06:03,320 INFO [train.py:904] (2/8) Epoch 5, batch 3300, loss[loss=0.2942, simple_loss=0.3544, pruned_loss=0.1169, over 11899.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2934, pruned_loss=0.06982, over 3312766.21 frames. ], batch size: 246, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:06:14,707 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:12,748 INFO [train.py:904] (2/8) Epoch 5, batch 3350, loss[loss=0.2078, simple_loss=0.2954, pruned_loss=0.06015, over 17060.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.293, pruned_loss=0.06932, over 3312209.23 frames. ], batch size: 50, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:07:24,200 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:26,940 INFO [optim.py:368] (2/8) 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 07:07:39,243 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:08:22,243 INFO [train.py:904] (2/8) Epoch 5, batch 3400, loss[loss=0.2028, simple_loss=0.2751, pruned_loss=0.06522, over 16752.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2936, pruned_loss=0.06957, over 3308244.11 frames. ], batch size: 39, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:08:47,618 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:09:31,566 INFO [train.py:904] (2/8) Epoch 5, batch 3450, loss[loss=0.2504, simple_loss=0.3139, pruned_loss=0.09343, over 16771.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2927, pruned_loss=0.06917, over 3313864.77 frames. ], batch size: 124, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:09:44,808 INFO [optim.py:368] (2/8) 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 07:10:39,130 INFO [train.py:904] (2/8) Epoch 5, batch 3500, loss[loss=0.1959, simple_loss=0.2853, pruned_loss=0.05327, over 17192.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2914, pruned_loss=0.06829, over 3312138.46 frames. ], batch size: 44, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:11:04,122 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9491, 4.3998, 4.4780, 1.8595, 4.7532, 4.7377, 3.5481, 3.7831], device='cuda:2'), covar=tensor([0.0638, 0.0082, 0.0149, 0.1116, 0.0032, 0.0050, 0.0247, 0.0284], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0086, 0.0085, 0.0141, 0.0073, 0.0081, 0.0116, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 07:11:10,768 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1522, 4.3564, 4.4460, 3.5232, 4.0885, 4.6130, 4.1545, 2.9310], device='cuda:2'), covar=tensor([0.0213, 0.0024, 0.0026, 0.0156, 0.0029, 0.0026, 0.0027, 0.0210], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0061, 0.0062, 0.0114, 0.0062, 0.0069, 0.0067, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:11:37,786 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:11:49,615 INFO [train.py:904] (2/8) Epoch 5, batch 3550, loss[loss=0.1792, simple_loss=0.2694, pruned_loss=0.04448, over 17119.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2896, pruned_loss=0.06722, over 3312242.53 frames. ], batch size: 47, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:11:49,964 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0898, 5.0504, 4.8286, 4.1245, 4.8111, 1.7582, 4.6014, 4.8638], device='cuda:2'), covar=tensor([0.0070, 0.0056, 0.0092, 0.0364, 0.0077, 0.1607, 0.0111, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0086, 0.0129, 0.0137, 0.0100, 0.0137, 0.0115, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:12:03,021 INFO [optim.py:368] (2/8) 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 07:12:17,254 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:12:52,248 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4717, 4.2075, 3.5690, 2.0287, 3.0860, 2.4035, 3.8234, 3.9157], device='cuda:2'), covar=tensor([0.0248, 0.0511, 0.0544, 0.1524, 0.0685, 0.0964, 0.0542, 0.0788], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0133, 0.0153, 0.0138, 0.0131, 0.0124, 0.0140, 0.0142], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 07:12:59,202 INFO [train.py:904] (2/8) Epoch 5, batch 3600, loss[loss=0.2253, simple_loss=0.2947, pruned_loss=0.07791, over 11717.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.288, pruned_loss=0.06689, over 3312455.72 frames. ], batch size: 246, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:13:23,274 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:13:28,087 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6613, 4.5830, 4.6314, 4.7006, 4.5738, 5.1823, 4.8168, 4.4486], device='cuda:2'), covar=tensor([0.1130, 0.1686, 0.1425, 0.1782, 0.2805, 0.0983, 0.1404, 0.3073], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0409, 0.0397, 0.0347, 0.0464, 0.0425, 0.0332, 0.0477], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:14:09,666 INFO [train.py:904] (2/8) Epoch 5, batch 3650, loss[loss=0.2183, simple_loss=0.2832, pruned_loss=0.07675, over 16507.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.287, pruned_loss=0.06697, over 3308790.43 frames. ], batch size: 146, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:14:25,432 INFO [optim.py:368] (2/8) 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 07:14:30,508 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:14:37,863 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-28 07:15:22,474 INFO [train.py:904] (2/8) Epoch 5, batch 3700, loss[loss=0.2103, simple_loss=0.2762, pruned_loss=0.07217, over 16087.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2864, pruned_loss=0.06914, over 3279194.05 frames. ], batch size: 165, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:15:44,548 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 5, batch 3750, loss[loss=0.2049, simple_loss=0.2672, pruned_loss=0.07129, over 16473.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2865, pruned_loss=0.07076, over 3273374.97 frames. ], batch size: 75, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:16:52,716 INFO [optim.py:368] (2/8) 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 07:17:49,965 INFO [train.py:904] (2/8) Epoch 5, batch 3800, loss[loss=0.2012, simple_loss=0.2793, pruned_loss=0.06152, over 16544.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2878, pruned_loss=0.0725, over 3280848.71 frames. ], batch size: 68, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:18:14,614 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1348, 4.0805, 4.1210, 3.6635, 4.1205, 1.8019, 3.9797, 3.8916], device='cuda:2'), covar=tensor([0.0074, 0.0066, 0.0081, 0.0257, 0.0053, 0.1571, 0.0082, 0.0126], device='cuda:2'), in_proj_covar=tensor([0.0097, 0.0086, 0.0127, 0.0135, 0.0098, 0.0139, 0.0113, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:18:50,507 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:19:01,028 INFO [train.py:904] (2/8) Epoch 5, batch 3850, loss[loss=0.1944, simple_loss=0.2656, pruned_loss=0.0616, over 16797.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2888, pruned_loss=0.07369, over 3261219.21 frames. ], batch size: 102, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:19:16,584 INFO [optim.py:368] (2/8) 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 07:19:32,408 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5923, 4.2068, 3.7262, 2.0733, 2.9372, 2.6449, 3.7749, 4.0240], device='cuda:2'), covar=tensor([0.0178, 0.0456, 0.0455, 0.1443, 0.0688, 0.0775, 0.0498, 0.0623], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0132, 0.0151, 0.0137, 0.0129, 0.0123, 0.0137, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 07:19:42,197 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3705, 4.4483, 4.8715, 4.8485, 4.8178, 4.4204, 4.4786, 4.2728], device='cuda:2'), covar=tensor([0.0248, 0.0457, 0.0254, 0.0316, 0.0388, 0.0278, 0.0784, 0.0446], device='cuda:2'), in_proj_covar=tensor([0.0246, 0.0237, 0.0241, 0.0245, 0.0291, 0.0255, 0.0363, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 07:19:57,673 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 5, batch 3900, loss[loss=0.3007, simple_loss=0.3581, pruned_loss=0.1217, over 12387.00 frames. ], tot_loss[loss=0.218, simple_loss=0.288, pruned_loss=0.07401, over 3264075.31 frames. ], batch size: 246, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:20:24,580 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1182, 3.2253, 1.6725, 3.2622, 2.3275, 3.2301, 1.7833, 2.4684], device='cuda:2'), covar=tensor([0.0143, 0.0259, 0.1458, 0.0092, 0.0750, 0.0391, 0.1324, 0.0647], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0160, 0.0177, 0.0088, 0.0163, 0.0193, 0.0186, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 07:20:49,851 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:21:22,517 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:21:23,249 INFO [train.py:904] (2/8) Epoch 5, batch 3950, loss[loss=0.2117, simple_loss=0.2776, pruned_loss=0.07288, over 16527.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2879, pruned_loss=0.07507, over 3263381.37 frames. ], batch size: 75, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:21:37,715 INFO [optim.py:368] (2/8) 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 07:21:44,218 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:15,289 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:34,890 INFO [train.py:904] (2/8) Epoch 5, batch 4000, loss[loss=0.1922, simple_loss=0.2659, pruned_loss=0.05925, over 16471.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.2874, pruned_loss=0.07523, over 3270965.06 frames. ], batch size: 68, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:22:43,918 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 07:22:49,519 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:51,758 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:54,876 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:23:45,848 INFO [train.py:904] (2/8) Epoch 5, batch 4050, loss[loss=0.2029, simple_loss=0.2763, pruned_loss=0.06472, over 16619.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2862, pruned_loss=0.07274, over 3279437.93 frames. ], batch size: 57, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:24:02,949 INFO [optim.py:368] (2/8) 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 07:24:04,434 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:25:01,112 INFO [train.py:904] (2/8) Epoch 5, batch 4100, loss[loss=0.2117, simple_loss=0.2838, pruned_loss=0.06974, over 16297.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2874, pruned_loss=0.07191, over 3265050.81 frames. ], batch size: 35, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:14,992 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0853, 3.0254, 2.6170, 1.9212, 2.5503, 2.1685, 2.6746, 2.8826], device='cuda:2'), covar=tensor([0.0318, 0.0420, 0.0521, 0.1391, 0.0634, 0.0748, 0.0678, 0.0530], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0133, 0.0153, 0.0140, 0.0133, 0.0125, 0.0141, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 07:26:16,728 INFO [train.py:904] (2/8) Epoch 5, batch 4150, loss[loss=0.2682, simple_loss=0.3423, pruned_loss=0.09703, over 15434.00 frames. ], tot_loss[loss=0.224, simple_loss=0.2961, pruned_loss=0.07593, over 3235390.86 frames. ], batch size: 190, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:34,908 INFO [optim.py:368] (2/8) 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 07:27:12,878 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4682, 3.5408, 3.2475, 3.2073, 3.0781, 3.3233, 3.2136, 3.0743], device='cuda:2'), covar=tensor([0.0355, 0.0218, 0.0182, 0.0151, 0.0453, 0.0244, 0.0883, 0.0356], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0177, 0.0209, 0.0184, 0.0241, 0.0207, 0.0151, 0.0229], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:27:35,372 INFO [train.py:904] (2/8) Epoch 5, batch 4200, loss[loss=0.2568, simple_loss=0.3406, pruned_loss=0.08651, over 16210.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3043, pruned_loss=0.07813, over 3222251.26 frames. ], batch size: 165, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:27:53,603 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9632, 2.7102, 2.6779, 1.7310, 2.8082, 2.8364, 2.3499, 2.3240], device='cuda:2'), covar=tensor([0.0709, 0.0130, 0.0165, 0.0902, 0.0071, 0.0087, 0.0367, 0.0386], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0087, 0.0081, 0.0142, 0.0071, 0.0076, 0.0116, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 07:28:51,234 INFO [train.py:904] (2/8) Epoch 5, batch 4250, loss[loss=0.2171, simple_loss=0.3079, pruned_loss=0.06313, over 16707.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3078, pruned_loss=0.07839, over 3210681.75 frames. ], batch size: 134, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:29:07,183 INFO [optim.py:368] (2/8) 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 07:29:15,231 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8869, 2.3400, 2.2109, 4.6036, 1.8832, 3.3386, 2.4444, 2.5371], device='cuda:2'), covar=tensor([0.0583, 0.2043, 0.1133, 0.0256, 0.3075, 0.0824, 0.1810, 0.2409], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0318, 0.0257, 0.0306, 0.0360, 0.0303, 0.0285, 0.0387], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:29:38,872 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44883.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:30:06,000 INFO [train.py:904] (2/8) Epoch 5, batch 4300, loss[loss=0.2152, simple_loss=0.303, pruned_loss=0.06372, over 16549.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3086, pruned_loss=0.07743, over 3203555.57 frames. ], batch size: 75, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:30:13,091 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:30:38,109 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 07:30:50,106 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2371, 4.4748, 4.5473, 2.4580, 4.1228, 4.5056, 4.3209, 2.3424], device='cuda:2'), covar=tensor([0.0305, 0.0008, 0.0018, 0.0261, 0.0024, 0.0020, 0.0014, 0.0250], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0054, 0.0057, 0.0111, 0.0058, 0.0065, 0.0061, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:31:10,096 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 07:31:19,635 INFO [train.py:904] (2/8) Epoch 5, batch 4350, loss[loss=0.2497, simple_loss=0.33, pruned_loss=0.08466, over 16422.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3129, pruned_loss=0.07885, over 3209106.07 frames. ], batch size: 68, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:31:36,570 INFO [optim.py:368] (2/8) 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 07:32:08,284 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4022, 4.1660, 4.4250, 4.6242, 4.7181, 4.2547, 4.7035, 4.7113], device='cuda:2'), covar=tensor([0.0766, 0.0719, 0.0953, 0.0360, 0.0335, 0.0629, 0.0364, 0.0357], device='cuda:2'), in_proj_covar=tensor([0.0353, 0.0438, 0.0555, 0.0450, 0.0333, 0.0336, 0.0347, 0.0375], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:32:35,453 INFO [train.py:904] (2/8) Epoch 5, batch 4400, loss[loss=0.2463, simple_loss=0.3201, pruned_loss=0.08622, over 16340.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3143, pruned_loss=0.07952, over 3198026.64 frames. ], batch size: 35, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:32:45,576 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-04-28 07:33:14,600 INFO [scaling.py:679] (2/8) 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] (2/8) Epoch 5, batch 4450, loss[loss=0.252, simple_loss=0.3356, pruned_loss=0.08421, over 15271.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3166, pruned_loss=0.07955, over 3207874.17 frames. ], batch size: 190, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:56,517 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:34:06,116 INFO [optim.py:368] (2/8) 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 07:34:33,619 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:35:01,130 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6989, 2.5475, 2.3411, 3.7513, 2.9785, 3.4893, 1.5110, 2.7631], device='cuda:2'), covar=tensor([0.1335, 0.0613, 0.1082, 0.0082, 0.0217, 0.0337, 0.1435, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0139, 0.0163, 0.0083, 0.0172, 0.0168, 0.0158, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 07:35:01,843 INFO [train.py:904] (2/8) Epoch 5, batch 4500, loss[loss=0.2394, simple_loss=0.3162, pruned_loss=0.0813, over 15489.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3166, pruned_loss=0.07927, over 3212440.36 frames. ], batch size: 190, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:35:14,616 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0226, 5.0675, 4.8376, 4.7093, 4.3561, 4.8859, 4.7879, 4.5499], device='cuda:2'), covar=tensor([0.0369, 0.0106, 0.0144, 0.0129, 0.0929, 0.0151, 0.0163, 0.0376], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0171, 0.0204, 0.0178, 0.0238, 0.0199, 0.0148, 0.0220], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:35:26,538 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:35:44,084 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5843, 3.6514, 1.7395, 4.0734, 2.5806, 3.9320, 1.9263, 2.6191], device='cuda:2'), covar=tensor([0.0119, 0.0214, 0.1590, 0.0033, 0.0791, 0.0274, 0.1473, 0.0691], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0151, 0.0173, 0.0081, 0.0159, 0.0181, 0.0184, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 07:36:02,272 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:36:13,663 INFO [train.py:904] (2/8) Epoch 5, batch 4550, loss[loss=0.2738, simple_loss=0.3469, pruned_loss=0.1004, over 16396.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3166, pruned_loss=0.07935, over 3226756.47 frames. ], batch size: 35, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:36:30,358 INFO [optim.py:368] (2/8) 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 07:37:00,714 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:37:08,717 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7598, 4.5314, 4.7566, 4.9800, 5.1008, 4.5355, 5.0571, 5.0715], device='cuda:2'), covar=tensor([0.0840, 0.0676, 0.1041, 0.0358, 0.0257, 0.0475, 0.0302, 0.0294], device='cuda:2'), in_proj_covar=tensor([0.0349, 0.0436, 0.0551, 0.0444, 0.0331, 0.0335, 0.0344, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:37:27,614 INFO [train.py:904] (2/8) Epoch 5, batch 4600, loss[loss=0.2149, simple_loss=0.3065, pruned_loss=0.06161, over 16867.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3172, pruned_loss=0.07917, over 3215910.38 frames. ], batch size: 116, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:37:35,912 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:12,284 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:22,022 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-04-28 07:38:40,729 INFO [train.py:904] (2/8) Epoch 5, batch 4650, loss[loss=0.2503, simple_loss=0.3209, pruned_loss=0.08987, over 16709.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3154, pruned_loss=0.07856, over 3222342.22 frames. ], batch size: 57, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:38:45,218 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:57,178 INFO [optim.py:368] (2/8) 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 07:39:25,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4946, 1.9632, 2.1008, 4.0983, 1.8073, 2.8989, 2.2259, 2.1096], device='cuda:2'), covar=tensor([0.0594, 0.2148, 0.1139, 0.0276, 0.2957, 0.0886, 0.1776, 0.2398], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0319, 0.0257, 0.0306, 0.0368, 0.0304, 0.0283, 0.0388], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:39:55,267 INFO [train.py:904] (2/8) Epoch 5, batch 4700, loss[loss=0.2176, simple_loss=0.2989, pruned_loss=0.06817, over 16852.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3122, pruned_loss=0.07686, over 3222305.30 frames. ], batch size: 39, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:40:33,589 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1237, 3.9617, 4.1197, 4.3766, 4.4428, 4.0401, 4.4168, 4.4340], device='cuda:2'), covar=tensor([0.0884, 0.0889, 0.1196, 0.0457, 0.0377, 0.0809, 0.0464, 0.0370], device='cuda:2'), in_proj_covar=tensor([0.0362, 0.0445, 0.0564, 0.0453, 0.0342, 0.0341, 0.0358, 0.0373], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:41:03,088 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:41:04,548 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 07:41:06,793 INFO [train.py:904] (2/8) Epoch 5, batch 4750, loss[loss=0.1844, simple_loss=0.2739, pruned_loss=0.04745, over 16876.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3078, pruned_loss=0.07436, over 3231097.55 frames. ], batch size: 96, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:41:22,808 INFO [optim.py:368] (2/8) 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 07:42:20,534 INFO [train.py:904] (2/8) Epoch 5, batch 4800, loss[loss=0.2191, simple_loss=0.3066, pruned_loss=0.06584, over 16703.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3038, pruned_loss=0.0723, over 3230135.84 frames. ], batch size: 89, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:42:33,272 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:42:38,023 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:43:14,511 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:43:33,759 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3948, 3.6158, 1.6958, 3.8246, 2.4288, 3.7376, 1.9242, 2.6642], device='cuda:2'), covar=tensor([0.0133, 0.0201, 0.1606, 0.0038, 0.0784, 0.0337, 0.1529, 0.0698], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0147, 0.0172, 0.0079, 0.0157, 0.0178, 0.0183, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 07:43:34,380 INFO [train.py:904] (2/8) Epoch 5, batch 4850, loss[loss=0.21, simple_loss=0.3013, pruned_loss=0.05935, over 16685.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.305, pruned_loss=0.07194, over 3220274.29 frames. ], batch size: 124, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:43:50,673 INFO [optim.py:368] (2/8) 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 07:44:47,402 INFO [train.py:904] (2/8) Epoch 5, batch 4900, loss[loss=0.1907, simple_loss=0.2849, pruned_loss=0.04827, over 16248.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.305, pruned_loss=0.07135, over 3196146.78 frames. ], batch size: 165, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:45:30,730 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5206, 4.8226, 4.9829, 4.7939, 4.9092, 5.4252, 5.0476, 4.7937], device='cuda:2'), covar=tensor([0.0857, 0.1515, 0.1139, 0.1590, 0.2073, 0.0891, 0.1074, 0.2221], device='cuda:2'), in_proj_covar=tensor([0.0271, 0.0367, 0.0360, 0.0318, 0.0422, 0.0386, 0.0302, 0.0440], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:46:01,017 INFO [train.py:904] (2/8) Epoch 5, batch 4950, loss[loss=0.2428, simple_loss=0.3299, pruned_loss=0.0779, over 16394.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3056, pruned_loss=0.07119, over 3206592.92 frames. ], batch size: 146, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:04,112 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-04-28 07:46:15,481 INFO [optim.py:368] (2/8) 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 07:47:09,025 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6499, 4.8248, 4.8036, 2.6915, 4.1194, 4.7362, 4.3981, 2.5828], device='cuda:2'), covar=tensor([0.0307, 0.0009, 0.0019, 0.0259, 0.0030, 0.0024, 0.0024, 0.0266], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0054, 0.0058, 0.0115, 0.0059, 0.0067, 0.0063, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 07:47:13,804 INFO [train.py:904] (2/8) Epoch 5, batch 5000, loss[loss=0.1917, simple_loss=0.279, pruned_loss=0.05214, over 17229.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3069, pruned_loss=0.0711, over 3209784.04 frames. ], batch size: 44, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:27,065 INFO [train.py:904] (2/8) Epoch 5, batch 5050, loss[loss=0.2417, simple_loss=0.3219, pruned_loss=0.08075, over 16265.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3071, pruned_loss=0.07062, over 3220089.57 frames. ], batch size: 165, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:43,127 INFO [optim.py:368] (2/8) 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 07:49:39,972 INFO [train.py:904] (2/8) Epoch 5, batch 5100, loss[loss=0.2016, simple_loss=0.2803, pruned_loss=0.0615, over 16699.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3057, pruned_loss=0.07039, over 3197018.82 frames. ], batch size: 62, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:49:42,354 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:49:44,451 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:49:55,979 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:50:28,270 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7911, 3.6765, 3.1033, 1.6602, 2.8048, 2.0782, 3.3332, 3.5345], device='cuda:2'), covar=tensor([0.0227, 0.0410, 0.0537, 0.1597, 0.0664, 0.0952, 0.0559, 0.0519], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0129, 0.0154, 0.0141, 0.0135, 0.0127, 0.0140, 0.0137], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 07:50:33,732 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:50:53,046 INFO [train.py:904] (2/8) Epoch 5, batch 5150, loss[loss=0.1955, simple_loss=0.2794, pruned_loss=0.0558, over 17239.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3058, pruned_loss=0.06973, over 3204329.82 frames. ], batch size: 45, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:51:04,014 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2547, 5.5806, 5.1882, 5.2489, 4.8937, 4.6909, 4.9886, 5.5877], device='cuda:2'), covar=tensor([0.0547, 0.0565, 0.0836, 0.0441, 0.0632, 0.0544, 0.0607, 0.0628], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0455, 0.0399, 0.0304, 0.0294, 0.0304, 0.0374, 0.0330], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:51:06,958 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:08,849 INFO [optim.py:368] (2/8) 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 07:51:10,571 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:26,033 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:42,659 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:52:05,634 INFO [train.py:904] (2/8) Epoch 5, batch 5200, loss[loss=0.2181, simple_loss=0.3034, pruned_loss=0.06645, over 16715.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3061, pruned_loss=0.07036, over 3195990.73 frames. ], batch size: 134, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:52:17,753 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 07:52:54,233 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:16,334 INFO [train.py:904] (2/8) Epoch 5, batch 5250, loss[loss=0.2151, simple_loss=0.2956, pruned_loss=0.06727, over 16382.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3032, pruned_loss=0.07006, over 3206018.45 frames. ], batch size: 146, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:53:26,765 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:31,052 INFO [optim.py:368] (2/8) 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 07:54:26,352 INFO [train.py:904] (2/8) Epoch 5, batch 5300, loss[loss=0.2019, simple_loss=0.2769, pruned_loss=0.06347, over 16424.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2984, pruned_loss=0.06818, over 3217385.02 frames. ], batch size: 75, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:54:53,120 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:55:24,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1343, 1.9479, 2.0916, 3.5774, 1.8454, 2.7428, 2.1879, 2.0870], device='cuda:2'), covar=tensor([0.0608, 0.1968, 0.1048, 0.0344, 0.2838, 0.0923, 0.1871, 0.2175], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0315, 0.0256, 0.0304, 0.0363, 0.0301, 0.0285, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:55:36,417 INFO [train.py:904] (2/8) Epoch 5, batch 5350, loss[loss=0.2238, simple_loss=0.3073, pruned_loss=0.07015, over 16704.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2964, pruned_loss=0.06691, over 3221626.95 frames. ], batch size: 89, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:55:53,055 INFO [optim.py:368] (2/8) 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 07:56:52,489 INFO [train.py:904] (2/8) Epoch 5, batch 5400, loss[loss=0.2285, simple_loss=0.3123, pruned_loss=0.07238, over 16556.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2988, pruned_loss=0.0677, over 3211189.75 frames. ], batch size: 68, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:56:57,464 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:57:10,994 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9321, 4.1105, 3.2743, 2.4882, 3.2970, 2.5483, 4.5134, 4.3440], device='cuda:2'), covar=tensor([0.1955, 0.0659, 0.1144, 0.1449, 0.1902, 0.1257, 0.0362, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0249, 0.0264, 0.0242, 0.0290, 0.0200, 0.0242, 0.0250], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 07:58:09,282 INFO [train.py:904] (2/8) Epoch 5, batch 5450, loss[loss=0.2798, simple_loss=0.3514, pruned_loss=0.1041, over 12107.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.303, pruned_loss=0.07033, over 3193038.25 frames. ], batch size: 248, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:58:11,174 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:58:19,235 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:58:24,984 INFO [optim.py:368] (2/8) 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 07:59:22,159 INFO [train.py:904] (2/8) Epoch 5, batch 5500, loss[loss=0.2282, simple_loss=0.3165, pruned_loss=0.06997, over 16775.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3133, pruned_loss=0.07798, over 3171340.72 frames. ], batch size: 83, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:07,584 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:00:20,116 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 08:00:24,141 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 08:00:32,871 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7110, 2.5386, 2.3066, 3.6255, 2.9434, 3.6068, 1.6792, 2.7116], device='cuda:2'), covar=tensor([0.1275, 0.0545, 0.1137, 0.0090, 0.0307, 0.0336, 0.1324, 0.0720], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0141, 0.0167, 0.0083, 0.0176, 0.0174, 0.0162, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 08:00:39,140 INFO [train.py:904] (2/8) Epoch 5, batch 5550, loss[loss=0.2884, simple_loss=0.3605, pruned_loss=0.1082, over 16521.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3222, pruned_loss=0.08469, over 3158728.18 frames. ], batch size: 75, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:47,697 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 08:00:56,876 INFO [optim.py:368] (2/8) 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 08:01:30,115 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-28 08:01:58,727 INFO [train.py:904] (2/8) Epoch 5, batch 5600, loss[loss=0.2892, simple_loss=0.3526, pruned_loss=0.1129, over 15324.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.328, pruned_loss=0.08979, over 3126850.92 frames. ], batch size: 190, lr: 1.33e-02, grad_scale: 16.0 2023-04-28 08:02:23,969 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:02:27,267 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6256, 4.8988, 4.5877, 4.6440, 4.2996, 4.2041, 4.4733, 4.9519], device='cuda:2'), covar=tensor([0.0754, 0.0689, 0.1021, 0.0494, 0.0657, 0.1050, 0.0609, 0.0657], device='cuda:2'), in_proj_covar=tensor([0.0352, 0.0469, 0.0408, 0.0305, 0.0291, 0.0311, 0.0380, 0.0336], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:02:51,738 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1270, 2.9828, 2.4902, 2.0268, 2.4997, 2.0040, 2.7016, 2.8276], device='cuda:2'), covar=tensor([0.0322, 0.0419, 0.0482, 0.1381, 0.0688, 0.0990, 0.0550, 0.0591], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0126, 0.0153, 0.0140, 0.0134, 0.0126, 0.0140, 0.0134], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 08:03:20,614 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5626, 3.8168, 3.9612, 1.6759, 4.0877, 4.3240, 3.2069, 3.1002], device='cuda:2'), covar=tensor([0.0645, 0.0121, 0.0128, 0.1105, 0.0052, 0.0034, 0.0250, 0.0365], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0087, 0.0080, 0.0138, 0.0070, 0.0075, 0.0115, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 08:03:23,171 INFO [train.py:904] (2/8) Epoch 5, batch 5650, loss[loss=0.2275, simple_loss=0.3131, pruned_loss=0.07094, over 17081.00 frames. ], tot_loss[loss=0.2631, simple_loss=0.3346, pruned_loss=0.0958, over 3101424.53 frames. ], batch size: 49, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:03:26,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5641, 3.3435, 2.7736, 2.1593, 2.4738, 2.1300, 3.4935, 3.5152], device='cuda:2'), covar=tensor([0.1978, 0.0705, 0.1225, 0.1613, 0.1925, 0.1437, 0.0398, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0273, 0.0246, 0.0262, 0.0238, 0.0287, 0.0197, 0.0237, 0.0246], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:03:42,414 INFO [optim.py:368] (2/8) 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 08:03:44,744 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3422, 1.4031, 1.8500, 2.1562, 2.2053, 2.4011, 1.3617, 2.2379], device='cuda:2'), covar=tensor([0.0067, 0.0225, 0.0106, 0.0109, 0.0089, 0.0059, 0.0217, 0.0040], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0142, 0.0127, 0.0124, 0.0126, 0.0089, 0.0140, 0.0080], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 08:04:04,192 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:04:28,628 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4015, 4.6545, 4.7232, 4.7526, 4.6520, 5.2079, 4.7701, 4.6467], device='cuda:2'), covar=tensor([0.1011, 0.1610, 0.1347, 0.1401, 0.2302, 0.0912, 0.1333, 0.2144], device='cuda:2'), in_proj_covar=tensor([0.0278, 0.0378, 0.0375, 0.0328, 0.0439, 0.0400, 0.0312, 0.0454], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 08:04:43,824 INFO [train.py:904] (2/8) Epoch 5, batch 5700, loss[loss=0.2564, simple_loss=0.3362, pruned_loss=0.08828, over 16613.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.3371, pruned_loss=0.09887, over 3072018.93 frames. ], batch size: 76, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:05:37,061 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 08:05:41,863 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:04,708 INFO [train.py:904] (2/8) Epoch 5, batch 5750, loss[loss=0.2719, simple_loss=0.3408, pruned_loss=0.1015, over 15333.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3402, pruned_loss=0.1012, over 3036985.67 frames. ], batch size: 190, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:06:16,556 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:23,498 INFO [optim.py:368] (2/8) 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 08:07:26,288 INFO [train.py:904] (2/8) Epoch 5, batch 5800, loss[loss=0.349, simple_loss=0.3827, pruned_loss=0.1576, over 12022.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3392, pruned_loss=0.09903, over 3039005.98 frames. ], batch size: 247, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:07:35,757 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:07:46,837 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3326, 2.9577, 2.4997, 2.3335, 2.2573, 2.0815, 2.8563, 3.0197], device='cuda:2'), covar=tensor([0.1704, 0.0580, 0.1096, 0.1229, 0.1770, 0.1322, 0.0390, 0.0604], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0245, 0.0264, 0.0239, 0.0290, 0.0199, 0.0238, 0.0246], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:08:13,829 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:46,734 INFO [train.py:904] (2/8) Epoch 5, batch 5850, loss[loss=0.2479, simple_loss=0.3299, pruned_loss=0.08293, over 16270.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3362, pruned_loss=0.09691, over 3024690.21 frames. ], batch size: 165, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:09:06,683 INFO [optim.py:368] (2/8) 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 08:09:29,289 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:10:09,067 INFO [train.py:904] (2/8) Epoch 5, batch 5900, loss[loss=0.2612, simple_loss=0.3383, pruned_loss=0.09204, over 16661.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3347, pruned_loss=0.09473, over 3051570.32 frames. ], batch size: 62, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:10:36,171 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:11:30,073 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6272, 3.2075, 2.6378, 4.9667, 4.1601, 4.3174, 1.9769, 2.8467], device='cuda:2'), covar=tensor([0.1568, 0.0596, 0.1225, 0.0064, 0.0352, 0.0289, 0.1363, 0.0941], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0141, 0.0167, 0.0083, 0.0175, 0.0175, 0.0163, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 08:11:31,951 INFO [train.py:904] (2/8) Epoch 5, batch 5950, loss[loss=0.2705, simple_loss=0.3559, pruned_loss=0.09251, over 16495.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3349, pruned_loss=0.09305, over 3050006.66 frames. ], batch size: 75, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:11:52,724 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 08:12:43,465 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:12:51,974 INFO [train.py:904] (2/8) Epoch 5, batch 6000, loss[loss=0.2228, simple_loss=0.3065, pruned_loss=0.06957, over 16899.00 frames. ], tot_loss[loss=0.2597, simple_loss=0.3339, pruned_loss=0.0928, over 3064014.65 frames. ], batch size: 96, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:12:51,974 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 08:13:04,045 INFO [train.py:938] (2/8) Epoch 5, validation: loss=0.1879, simple_loss=0.2992, pruned_loss=0.03826, over 944034.00 frames. 2023-04-28 08:13:04,046 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 08:13:05,693 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:13:51,747 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:27,390 INFO [train.py:904] (2/8) Epoch 5, batch 6050, loss[loss=0.2807, simple_loss=0.3613, pruned_loss=0.1001, over 16459.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.332, pruned_loss=0.09195, over 3076896.15 frames. ], batch size: 62, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:14:37,510 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:14:48,340 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:48,998 INFO [optim.py:368] (2/8) 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 08:15:45,451 INFO [train.py:904] (2/8) Epoch 5, batch 6100, loss[loss=0.279, simple_loss=0.3474, pruned_loss=0.1053, over 15311.00 frames. ], tot_loss[loss=0.2557, simple_loss=0.3313, pruned_loss=0.09006, over 3106013.67 frames. ], batch size: 191, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:16:35,426 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8106, 1.2656, 1.5328, 1.6935, 1.8264, 1.8534, 1.4097, 1.6514], device='cuda:2'), covar=tensor([0.0082, 0.0182, 0.0087, 0.0106, 0.0087, 0.0056, 0.0162, 0.0040], device='cuda:2'), in_proj_covar=tensor([0.0114, 0.0138, 0.0123, 0.0119, 0.0122, 0.0086, 0.0135, 0.0076], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 08:16:50,368 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 08:17:01,187 INFO [train.py:904] (2/8) Epoch 5, batch 6150, loss[loss=0.2115, simple_loss=0.302, pruned_loss=0.06049, over 16886.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.329, pruned_loss=0.08921, over 3100481.72 frames. ], batch size: 96, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:17:09,817 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2074, 3.3949, 1.4008, 3.6413, 2.3438, 3.5030, 1.6982, 2.6810], device='cuda:2'), covar=tensor([0.0131, 0.0249, 0.1671, 0.0043, 0.0663, 0.0382, 0.1390, 0.0502], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0145, 0.0178, 0.0080, 0.0159, 0.0183, 0.0184, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 08:17:12,196 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4623, 3.5228, 3.2563, 3.2174, 3.0477, 3.3751, 3.1845, 3.1595], device='cuda:2'), covar=tensor([0.0409, 0.0251, 0.0185, 0.0173, 0.0496, 0.0270, 0.0770, 0.0365], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0182, 0.0205, 0.0177, 0.0235, 0.0206, 0.0149, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:17:22,717 INFO [optim.py:368] (2/8) 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 08:18:10,615 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2083, 4.2027, 4.0218, 3.9497, 3.6003, 4.1347, 3.9542, 3.7808], device='cuda:2'), covar=tensor([0.0447, 0.0299, 0.0207, 0.0190, 0.0822, 0.0304, 0.0426, 0.0490], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0184, 0.0207, 0.0180, 0.0238, 0.0209, 0.0152, 0.0227], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:18:21,102 INFO [train.py:904] (2/8) Epoch 5, batch 6200, loss[loss=0.2356, simple_loss=0.315, pruned_loss=0.07805, over 16782.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3266, pruned_loss=0.08822, over 3108214.16 frames. ], batch size: 124, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:18:36,392 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:18:59,229 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 08:19:25,057 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5362, 3.5586, 4.0148, 3.9780, 3.9957, 3.6372, 3.6692, 3.6543], device='cuda:2'), covar=tensor([0.0337, 0.0506, 0.0367, 0.0441, 0.0445, 0.0343, 0.0863, 0.0457], device='cuda:2'), in_proj_covar=tensor([0.0238, 0.0227, 0.0234, 0.0232, 0.0279, 0.0246, 0.0352, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 08:19:37,502 INFO [train.py:904] (2/8) Epoch 5, batch 6250, loss[loss=0.242, simple_loss=0.329, pruned_loss=0.07751, over 16809.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3268, pruned_loss=0.08812, over 3115087.82 frames. ], batch size: 102, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:19:57,344 INFO [optim.py:368] (2/8) 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 08:20:09,414 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:20:55,067 INFO [train.py:904] (2/8) Epoch 5, batch 6300, loss[loss=0.2778, simple_loss=0.3563, pruned_loss=0.09962, over 16956.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3265, pruned_loss=0.08722, over 3108587.39 frames. ], batch size: 109, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:21:08,554 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 08:21:43,814 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:21:47,385 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-04-28 08:22:12,094 INFO [train.py:904] (2/8) Epoch 5, batch 6350, loss[loss=0.2321, simple_loss=0.3155, pruned_loss=0.07433, over 16247.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3286, pruned_loss=0.08947, over 3107882.83 frames. ], batch size: 165, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:22:13,433 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 08:22:22,791 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:31,677 INFO [optim.py:368] (2/8) 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 08:22:56,442 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:23:28,012 INFO [train.py:904] (2/8) Epoch 5, batch 6400, loss[loss=0.2281, simple_loss=0.3101, pruned_loss=0.07302, over 16481.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3293, pruned_loss=0.09101, over 3090902.42 frames. ], batch size: 75, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:24:42,540 INFO [train.py:904] (2/8) Epoch 5, batch 6450, loss[loss=0.2257, simple_loss=0.3139, pruned_loss=0.06872, over 16603.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3276, pruned_loss=0.08892, over 3104824.50 frames. ], batch size: 68, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:25:02,353 INFO [optim.py:368] (2/8) 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 08:25:28,102 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 08:25:59,690 INFO [train.py:904] (2/8) Epoch 5, batch 6500, loss[loss=0.2489, simple_loss=0.3231, pruned_loss=0.08731, over 16711.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3251, pruned_loss=0.08786, over 3106145.75 frames. ], batch size: 134, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:26:13,082 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-28 08:26:24,035 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-28 08:27:18,007 INFO [train.py:904] (2/8) Epoch 5, batch 6550, loss[loss=0.2525, simple_loss=0.3389, pruned_loss=0.08311, over 16540.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3277, pruned_loss=0.08893, over 3093332.08 frames. ], batch size: 68, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:27:37,104 INFO [optim.py:368] (2/8) 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 08:27:41,299 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:28:33,655 INFO [train.py:904] (2/8) Epoch 5, batch 6600, loss[loss=0.3556, simple_loss=0.3851, pruned_loss=0.163, over 11222.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3303, pruned_loss=0.09001, over 3083853.50 frames. ], batch size: 247, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:35,191 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:29:51,330 INFO [train.py:904] (2/8) Epoch 5, batch 6650, loss[loss=0.2389, simple_loss=0.3199, pruned_loss=0.07895, over 16897.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3304, pruned_loss=0.09039, over 3102295.85 frames. ], batch size: 96, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:51,902 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:29:52,130 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 08:30:02,690 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:11,434 INFO [optim.py:368] (2/8) 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 08:31:04,024 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.41 vs. limit=5.0 2023-04-28 08:31:04,888 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:06,914 INFO [train.py:904] (2/8) Epoch 5, batch 6700, loss[loss=0.2476, simple_loss=0.319, pruned_loss=0.08808, over 16738.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3281, pruned_loss=0.08947, over 3114183.86 frames. ], batch size: 124, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:31:09,289 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:16,566 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:47,605 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7273, 4.7451, 5.2790, 5.2128, 5.2207, 4.8449, 4.8094, 4.4151], device='cuda:2'), covar=tensor([0.0256, 0.0338, 0.0270, 0.0375, 0.0401, 0.0249, 0.0814, 0.0356], device='cuda:2'), in_proj_covar=tensor([0.0235, 0.0225, 0.0234, 0.0234, 0.0282, 0.0247, 0.0358, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 08:32:18,297 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 08:32:25,989 INFO [train.py:904] (2/8) Epoch 5, batch 6750, loss[loss=0.2463, simple_loss=0.33, pruned_loss=0.08132, over 16694.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3273, pruned_loss=0.08968, over 3107704.37 frames. ], batch size: 76, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:32:45,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7909, 3.5653, 2.9845, 1.6742, 2.6819, 2.1732, 3.1774, 3.3062], device='cuda:2'), covar=tensor([0.0293, 0.0386, 0.0591, 0.1702, 0.0749, 0.0916, 0.0570, 0.0604], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0126, 0.0153, 0.0141, 0.0131, 0.0124, 0.0139, 0.0134], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 08:32:45,840 INFO [optim.py:368] (2/8) 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 08:33:08,581 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:33:23,623 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6008, 2.1629, 2.1633, 4.2483, 1.9617, 3.1189, 2.3064, 2.3490], device='cuda:2'), covar=tensor([0.0541, 0.2085, 0.1171, 0.0258, 0.2865, 0.0874, 0.1897, 0.2214], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0320, 0.0263, 0.0311, 0.0375, 0.0309, 0.0288, 0.0382], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:33:40,656 INFO [train.py:904] (2/8) Epoch 5, batch 6800, loss[loss=0.2354, simple_loss=0.3191, pruned_loss=0.07587, over 16497.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3271, pruned_loss=0.08968, over 3092241.89 frames. ], batch size: 68, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:34:40,925 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:34:46,999 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 08:34:57,631 INFO [train.py:904] (2/8) Epoch 5, batch 6850, loss[loss=0.2625, simple_loss=0.3474, pruned_loss=0.08877, over 16768.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3276, pruned_loss=0.08941, over 3103650.76 frames. ], batch size: 124, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:35:17,578 INFO [optim.py:368] (2/8) 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 08:35:21,076 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:36:12,142 INFO [train.py:904] (2/8) Epoch 5, batch 6900, loss[loss=0.2816, simple_loss=0.3476, pruned_loss=0.1078, over 16287.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.33, pruned_loss=0.08878, over 3112333.16 frames. ], batch size: 165, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:36:33,878 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:36:46,843 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7355, 3.4692, 3.0019, 1.7674, 2.5732, 2.0314, 3.0911, 3.3474], device='cuda:2'), covar=tensor([0.0277, 0.0466, 0.0549, 0.1665, 0.0773, 0.0923, 0.0661, 0.0610], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0124, 0.0152, 0.0140, 0.0131, 0.0123, 0.0138, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 08:37:30,517 INFO [train.py:904] (2/8) Epoch 5, batch 6950, loss[loss=0.2703, simple_loss=0.338, pruned_loss=0.1013, over 15450.00 frames. ], tot_loss[loss=0.2592, simple_loss=0.3338, pruned_loss=0.09231, over 3099079.46 frames. ], batch size: 190, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:37:50,868 INFO [optim.py:368] (2/8) 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 08:37:53,611 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 08:38:42,825 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:38:49,094 INFO [train.py:904] (2/8) Epoch 5, batch 7000, loss[loss=0.228, simple_loss=0.324, pruned_loss=0.06598, over 16783.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3342, pruned_loss=0.09139, over 3105743.41 frames. ], batch size: 83, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:38:54,172 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2526, 3.7969, 3.9870, 1.4350, 4.1392, 4.2057, 3.0818, 3.0139], device='cuda:2'), covar=tensor([0.0823, 0.0103, 0.0149, 0.1317, 0.0047, 0.0046, 0.0270, 0.0374], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0087, 0.0079, 0.0140, 0.0071, 0.0076, 0.0115, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 08:40:06,954 INFO [train.py:904] (2/8) Epoch 5, batch 7050, loss[loss=0.2537, simple_loss=0.3305, pruned_loss=0.08848, over 16275.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3345, pruned_loss=0.09127, over 3104797.40 frames. ], batch size: 165, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:40:18,020 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8959, 3.0997, 3.0873, 1.5943, 3.3041, 3.3299, 2.5397, 2.4148], device='cuda:2'), covar=tensor([0.0914, 0.0150, 0.0193, 0.1215, 0.0071, 0.0068, 0.0379, 0.0479], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0087, 0.0080, 0.0141, 0.0071, 0.0076, 0.0116, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 08:40:26,886 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 5, batch 7100, loss[loss=0.2274, simple_loss=0.3066, pruned_loss=0.07405, over 16918.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3323, pruned_loss=0.09036, over 3093806.72 frames. ], batch size: 109, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:42:18,611 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7164, 1.3945, 2.1715, 2.5849, 2.4477, 2.9334, 1.5245, 2.6305], device='cuda:2'), covar=tensor([0.0079, 0.0307, 0.0149, 0.0149, 0.0127, 0.0076, 0.0247, 0.0068], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0144, 0.0127, 0.0122, 0.0127, 0.0092, 0.0139, 0.0081], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 08:42:19,693 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:24,191 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:44,668 INFO [train.py:904] (2/8) Epoch 5, batch 7150, loss[loss=0.2537, simple_loss=0.3257, pruned_loss=0.0908, over 16330.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.329, pruned_loss=0.08864, over 3103595.69 frames. ], batch size: 146, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:43:00,358 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1255, 3.2994, 3.4148, 1.4976, 3.6301, 3.6553, 2.7569, 2.6115], device='cuda:2'), covar=tensor([0.0822, 0.0174, 0.0201, 0.1315, 0.0060, 0.0060, 0.0390, 0.0454], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0087, 0.0079, 0.0140, 0.0071, 0.0075, 0.0115, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 08:43:03,967 INFO [optim.py:368] (2/8) 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 08:43:37,212 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-04-28 08:43:39,944 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:43:54,668 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-28 08:43:57,119 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:44:00,015 INFO [train.py:904] (2/8) Epoch 5, batch 7200, loss[loss=0.261, simple_loss=0.3233, pruned_loss=0.09942, over 11506.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.327, pruned_loss=0.08713, over 3085088.76 frames. ], batch size: 248, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:44:14,002 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:45:22,089 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:45:24,155 INFO [train.py:904] (2/8) Epoch 5, batch 7250, loss[loss=0.2164, simple_loss=0.2889, pruned_loss=0.07191, over 16871.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3242, pruned_loss=0.08541, over 3093163.37 frames. ], batch size: 116, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:45:43,494 INFO [optim.py:368] (2/8) 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 08:45:53,830 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 08:46:34,137 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:46:40,681 INFO [train.py:904] (2/8) Epoch 5, batch 7300, loss[loss=0.246, simple_loss=0.3237, pruned_loss=0.08417, over 15361.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3233, pruned_loss=0.08501, over 3103157.07 frames. ], batch size: 190, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:47:49,458 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:47:58,198 INFO [train.py:904] (2/8) Epoch 5, batch 7350, loss[loss=0.2548, simple_loss=0.3297, pruned_loss=0.08992, over 15402.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3234, pruned_loss=0.08546, over 3092657.91 frames. ], batch size: 190, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:48:17,667 INFO [optim.py:368] (2/8) 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 08:48:19,948 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5408, 4.7482, 4.8599, 4.8164, 4.7474, 5.3062, 4.8945, 4.6883], device='cuda:2'), covar=tensor([0.0938, 0.1437, 0.1359, 0.1370, 0.2159, 0.0903, 0.1139, 0.2011], device='cuda:2'), in_proj_covar=tensor([0.0273, 0.0369, 0.0377, 0.0329, 0.0427, 0.0398, 0.0307, 0.0445], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 08:49:18,342 INFO [train.py:904] (2/8) Epoch 5, batch 7400, loss[loss=0.2317, simple_loss=0.3157, pruned_loss=0.07383, over 16496.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3245, pruned_loss=0.08606, over 3100937.44 frames. ], batch size: 68, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:50:13,347 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:50:18,273 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-04-28 08:50:38,165 INFO [train.py:904] (2/8) Epoch 5, batch 7450, loss[loss=0.2427, simple_loss=0.3195, pruned_loss=0.083, over 16620.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3259, pruned_loss=0.0878, over 3086467.98 frames. ], batch size: 134, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:50:57,116 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8757, 3.7258, 3.9245, 4.1523, 4.2024, 3.7764, 4.1499, 4.1776], device='cuda:2'), covar=tensor([0.0978, 0.0768, 0.1101, 0.0464, 0.0426, 0.1110, 0.0521, 0.0448], device='cuda:2'), in_proj_covar=tensor([0.0367, 0.0453, 0.0574, 0.0466, 0.0345, 0.0342, 0.0368, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:51:00,811 INFO [optim.py:368] (2/8) 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 08:51:14,599 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:31,406 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:49,934 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:52:00,147 INFO [train.py:904] (2/8) Epoch 5, batch 7500, loss[loss=0.2202, simple_loss=0.3009, pruned_loss=0.06977, over 16926.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.3263, pruned_loss=0.08763, over 3079836.85 frames. ], batch size: 109, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:52:50,912 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:08,104 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:18,029 INFO [train.py:904] (2/8) Epoch 5, batch 7550, loss[loss=0.2416, simple_loss=0.3175, pruned_loss=0.08288, over 16476.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.326, pruned_loss=0.08845, over 3073209.57 frames. ], batch size: 75, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:53:38,523 INFO [optim.py:368] (2/8) 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 08:53:40,653 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:54:14,716 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7107, 4.5953, 4.5203, 4.4241, 4.0538, 4.5355, 4.5337, 4.3031], device='cuda:2'), covar=tensor([0.0467, 0.0252, 0.0211, 0.0168, 0.0905, 0.0305, 0.0241, 0.0486], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0184, 0.0207, 0.0179, 0.0236, 0.0211, 0.0152, 0.0233], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:54:34,588 INFO [train.py:904] (2/8) Epoch 5, batch 7600, loss[loss=0.2901, simple_loss=0.3421, pruned_loss=0.119, over 11566.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.325, pruned_loss=0.08811, over 3089188.11 frames. ], batch size: 246, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:54:51,921 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:55:48,839 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2225, 3.3363, 1.6842, 3.5548, 2.3184, 3.5486, 1.8597, 2.6192], device='cuda:2'), covar=tensor([0.0146, 0.0324, 0.1696, 0.0053, 0.0799, 0.0435, 0.1463, 0.0583], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0151, 0.0178, 0.0080, 0.0162, 0.0180, 0.0188, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 08:55:51,252 INFO [train.py:904] (2/8) Epoch 5, batch 7650, loss[loss=0.271, simple_loss=0.3403, pruned_loss=0.1008, over 15341.00 frames. ], tot_loss[loss=0.2522, simple_loss=0.326, pruned_loss=0.08916, over 3085595.06 frames. ], batch size: 190, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:56:12,528 INFO [optim.py:368] (2/8) 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 08:56:25,996 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 08:57:08,648 INFO [train.py:904] (2/8) Epoch 5, batch 7700, loss[loss=0.216, simple_loss=0.3035, pruned_loss=0.06423, over 16791.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3275, pruned_loss=0.09059, over 3074516.23 frames. ], batch size: 102, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:57:09,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6397, 3.9752, 4.1272, 1.9440, 4.2705, 4.3311, 3.3155, 3.1719], device='cuda:2'), covar=tensor([0.0740, 0.0097, 0.0148, 0.1114, 0.0043, 0.0059, 0.0306, 0.0381], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0087, 0.0082, 0.0142, 0.0071, 0.0078, 0.0116, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 08:57:12,663 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5666, 5.9392, 5.6309, 5.8111, 5.1636, 4.8622, 5.4321, 6.0804], device='cuda:2'), covar=tensor([0.0693, 0.0646, 0.0881, 0.0445, 0.0653, 0.0648, 0.0584, 0.0642], device='cuda:2'), in_proj_covar=tensor([0.0365, 0.0480, 0.0413, 0.0306, 0.0297, 0.0320, 0.0386, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:57:32,427 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7663, 4.4915, 4.4260, 5.0474, 5.1034, 4.6178, 5.1629, 5.1101], device='cuda:2'), covar=tensor([0.0988, 0.0923, 0.2050, 0.0680, 0.0651, 0.0631, 0.0637, 0.0626], device='cuda:2'), in_proj_covar=tensor([0.0371, 0.0461, 0.0587, 0.0475, 0.0353, 0.0349, 0.0374, 0.0394], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:57:44,099 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:58:26,511 INFO [train.py:904] (2/8) Epoch 5, batch 7750, loss[loss=0.2081, simple_loss=0.3002, pruned_loss=0.05795, over 16880.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3271, pruned_loss=0.09005, over 3074002.23 frames. ], batch size: 96, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:58:26,967 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2973, 4.6312, 4.3202, 4.4174, 3.9836, 4.0384, 4.1612, 4.6110], device='cuda:2'), covar=tensor([0.0766, 0.0713, 0.0946, 0.0481, 0.0641, 0.0990, 0.0679, 0.0771], device='cuda:2'), in_proj_covar=tensor([0.0361, 0.0473, 0.0409, 0.0303, 0.0293, 0.0318, 0.0384, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:58:47,830 INFO [optim.py:368] (2/8) 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 08:58:53,639 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9210, 1.8987, 2.2068, 3.0452, 2.0809, 2.4529, 2.2684, 1.9179], device='cuda:2'), covar=tensor([0.0619, 0.1992, 0.0988, 0.0383, 0.2583, 0.1166, 0.1626, 0.2233], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0319, 0.0260, 0.0307, 0.0373, 0.0310, 0.0284, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 08:59:20,053 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:34,774 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:39,602 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:45,049 INFO [train.py:904] (2/8) Epoch 5, batch 7800, loss[loss=0.2378, simple_loss=0.322, pruned_loss=0.07678, over 16726.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3287, pruned_loss=0.09126, over 3069770.62 frames. ], batch size: 76, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:00:21,166 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:28,135 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:49,116 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:52,204 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:01:02,289 INFO [train.py:904] (2/8) Epoch 5, batch 7850, loss[loss=0.3152, simple_loss=0.3628, pruned_loss=0.1338, over 11565.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3297, pruned_loss=0.09108, over 3077214.37 frames. ], batch size: 247, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:01:13,092 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:01:23,602 INFO [optim.py:368] (2/8) 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 09:01:23,942 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:01:52,510 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:16,621 INFO [train.py:904] (2/8) Epoch 5, batch 7900, loss[loss=0.2984, simple_loss=0.3618, pruned_loss=0.1175, over 15287.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3283, pruned_loss=0.09041, over 3063443.62 frames. ], batch size: 190, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:02:35,252 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:02:48,325 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 09:02:48,405 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 09:03:25,553 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 09:03:36,607 INFO [train.py:904] (2/8) Epoch 5, batch 7950, loss[loss=0.2601, simple_loss=0.3356, pruned_loss=0.0923, over 16874.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3293, pruned_loss=0.09123, over 3067089.02 frames. ], batch size: 116, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:03:56,966 INFO [optim.py:368] (2/8) 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 09:04:02,935 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:04:50,721 INFO [train.py:904] (2/8) Epoch 5, batch 8000, loss[loss=0.2409, simple_loss=0.3172, pruned_loss=0.08226, over 16607.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.33, pruned_loss=0.09203, over 3055110.51 frames. ], batch size: 57, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:06:04,210 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8006, 2.5360, 2.5456, 1.8816, 2.5010, 2.5730, 2.4888, 1.6742], device='cuda:2'), covar=tensor([0.0278, 0.0034, 0.0042, 0.0201, 0.0055, 0.0058, 0.0043, 0.0291], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0054, 0.0060, 0.0117, 0.0061, 0.0069, 0.0064, 0.0109], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 09:06:06,080 INFO [train.py:904] (2/8) Epoch 5, batch 8050, loss[loss=0.2301, simple_loss=0.3185, pruned_loss=0.07092, over 16740.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.3302, pruned_loss=0.09202, over 3046988.85 frames. ], batch size: 89, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:06:26,963 INFO [optim.py:368] (2/8) 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 09:06:48,211 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 09:06:49,660 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:07:21,588 INFO [train.py:904] (2/8) Epoch 5, batch 8100, loss[loss=0.2359, simple_loss=0.3133, pruned_loss=0.07929, over 16990.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3291, pruned_loss=0.09071, over 3067935.92 frames. ], batch size: 55, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:05,522 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:08:40,602 INFO [train.py:904] (2/8) Epoch 5, batch 8150, loss[loss=0.2564, simple_loss=0.318, pruned_loss=0.09744, over 11526.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3261, pruned_loss=0.0888, over 3086981.25 frames. ], batch size: 248, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:44,093 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:01,946 INFO [optim.py:368] (2/8) 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 09:09:09,007 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 09:09:20,845 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:24,937 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:43,753 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-28 09:09:43,832 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 09:10:00,530 INFO [train.py:904] (2/8) Epoch 5, batch 8200, loss[loss=0.2241, simple_loss=0.3057, pruned_loss=0.07124, over 16527.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3228, pruned_loss=0.0877, over 3086147.26 frames. ], batch size: 75, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:10:18,789 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-04-28 09:10:32,541 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:10:56,550 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-28 09:11:00,868 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:11:22,105 INFO [train.py:904] (2/8) Epoch 5, batch 8250, loss[loss=0.2531, simple_loss=0.3328, pruned_loss=0.08666, over 15214.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3219, pruned_loss=0.08533, over 3075882.96 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:11:44,535 INFO [optim.py:368] (2/8) 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 09:11:50,745 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:12:13,763 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:40,605 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:43,301 INFO [train.py:904] (2/8) Epoch 5, batch 8300, loss[loss=0.2191, simple_loss=0.3114, pruned_loss=0.06342, over 16906.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3186, pruned_loss=0.08191, over 3053121.84 frames. ], batch size: 102, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:12:43,987 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2619, 1.4064, 1.8956, 2.3297, 2.3317, 2.4688, 1.5445, 2.3753], device='cuda:2'), covar=tensor([0.0076, 0.0259, 0.0141, 0.0117, 0.0104, 0.0088, 0.0230, 0.0060], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0146, 0.0130, 0.0124, 0.0132, 0.0094, 0.0142, 0.0080], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 09:12:48,305 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.24 vs. limit=5.0 2023-04-28 09:13:08,660 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:13:26,978 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-28 09:14:06,484 INFO [train.py:904] (2/8) Epoch 5, batch 8350, loss[loss=0.2421, simple_loss=0.3102, pruned_loss=0.08698, over 12216.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3169, pruned_loss=0.07942, over 3033086.30 frames. ], batch size: 247, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:14:30,518 INFO [optim.py:368] (2/8) 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 09:14:41,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3309, 4.1552, 4.7018, 4.6545, 4.6883, 4.2876, 4.3652, 4.1844], device='cuda:2'), covar=tensor([0.0213, 0.0420, 0.0373, 0.0382, 0.0361, 0.0263, 0.0752, 0.0310], device='cuda:2'), in_proj_covar=tensor([0.0227, 0.0216, 0.0226, 0.0224, 0.0272, 0.0237, 0.0338, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 09:14:54,735 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:15:30,229 INFO [train.py:904] (2/8) Epoch 5, batch 8400, loss[loss=0.1909, simple_loss=0.2826, pruned_loss=0.04957, over 16092.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3134, pruned_loss=0.0759, over 3057375.99 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:14,935 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:16:51,003 INFO [train.py:904] (2/8) Epoch 5, batch 8450, loss[loss=0.225, simple_loss=0.3125, pruned_loss=0.06874, over 15259.00 frames. ], tot_loss[loss=0.229, simple_loss=0.311, pruned_loss=0.07354, over 3067834.44 frames. ], batch size: 190, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:54,812 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:17:13,778 INFO [optim.py:368] (2/8) 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 09:17:38,437 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:13,183 INFO [train.py:904] (2/8) Epoch 5, batch 8500, loss[loss=0.2244, simple_loss=0.3009, pruned_loss=0.07394, over 15135.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3055, pruned_loss=0.06989, over 3058364.14 frames. ], batch size: 190, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:18:13,693 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:57,905 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 5, batch 8550, loss[loss=0.2249, simple_loss=0.3163, pruned_loss=0.06674, over 16230.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3023, pruned_loss=0.0683, over 3030396.98 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:20:04,109 INFO [optim.py:368] (2/8) 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 09:20:29,013 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:04,489 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:17,318 INFO [train.py:904] (2/8) Epoch 5, batch 8600, loss[loss=0.2014, simple_loss=0.2922, pruned_loss=0.0553, over 16607.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3024, pruned_loss=0.06709, over 3042056.29 frames. ], batch size: 57, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:22:28,805 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:22:58,665 INFO [train.py:904] (2/8) Epoch 5, batch 8650, loss[loss=0.1975, simple_loss=0.2868, pruned_loss=0.0541, over 15418.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.3004, pruned_loss=0.06543, over 3034860.51 frames. ], batch size: 190, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:23:33,946 INFO [optim.py:368] (2/8) 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 09:24:39,680 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:24:47,015 INFO [train.py:904] (2/8) Epoch 5, batch 8700, loss[loss=0.2043, simple_loss=0.2811, pruned_loss=0.06378, over 12550.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2969, pruned_loss=0.06358, over 3040819.60 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:26:24,150 INFO [train.py:904] (2/8) Epoch 5, batch 8750, loss[loss=0.2446, simple_loss=0.3277, pruned_loss=0.08073, over 16700.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2972, pruned_loss=0.0631, over 3055214.67 frames. ], batch size: 134, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:27:05,487 INFO [optim.py:368] (2/8) 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 09:28:17,594 INFO [train.py:904] (2/8) Epoch 5, batch 8800, loss[loss=0.2214, simple_loss=0.3094, pruned_loss=0.06672, over 16837.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.296, pruned_loss=0.06212, over 3063122.70 frames. ], batch size: 102, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:29:01,955 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:29:24,429 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:30:04,511 INFO [train.py:904] (2/8) Epoch 5, batch 8850, loss[loss=0.2025, simple_loss=0.2807, pruned_loss=0.06213, over 12430.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2977, pruned_loss=0.06137, over 3057984.80 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:30:38,665 INFO [optim.py:368] (2/8) 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 09:30:42,696 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8157, 1.3048, 1.6384, 1.7476, 1.8702, 1.8666, 1.4940, 1.8421], device='cuda:2'), covar=tensor([0.0113, 0.0199, 0.0098, 0.0153, 0.0137, 0.0091, 0.0180, 0.0048], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0143, 0.0128, 0.0124, 0.0130, 0.0091, 0.0141, 0.0077], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 09:31:02,575 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:16,994 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:37,580 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 09:31:39,633 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:54,237 INFO [train.py:904] (2/8) Epoch 5, batch 8900, loss[loss=0.1967, simple_loss=0.2815, pruned_loss=0.05589, over 12974.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2976, pruned_loss=0.06045, over 3052674.39 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:31:57,682 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6324, 5.9822, 5.6792, 5.7209, 5.3713, 5.0773, 5.5104, 6.0100], device='cuda:2'), covar=tensor([0.0559, 0.0539, 0.0860, 0.0357, 0.0470, 0.0552, 0.0628, 0.0617], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0460, 0.0388, 0.0298, 0.0289, 0.0308, 0.0373, 0.0337], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 09:32:48,144 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:39,419 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:59,796 INFO [train.py:904] (2/8) Epoch 5, batch 8950, loss[loss=0.2156, simple_loss=0.2944, pruned_loss=0.06833, over 12569.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2981, pruned_loss=0.06129, over 3057944.16 frames. ], batch size: 246, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:34:35,628 INFO [optim.py:368] (2/8) 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 09:35:10,816 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7589, 3.0768, 3.0030, 1.4790, 3.1622, 3.2409, 2.9515, 2.4938], device='cuda:2'), covar=tensor([0.1106, 0.0144, 0.0160, 0.1349, 0.0080, 0.0087, 0.0311, 0.0502], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0087, 0.0077, 0.0140, 0.0067, 0.0075, 0.0114, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 09:35:33,160 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:35:51,807 INFO [train.py:904] (2/8) Epoch 5, batch 9000, loss[loss=0.1871, simple_loss=0.2762, pruned_loss=0.04894, over 16673.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2946, pruned_loss=0.05945, over 3075363.31 frames. ], batch size: 134, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:35:51,807 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 09:36:02,096 INFO [train.py:938] (2/8) Epoch 5, validation: loss=0.1735, simple_loss=0.2766, pruned_loss=0.0352, over 944034.00 frames. 2023-04-28 09:36:02,097 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 09:37:44,738 INFO [train.py:904] (2/8) Epoch 5, batch 9050, loss[loss=0.2155, simple_loss=0.2989, pruned_loss=0.06611, over 16324.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2965, pruned_loss=0.0608, over 3079633.30 frames. ], batch size: 146, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:38:18,679 INFO [optim.py:368] (2/8) 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 09:38:47,041 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6055, 2.0009, 2.1826, 3.9940, 1.7391, 2.7425, 2.0730, 1.9575], device='cuda:2'), covar=tensor([0.0610, 0.2683, 0.1434, 0.0380, 0.4140, 0.1428, 0.2524, 0.3241], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0306, 0.0257, 0.0298, 0.0362, 0.0305, 0.0284, 0.0361], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 09:39:29,263 INFO [train.py:904] (2/8) Epoch 5, batch 9100, loss[loss=0.2185, simple_loss=0.3077, pruned_loss=0.06467, over 16846.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2964, pruned_loss=0.06134, over 3075786.41 frames. ], batch size: 116, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:41:26,340 INFO [train.py:904] (2/8) Epoch 5, batch 9150, loss[loss=0.202, simple_loss=0.282, pruned_loss=0.06105, over 12356.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2964, pruned_loss=0.06071, over 3070643.26 frames. ], batch size: 250, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:41:36,332 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 09:42:00,609 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:01,427 INFO [optim.py:368] (2/8) 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 09:42:25,693 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:36,769 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9977, 2.7967, 2.7084, 1.7205, 2.9316, 2.9390, 2.5714, 2.4383], device='cuda:2'), covar=tensor([0.0665, 0.0139, 0.0171, 0.1023, 0.0075, 0.0084, 0.0343, 0.0408], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0086, 0.0076, 0.0137, 0.0066, 0.0073, 0.0112, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 09:42:47,420 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:43:09,755 INFO [train.py:904] (2/8) Epoch 5, batch 9200, loss[loss=0.2162, simple_loss=0.2992, pruned_loss=0.06659, over 15231.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.291, pruned_loss=0.05935, over 3062021.42 frames. ], batch size: 190, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:43:58,064 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:44:46,686 INFO [train.py:904] (2/8) Epoch 5, batch 9250, loss[loss=0.1826, simple_loss=0.2578, pruned_loss=0.05366, over 11990.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2902, pruned_loss=0.05889, over 3075087.19 frames. ], batch size: 248, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:45:18,929 INFO [optim.py:368] (2/8) 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 09:46:21,461 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:46:32,019 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5050, 3.6963, 1.5392, 3.8232, 2.5733, 3.7766, 2.0261, 2.9185], device='cuda:2'), covar=tensor([0.0165, 0.0237, 0.1792, 0.0058, 0.0740, 0.0344, 0.1407, 0.0529], device='cuda:2'), in_proj_covar=tensor([0.0114, 0.0141, 0.0171, 0.0074, 0.0151, 0.0166, 0.0180, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 09:46:40,050 INFO [train.py:904] (2/8) Epoch 5, batch 9300, loss[loss=0.2004, simple_loss=0.2804, pruned_loss=0.06023, over 16295.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2879, pruned_loss=0.05792, over 3065443.93 frames. ], batch size: 165, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:48:05,860 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:48:26,420 INFO [train.py:904] (2/8) Epoch 5, batch 9350, loss[loss=0.2245, simple_loss=0.3101, pruned_loss=0.06951, over 16259.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2882, pruned_loss=0.05798, over 3063907.54 frames. ], batch size: 165, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:49:00,529 INFO [optim.py:368] (2/8) 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 09:49:19,005 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6480, 3.5426, 4.0838, 4.0146, 4.0414, 3.6807, 3.7900, 3.7338], device='cuda:2'), covar=tensor([0.0250, 0.0528, 0.0283, 0.0460, 0.0391, 0.0382, 0.0677, 0.0350], device='cuda:2'), in_proj_covar=tensor([0.0213, 0.0209, 0.0214, 0.0215, 0.0261, 0.0230, 0.0312, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-28 09:50:03,273 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9146, 4.2027, 4.0135, 4.0598, 3.6923, 3.7458, 3.9112, 4.1608], device='cuda:2'), covar=tensor([0.0685, 0.0788, 0.0745, 0.0474, 0.0621, 0.1377, 0.0625, 0.0795], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0465, 0.0381, 0.0300, 0.0292, 0.0303, 0.0375, 0.0334], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 09:50:11,380 INFO [train.py:904] (2/8) Epoch 5, batch 9400, loss[loss=0.1947, simple_loss=0.2956, pruned_loss=0.04687, over 16829.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.288, pruned_loss=0.05791, over 3052203.30 frames. ], batch size: 124, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:50:14,149 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1361, 1.3350, 1.6859, 2.0089, 2.1258, 2.2030, 1.5468, 2.0070], device='cuda:2'), covar=tensor([0.0087, 0.0220, 0.0136, 0.0135, 0.0110, 0.0085, 0.0215, 0.0052], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0140, 0.0126, 0.0122, 0.0126, 0.0088, 0.0138, 0.0075], device='cuda:2'), out_proj_covar=tensor([1.6271e-04, 1.9134e-04, 1.7591e-04, 1.6949e-04, 1.7254e-04, 1.1581e-04, 1.8890e-04, 9.9974e-05], device='cuda:2') 2023-04-28 09:50:19,900 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1197, 4.0711, 4.1990, 4.1365, 4.2286, 4.6888, 4.3992, 4.0830], device='cuda:2'), covar=tensor([0.1340, 0.1887, 0.1433, 0.2190, 0.2574, 0.1070, 0.1204, 0.2393], device='cuda:2'), in_proj_covar=tensor([0.0253, 0.0357, 0.0354, 0.0312, 0.0412, 0.0381, 0.0298, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 09:51:39,796 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-28 09:51:46,449 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2712, 4.0080, 4.3314, 4.4356, 4.5834, 4.0769, 4.5758, 4.5665], device='cuda:2'), covar=tensor([0.0742, 0.0766, 0.0835, 0.0460, 0.0357, 0.0754, 0.0377, 0.0331], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0430, 0.0535, 0.0444, 0.0335, 0.0328, 0.0352, 0.0368], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 09:51:51,664 INFO [train.py:904] (2/8) Epoch 5, batch 9450, loss[loss=0.1858, simple_loss=0.2799, pruned_loss=0.04581, over 16624.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2899, pruned_loss=0.05822, over 3056998.55 frames. ], batch size: 62, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:52:21,841 INFO [optim.py:368] (2/8) 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 09:52:25,871 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-28 09:52:47,323 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:53:09,009 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:53:18,583 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1173, 3.7935, 3.4134, 1.8565, 2.9623, 2.3695, 3.3811, 3.5742], device='cuda:2'), covar=tensor([0.0250, 0.0525, 0.0442, 0.1573, 0.0700, 0.0925, 0.0650, 0.0621], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0117, 0.0151, 0.0138, 0.0130, 0.0127, 0.0136, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 09:53:20,148 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-28 09:53:33,285 INFO [train.py:904] (2/8) Epoch 5, batch 9500, loss[loss=0.2092, simple_loss=0.2996, pruned_loss=0.05944, over 15343.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2893, pruned_loss=0.05751, over 3052129.70 frames. ], batch size: 191, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:54:17,580 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:20,764 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 09:54:25,575 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:29,291 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:55:18,505 INFO [train.py:904] (2/8) Epoch 5, batch 9550, loss[loss=0.1989, simple_loss=0.2798, pruned_loss=0.05906, over 12355.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2894, pruned_loss=0.05797, over 3038488.60 frames. ], batch size: 247, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:55:53,349 INFO [optim.py:368] (2/8) 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 09:56:29,037 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 09:56:38,128 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:56:52,900 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:56:54,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6786, 4.4437, 4.7224, 4.9209, 5.0760, 4.3405, 5.0835, 4.9933], device='cuda:2'), covar=tensor([0.0852, 0.0721, 0.1001, 0.0455, 0.0357, 0.0690, 0.0360, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0428, 0.0535, 0.0442, 0.0331, 0.0328, 0.0348, 0.0366], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 09:56:59,037 INFO [train.py:904] (2/8) Epoch 5, batch 9600, loss[loss=0.216, simple_loss=0.3094, pruned_loss=0.06128, over 16318.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2914, pruned_loss=0.05863, over 3047077.59 frames. ], batch size: 146, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:58:47,432 INFO [train.py:904] (2/8) Epoch 5, batch 9650, loss[loss=0.2059, simple_loss=0.2909, pruned_loss=0.06042, over 12367.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2936, pruned_loss=0.05903, over 3067729.38 frames. ], batch size: 246, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:59:05,788 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:59:27,461 INFO [optim.py:368] (2/8) 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 10:00:35,786 INFO [train.py:904] (2/8) Epoch 5, batch 9700, loss[loss=0.1898, simple_loss=0.2803, pruned_loss=0.04968, over 16921.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.292, pruned_loss=0.05846, over 3070705.03 frames. ], batch size: 116, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:00:59,074 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6895, 4.6373, 4.5206, 4.0161, 4.4620, 1.6400, 4.3318, 4.4487], device='cuda:2'), covar=tensor([0.0054, 0.0047, 0.0079, 0.0187, 0.0054, 0.1654, 0.0069, 0.0105], device='cuda:2'), in_proj_covar=tensor([0.0087, 0.0075, 0.0115, 0.0108, 0.0087, 0.0141, 0.0099, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:01:37,711 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0879, 2.9731, 2.8110, 2.0046, 2.5864, 2.1665, 2.7222, 2.8275], device='cuda:2'), covar=tensor([0.0295, 0.0557, 0.0423, 0.1291, 0.0639, 0.0843, 0.0596, 0.0516], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0116, 0.0150, 0.0139, 0.0130, 0.0126, 0.0135, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 10:01:37,755 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9237, 2.1008, 2.3243, 3.2423, 2.0818, 2.5645, 2.3717, 2.0480], device='cuda:2'), covar=tensor([0.0486, 0.1903, 0.0949, 0.0331, 0.2755, 0.1190, 0.1724, 0.2390], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0306, 0.0259, 0.0301, 0.0362, 0.0306, 0.0285, 0.0361], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:01:48,327 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 10:01:51,115 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9949, 4.4931, 3.5802, 2.6595, 3.1541, 2.6401, 4.6332, 4.2422], device='cuda:2'), covar=tensor([0.2000, 0.0444, 0.1063, 0.1403, 0.1651, 0.1328, 0.0287, 0.0540], device='cuda:2'), in_proj_covar=tensor([0.0268, 0.0234, 0.0258, 0.0232, 0.0234, 0.0192, 0.0230, 0.0227], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:02:09,984 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3132, 4.2798, 4.1823, 4.0240, 3.7291, 4.2433, 4.1403, 3.9478], device='cuda:2'), covar=tensor([0.0414, 0.0358, 0.0213, 0.0177, 0.0806, 0.0281, 0.0329, 0.0473], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0177, 0.0200, 0.0170, 0.0221, 0.0202, 0.0140, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:02:18,557 INFO [train.py:904] (2/8) Epoch 5, batch 9750, loss[loss=0.2092, simple_loss=0.3001, pruned_loss=0.05917, over 16791.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2909, pruned_loss=0.05879, over 3077882.09 frames. ], batch size: 124, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:02:50,578 INFO [optim.py:368] (2/8) 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 10:03:36,811 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1583, 5.1148, 4.8621, 4.4802, 4.8383, 1.6467, 4.6546, 4.7530], device='cuda:2'), covar=tensor([0.0035, 0.0029, 0.0056, 0.0140, 0.0039, 0.1756, 0.0060, 0.0094], device='cuda:2'), in_proj_covar=tensor([0.0087, 0.0075, 0.0115, 0.0108, 0.0087, 0.0141, 0.0099, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:03:56,536 INFO [train.py:904] (2/8) Epoch 5, batch 9800, loss[loss=0.1892, simple_loss=0.2929, pruned_loss=0.04275, over 16243.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2909, pruned_loss=0.05727, over 3084809.85 frames. ], batch size: 165, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:04:36,827 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:05:41,978 INFO [train.py:904] (2/8) Epoch 5, batch 9850, loss[loss=0.197, simple_loss=0.2739, pruned_loss=0.06003, over 12512.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2916, pruned_loss=0.0567, over 3083974.50 frames. ], batch size: 247, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:06:14,680 INFO [optim.py:368] (2/8) 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 10:06:21,304 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:06:35,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7721, 3.7242, 3.9057, 3.7657, 3.8185, 4.1911, 3.8756, 3.6434], device='cuda:2'), covar=tensor([0.1634, 0.1766, 0.1275, 0.1840, 0.2288, 0.1196, 0.1175, 0.2350], device='cuda:2'), in_proj_covar=tensor([0.0250, 0.0360, 0.0350, 0.0307, 0.0412, 0.0382, 0.0288, 0.0411], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:06:52,957 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:07:31,485 INFO [train.py:904] (2/8) Epoch 5, batch 9900, loss[loss=0.1952, simple_loss=0.294, pruned_loss=0.0482, over 16739.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2917, pruned_loss=0.05667, over 3075978.53 frames. ], batch size: 83, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:08:07,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6435, 1.3511, 2.0317, 2.6081, 2.3112, 2.7662, 1.3077, 2.4809], device='cuda:2'), covar=tensor([0.0066, 0.0285, 0.0154, 0.0110, 0.0126, 0.0079, 0.0322, 0.0071], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0146, 0.0130, 0.0124, 0.0130, 0.0091, 0.0143, 0.0079], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 10:09:17,007 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:09:17,326 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 10:09:27,672 INFO [train.py:904] (2/8) Epoch 5, batch 9950, loss[loss=0.2019, simple_loss=0.2934, pruned_loss=0.05515, over 16306.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2936, pruned_loss=0.05713, over 3074298.66 frames. ], batch size: 165, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:09:33,680 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 10:10:04,528 INFO [optim.py:368] (2/8) 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 10:10:35,311 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8560, 4.7940, 4.7003, 4.4510, 4.2618, 4.7403, 4.7220, 4.4092], device='cuda:2'), covar=tensor([0.0331, 0.0264, 0.0164, 0.0147, 0.0626, 0.0237, 0.0186, 0.0399], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0174, 0.0197, 0.0166, 0.0218, 0.0197, 0.0137, 0.0221], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:11:29,353 INFO [train.py:904] (2/8) Epoch 5, batch 10000, loss[loss=0.1831, simple_loss=0.2795, pruned_loss=0.04337, over 15459.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2918, pruned_loss=0.05636, over 3098855.40 frames. ], batch size: 191, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:11:43,844 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:13:08,523 INFO [train.py:904] (2/8) Epoch 5, batch 10050, loss[loss=0.2316, simple_loss=0.3201, pruned_loss=0.07152, over 12536.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2914, pruned_loss=0.05619, over 3081099.24 frames. ], batch size: 248, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:13:38,880 INFO [optim.py:368] (2/8) 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 10:14:28,154 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6056, 4.5969, 5.1339, 5.0179, 5.0610, 4.7349, 4.6567, 4.5350], device='cuda:2'), covar=tensor([0.0204, 0.0387, 0.0215, 0.0392, 0.0271, 0.0233, 0.0664, 0.0313], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0218, 0.0212, 0.0218, 0.0259, 0.0228, 0.0319, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-04-28 10:14:38,998 INFO [train.py:904] (2/8) Epoch 5, batch 10100, loss[loss=0.2079, simple_loss=0.2884, pruned_loss=0.06373, over 16358.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2921, pruned_loss=0.05705, over 3083262.72 frames. ], batch size: 146, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:15:44,009 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 10:16:20,268 INFO [train.py:904] (2/8) Epoch 6, batch 0, loss[loss=0.2326, simple_loss=0.3108, pruned_loss=0.07716, over 17178.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3108, pruned_loss=0.07716, over 17178.00 frames. ], batch size: 46, lr: 1.19e-02, grad_scale: 8.0 2023-04-28 10:16:20,269 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 10:16:27,640 INFO [train.py:938] (2/8) Epoch 6, validation: loss=0.1727, simple_loss=0.2755, pruned_loss=0.03501, over 944034.00 frames. 2023-04-28 10:16:27,641 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 10:16:50,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3706, 5.1559, 5.1962, 5.0304, 4.7314, 5.2621, 5.1387, 4.8194], device='cuda:2'), covar=tensor([0.0442, 0.0303, 0.0177, 0.0130, 0.0796, 0.0243, 0.0163, 0.0440], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0175, 0.0196, 0.0167, 0.0221, 0.0200, 0.0138, 0.0224], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:16:52,395 INFO [optim.py:368] (2/8) 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 10:17:10,809 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:13,002 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:19,698 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7237, 3.1240, 2.7229, 4.4343, 4.0561, 4.2268, 1.6015, 3.0396], device='cuda:2'), covar=tensor([0.1312, 0.0470, 0.0971, 0.0085, 0.0264, 0.0337, 0.1324, 0.0700], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0140, 0.0166, 0.0086, 0.0151, 0.0174, 0.0161, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 10:17:35,095 INFO [train.py:904] (2/8) Epoch 6, batch 50, loss[loss=0.232, simple_loss=0.305, pruned_loss=0.07944, over 16447.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3162, pruned_loss=0.08659, over 756153.49 frames. ], batch size: 68, lr: 1.19e-02, grad_scale: 2.0 2023-04-28 10:17:35,888 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 10:17:46,416 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:18:20,193 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:23,655 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8096, 4.7610, 4.5688, 4.0373, 4.5407, 1.7531, 4.3802, 4.5067], device='cuda:2'), covar=tensor([0.0070, 0.0059, 0.0109, 0.0253, 0.0077, 0.1912, 0.0097, 0.0157], device='cuda:2'), in_proj_covar=tensor([0.0090, 0.0078, 0.0122, 0.0113, 0.0090, 0.0145, 0.0104, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:18:31,730 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5761, 4.4448, 4.4387, 4.2584, 3.9508, 4.4521, 4.3822, 4.1252], device='cuda:2'), covar=tensor([0.0536, 0.0363, 0.0209, 0.0194, 0.0882, 0.0343, 0.0335, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0183, 0.0206, 0.0175, 0.0233, 0.0209, 0.0144, 0.0235], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:18:34,245 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:47,052 INFO [train.py:904] (2/8) Epoch 6, batch 100, loss[loss=0.2553, simple_loss=0.3156, pruned_loss=0.09752, over 16793.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3082, pruned_loss=0.08113, over 1325377.40 frames. ], batch size: 124, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:18:49,631 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:19:11,790 INFO [optim.py:368] (2/8) 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 10:19:12,264 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:19:25,272 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 10:19:54,850 INFO [train.py:904] (2/8) Epoch 6, batch 150, loss[loss=0.2055, simple_loss=0.287, pruned_loss=0.06199, over 17240.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3036, pruned_loss=0.0766, over 1764203.76 frames. ], batch size: 45, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:19:55,129 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:19:56,133 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:20:57,612 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:21:05,279 INFO [train.py:904] (2/8) Epoch 6, batch 200, loss[loss=0.2553, simple_loss=0.3002, pruned_loss=0.1052, over 16787.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3027, pruned_loss=0.07619, over 2114991.74 frames. ], batch size: 83, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:21:28,625 INFO [optim.py:368] (2/8) 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 10:22:12,665 INFO [train.py:904] (2/8) Epoch 6, batch 250, loss[loss=0.1975, simple_loss=0.2816, pruned_loss=0.0567, over 17058.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.2993, pruned_loss=0.07468, over 2388174.97 frames. ], batch size: 53, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:22:20,790 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:22:24,912 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 10:22:35,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3021, 1.9288, 2.4930, 3.1158, 2.9465, 3.2461, 2.0246, 3.4107], device='cuda:2'), covar=tensor([0.0076, 0.0216, 0.0131, 0.0126, 0.0111, 0.0106, 0.0210, 0.0066], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0146, 0.0131, 0.0127, 0.0133, 0.0095, 0.0144, 0.0080], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 10:22:37,415 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 10:22:59,088 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 10:23:20,604 INFO [train.py:904] (2/8) Epoch 6, batch 300, loss[loss=0.1903, simple_loss=0.2789, pruned_loss=0.05082, over 17224.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2958, pruned_loss=0.07349, over 2594154.29 frames. ], batch size: 45, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:23:37,441 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:23:45,701 INFO [optim.py:368] (2/8) 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 10:24:15,216 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 10:24:30,407 INFO [train.py:904] (2/8) Epoch 6, batch 350, loss[loss=0.2115, simple_loss=0.3019, pruned_loss=0.06058, over 17259.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2926, pruned_loss=0.07164, over 2755856.14 frames. ], batch size: 52, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:25:01,813 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:19,515 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:37,101 INFO [train.py:904] (2/8) Epoch 6, batch 400, loss[loss=0.1727, simple_loss=0.254, pruned_loss=0.04566, over 16879.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2903, pruned_loss=0.07044, over 2887449.71 frames. ], batch size: 42, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:25:55,420 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:26:01,707 INFO [optim.py:368] (2/8) 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 10:26:02,447 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-04-28 10:26:05,630 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:26:20,572 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8746, 4.3313, 3.1670, 2.3606, 3.0138, 2.3184, 4.4558, 4.2672], device='cuda:2'), covar=tensor([0.2228, 0.0519, 0.1297, 0.1801, 0.2238, 0.1627, 0.0359, 0.0627], device='cuda:2'), in_proj_covar=tensor([0.0278, 0.0250, 0.0271, 0.0246, 0.0271, 0.0203, 0.0244, 0.0256], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:26:45,067 INFO [train.py:904] (2/8) Epoch 6, batch 450, loss[loss=0.168, simple_loss=0.2596, pruned_loss=0.03822, over 16940.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2886, pruned_loss=0.06897, over 2989753.26 frames. ], batch size: 41, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:26:46,463 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:28,559 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-04-28 10:27:29,538 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:27:52,965 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:53,834 INFO [train.py:904] (2/8) Epoch 6, batch 500, loss[loss=0.1807, simple_loss=0.2606, pruned_loss=0.05035, over 16837.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2863, pruned_loss=0.067, over 3074186.73 frames. ], batch size: 42, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:28:15,996 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-04-28 10:28:17,360 INFO [optim.py:368] (2/8) 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 10:29:01,773 INFO [train.py:904] (2/8) Epoch 6, batch 550, loss[loss=0.1909, simple_loss=0.2797, pruned_loss=0.05102, over 17198.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2853, pruned_loss=0.06666, over 3123318.47 frames. ], batch size: 45, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:29:03,264 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:29:03,893 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-28 10:30:12,293 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 10:30:13,391 INFO [train.py:904] (2/8) Epoch 6, batch 600, loss[loss=0.1927, simple_loss=0.2886, pruned_loss=0.04842, over 16995.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2841, pruned_loss=0.0662, over 3163155.61 frames. ], batch size: 50, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:30:30,465 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4024, 5.2298, 5.1894, 4.8949, 4.6894, 5.1646, 5.2054, 4.8155], device='cuda:2'), covar=tensor([0.0446, 0.0287, 0.0188, 0.0185, 0.0911, 0.0319, 0.0157, 0.0521], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0209, 0.0229, 0.0199, 0.0262, 0.0236, 0.0162, 0.0265], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 10:30:38,609 INFO [optim.py:368] (2/8) 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 10:30:55,680 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:10,945 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:23,784 INFO [train.py:904] (2/8) Epoch 6, batch 650, loss[loss=0.2078, simple_loss=0.2688, pruned_loss=0.07338, over 12017.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2821, pruned_loss=0.06636, over 3195893.84 frames. ], batch size: 248, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:31:47,274 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:55,823 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 10:32:14,546 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:21,258 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:28,222 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2093, 2.2114, 2.3500, 4.7563, 1.9326, 3.1183, 2.3887, 2.3759], device='cuda:2'), covar=tensor([0.0502, 0.2529, 0.1294, 0.0259, 0.3504, 0.1472, 0.2129, 0.3129], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0327, 0.0272, 0.0318, 0.0374, 0.0338, 0.0300, 0.0396], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:32:31,562 INFO [train.py:904] (2/8) Epoch 6, batch 700, loss[loss=0.212, simple_loss=0.2916, pruned_loss=0.06623, over 17119.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2825, pruned_loss=0.06681, over 3219012.03 frames. ], batch size: 47, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:32:35,233 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:49,156 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:32:57,145 INFO [optim.py:368] (2/8) 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 10:33:19,514 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:27,636 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8905, 4.8951, 5.4720, 5.4801, 5.4811, 5.0785, 5.0816, 4.6937], device='cuda:2'), covar=tensor([0.0254, 0.0347, 0.0310, 0.0385, 0.0367, 0.0232, 0.0732, 0.0353], device='cuda:2'), in_proj_covar=tensor([0.0262, 0.0263, 0.0259, 0.0258, 0.0309, 0.0274, 0.0391, 0.0230], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 10:33:32,371 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:39,717 INFO [train.py:904] (2/8) Epoch 6, batch 750, loss[loss=0.2286, simple_loss=0.2871, pruned_loss=0.08503, over 16837.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2829, pruned_loss=0.06693, over 3241756.88 frames. ], batch size: 96, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:33:56,282 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:34:10,489 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6669, 4.5300, 4.5197, 4.3458, 4.1048, 4.5383, 4.4057, 4.2550], device='cuda:2'), covar=tensor([0.0441, 0.0327, 0.0203, 0.0189, 0.0731, 0.0290, 0.0401, 0.0457], device='cuda:2'), in_proj_covar=tensor([0.0203, 0.0215, 0.0236, 0.0204, 0.0269, 0.0242, 0.0169, 0.0271], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 10:34:14,212 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 10:34:19,279 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 10:34:53,034 INFO [train.py:904] (2/8) Epoch 6, batch 800, loss[loss=0.2079, simple_loss=0.2944, pruned_loss=0.06069, over 17142.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2828, pruned_loss=0.06603, over 3263558.78 frames. ], batch size: 48, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:34:59,940 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 10:35:26,914 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3902, 4.0894, 3.7287, 1.9852, 2.9029, 2.1831, 3.6041, 3.7746], device='cuda:2'), covar=tensor([0.0248, 0.0485, 0.0483, 0.1605, 0.0716, 0.0978, 0.0658, 0.0872], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0129, 0.0153, 0.0141, 0.0130, 0.0125, 0.0137, 0.0139], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 10:36:01,905 INFO [train.py:904] (2/8) Epoch 6, batch 850, loss[loss=0.2061, simple_loss=0.2893, pruned_loss=0.06147, over 16453.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2828, pruned_loss=0.06606, over 3265680.18 frames. ], batch size: 68, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:36:04,226 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:37:10,632 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:37:11,528 INFO [train.py:904] (2/8) Epoch 6, batch 900, loss[loss=0.2106, simple_loss=0.28, pruned_loss=0.07056, over 16496.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2813, pruned_loss=0.06407, over 3288737.30 frames. ], batch size: 146, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:37:39,516 INFO [optim.py:368] (2/8) 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 10:37:55,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5705, 5.9988, 5.7234, 5.8985, 5.2720, 5.0672, 5.5043, 6.1334], device='cuda:2'), covar=tensor([0.0734, 0.0724, 0.0986, 0.0454, 0.0633, 0.0569, 0.0710, 0.0719], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0550, 0.0447, 0.0350, 0.0334, 0.0344, 0.0446, 0.0384], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:38:07,336 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6171, 6.0336, 5.6960, 5.9476, 5.3367, 5.0256, 5.4997, 6.1183], device='cuda:2'), covar=tensor([0.0776, 0.0724, 0.0941, 0.0420, 0.0660, 0.0561, 0.0647, 0.0721], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0550, 0.0446, 0.0350, 0.0334, 0.0344, 0.0445, 0.0384], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:38:22,587 INFO [train.py:904] (2/8) Epoch 6, batch 950, loss[loss=0.1588, simple_loss=0.2405, pruned_loss=0.03855, over 16962.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2814, pruned_loss=0.0644, over 3293183.85 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:38:40,667 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1700, 4.9082, 5.1324, 5.3724, 5.5633, 4.7872, 5.5107, 5.4890], device='cuda:2'), covar=tensor([0.1053, 0.0788, 0.1269, 0.0483, 0.0359, 0.0555, 0.0381, 0.0368], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0536, 0.0687, 0.0545, 0.0411, 0.0401, 0.0431, 0.0462], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:38:46,224 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:12,999 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:25,703 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:26,802 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 6, batch 1000, loss[loss=0.2225, simple_loss=0.2776, pruned_loss=0.0837, over 16769.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.281, pruned_loss=0.06485, over 3293300.20 frames. ], batch size: 124, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:39:51,034 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:57,331 INFO [optim.py:368] (2/8) 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 10:40:39,945 INFO [train.py:904] (2/8) Epoch 6, batch 1050, loss[loss=0.2124, simple_loss=0.3008, pruned_loss=0.06202, over 17062.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2805, pruned_loss=0.06402, over 3304325.30 frames. ], batch size: 53, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:40:49,353 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:41:17,915 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:41:18,037 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9939, 4.1231, 4.3620, 3.2959, 4.0688, 4.3949, 4.1757, 2.7579], device='cuda:2'), covar=tensor([0.0266, 0.0034, 0.0024, 0.0181, 0.0036, 0.0033, 0.0029, 0.0248], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0063, 0.0060, 0.0115, 0.0062, 0.0070, 0.0064, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 10:41:49,451 INFO [train.py:904] (2/8) Epoch 6, batch 1100, loss[loss=0.2085, simple_loss=0.28, pruned_loss=0.06845, over 16720.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2796, pruned_loss=0.06321, over 3309780.40 frames. ], batch size: 89, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:41:49,769 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:42:16,638 INFO [optim.py:368] (2/8) 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 10:42:24,556 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:42:59,413 INFO [train.py:904] (2/8) Epoch 6, batch 1150, loss[loss=0.214, simple_loss=0.2796, pruned_loss=0.0742, over 16877.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2795, pruned_loss=0.06283, over 3314234.77 frames. ], batch size: 96, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:44:08,016 INFO [train.py:904] (2/8) Epoch 6, batch 1200, loss[loss=0.1733, simple_loss=0.257, pruned_loss=0.04475, over 16750.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.279, pruned_loss=0.0629, over 3308999.54 frames. ], batch size: 39, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:44:33,629 INFO [optim.py:368] (2/8) 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 10:44:55,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3958, 5.8140, 5.4948, 5.6726, 5.1117, 4.9763, 5.3447, 5.8967], device='cuda:2'), covar=tensor([0.0864, 0.0843, 0.1065, 0.0502, 0.0758, 0.0679, 0.0772, 0.0839], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0550, 0.0453, 0.0353, 0.0337, 0.0347, 0.0448, 0.0389], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:45:20,379 INFO [train.py:904] (2/8) Epoch 6, batch 1250, loss[loss=0.195, simple_loss=0.2886, pruned_loss=0.05068, over 17115.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2791, pruned_loss=0.06293, over 3320461.39 frames. ], batch size: 49, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:45:46,894 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6080, 5.9934, 5.7017, 5.8510, 5.2439, 5.0073, 5.5408, 6.0649], device='cuda:2'), covar=tensor([0.0834, 0.0685, 0.0905, 0.0465, 0.0661, 0.0552, 0.0683, 0.0657], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0546, 0.0448, 0.0349, 0.0334, 0.0342, 0.0444, 0.0385], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:46:13,384 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:15,813 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:26,672 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:30,469 INFO [train.py:904] (2/8) Epoch 6, batch 1300, loss[loss=0.2165, simple_loss=0.2947, pruned_loss=0.06912, over 17032.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2789, pruned_loss=0.06313, over 3321129.35 frames. ], batch size: 55, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:46:38,446 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 10:46:56,005 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6327, 2.1592, 2.3361, 4.3114, 2.0992, 2.9682, 2.2439, 2.3804], device='cuda:2'), covar=tensor([0.0634, 0.2297, 0.1241, 0.0303, 0.2795, 0.1206, 0.2087, 0.2132], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0328, 0.0274, 0.0320, 0.0374, 0.0347, 0.0302, 0.0402], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:46:58,327 INFO [optim.py:368] (2/8) 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 10:47:20,306 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:33,486 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52096.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:41,500 INFO [train.py:904] (2/8) Epoch 6, batch 1350, loss[loss=0.2198, simple_loss=0.2937, pruned_loss=0.0729, over 11982.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2789, pruned_loss=0.06338, over 3309982.58 frames. ], batch size: 247, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:47:42,025 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:44,072 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:48:51,202 INFO [train.py:904] (2/8) Epoch 6, batch 1400, loss[loss=0.1599, simple_loss=0.2482, pruned_loss=0.03583, over 17218.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2781, pruned_loss=0.06311, over 3308726.86 frames. ], batch size: 43, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:48:51,458 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:48:57,887 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0989, 4.8389, 5.0439, 5.3354, 5.5551, 4.7005, 5.4440, 5.4595], device='cuda:2'), covar=tensor([0.1153, 0.0936, 0.1645, 0.0595, 0.0419, 0.0543, 0.0410, 0.0477], device='cuda:2'), in_proj_covar=tensor([0.0442, 0.0544, 0.0696, 0.0557, 0.0417, 0.0410, 0.0434, 0.0469], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:49:19,190 INFO [optim.py:368] (2/8) 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 10:49:49,837 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 6, batch 1450, loss[loss=0.1953, simple_loss=0.2788, pruned_loss=0.05587, over 17097.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.277, pruned_loss=0.06299, over 3309473.94 frames. ], batch size: 53, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:50:33,788 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2841, 3.4448, 1.9949, 3.5907, 2.5381, 3.5449, 1.8759, 2.6821], device='cuda:2'), covar=tensor([0.0180, 0.0312, 0.1323, 0.0138, 0.0704, 0.0453, 0.1248, 0.0600], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0161, 0.0180, 0.0086, 0.0160, 0.0193, 0.0186, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 10:51:10,830 INFO [train.py:904] (2/8) Epoch 6, batch 1500, loss[loss=0.222, simple_loss=0.2827, pruned_loss=0.08067, over 16779.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.277, pruned_loss=0.06334, over 3308827.74 frames. ], batch size: 76, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:51:15,480 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:51:38,533 INFO [optim.py:368] (2/8) 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 10:52:18,640 INFO [train.py:904] (2/8) Epoch 6, batch 1550, loss[loss=0.2388, simple_loss=0.2964, pruned_loss=0.09061, over 16863.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.279, pruned_loss=0.06432, over 3312694.83 frames. ], batch size: 96, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:52:33,936 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:52:38,171 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-28 10:53:28,025 INFO [train.py:904] (2/8) Epoch 6, batch 1600, loss[loss=0.1695, simple_loss=0.2548, pruned_loss=0.04212, over 17235.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2805, pruned_loss=0.06505, over 3312019.92 frames. ], batch size: 44, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:53:55,823 INFO [optim.py:368] (2/8) 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 10:53:59,276 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:07,078 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 10:54:30,464 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:37,118 INFO [train.py:904] (2/8) Epoch 6, batch 1650, loss[loss=0.2368, simple_loss=0.3035, pruned_loss=0.08511, over 16868.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2827, pruned_loss=0.06596, over 3317897.02 frames. ], batch size: 116, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:54:40,197 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:53,245 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3977, 4.4337, 4.8716, 4.8744, 4.8832, 4.4862, 4.5224, 4.3249], device='cuda:2'), covar=tensor([0.0280, 0.0401, 0.0350, 0.0398, 0.0344, 0.0249, 0.0755, 0.0406], device='cuda:2'), in_proj_covar=tensor([0.0265, 0.0267, 0.0264, 0.0262, 0.0314, 0.0276, 0.0388, 0.0230], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 10:55:45,565 INFO [train.py:904] (2/8) Epoch 6, batch 1700, loss[loss=0.1991, simple_loss=0.2927, pruned_loss=0.05274, over 17217.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2855, pruned_loss=0.06751, over 3309173.83 frames. ], batch size: 52, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:55:45,910 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:56:14,196 INFO [optim.py:368] (2/8) 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 10:56:57,691 INFO [train.py:904] (2/8) Epoch 6, batch 1750, loss[loss=0.2479, simple_loss=0.3256, pruned_loss=0.08513, over 16729.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2865, pruned_loss=0.06761, over 3307038.69 frames. ], batch size: 57, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:05,490 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 6, batch 1800, loss[loss=0.2113, simple_loss=0.2824, pruned_loss=0.07012, over 16894.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2871, pruned_loss=0.06748, over 3304457.89 frames. ], batch size: 96, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:14,091 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:58:27,969 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6649, 4.4248, 4.6543, 4.9055, 5.0270, 4.4697, 4.9428, 4.9555], device='cuda:2'), covar=tensor([0.1000, 0.0877, 0.1255, 0.0478, 0.0391, 0.0710, 0.0613, 0.0422], device='cuda:2'), in_proj_covar=tensor([0.0443, 0.0548, 0.0695, 0.0560, 0.0416, 0.0406, 0.0435, 0.0467], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:58:36,304 INFO [optim.py:368] (2/8) 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 10:58:55,234 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8271, 2.1804, 2.3007, 4.5959, 1.9673, 3.2545, 2.3528, 2.4441], device='cuda:2'), covar=tensor([0.0581, 0.2439, 0.1276, 0.0261, 0.3097, 0.1257, 0.2092, 0.2730], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0331, 0.0274, 0.0318, 0.0374, 0.0348, 0.0301, 0.0403], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 10:58:58,314 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8248, 5.0164, 4.7568, 4.6946, 3.8500, 4.9375, 4.9081, 4.4489], device='cuda:2'), covar=tensor([0.0687, 0.0406, 0.0364, 0.0257, 0.1714, 0.0363, 0.0343, 0.0521], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0221, 0.0243, 0.0215, 0.0277, 0.0248, 0.0173, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 10:59:17,730 INFO [train.py:904] (2/8) Epoch 6, batch 1850, loss[loss=0.2026, simple_loss=0.2887, pruned_loss=0.05823, over 17099.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2877, pruned_loss=0.0672, over 3312544.64 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 10:59:38,370 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 10:59:39,363 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7444, 4.8585, 4.6748, 4.5329, 3.9195, 4.7445, 4.7480, 4.4301], device='cuda:2'), covar=tensor([0.0597, 0.0351, 0.0271, 0.0313, 0.1095, 0.0362, 0.0375, 0.0536], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0222, 0.0243, 0.0215, 0.0278, 0.0248, 0.0173, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 10:59:54,410 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4713, 3.3919, 2.3517, 2.1611, 2.4588, 1.9852, 3.4119, 3.3712], device='cuda:2'), covar=tensor([0.2146, 0.0676, 0.1427, 0.1623, 0.2258, 0.1759, 0.0456, 0.0803], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0251, 0.0268, 0.0244, 0.0282, 0.0203, 0.0244, 0.0267], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:00:24,440 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3409, 4.3270, 4.2657, 3.7697, 4.2800, 1.7081, 4.0880, 4.0863], device='cuda:2'), covar=tensor([0.0073, 0.0055, 0.0091, 0.0251, 0.0060, 0.1708, 0.0087, 0.0120], device='cuda:2'), in_proj_covar=tensor([0.0103, 0.0090, 0.0139, 0.0137, 0.0104, 0.0150, 0.0120, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:00:27,090 INFO [train.py:904] (2/8) Epoch 6, batch 1900, loss[loss=0.2126, simple_loss=0.2807, pruned_loss=0.07224, over 16718.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.287, pruned_loss=0.06644, over 3319706.78 frames. ], batch size: 89, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:00:51,240 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:00:52,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4649, 4.5723, 5.0231, 5.0627, 5.0353, 4.6337, 4.6029, 4.3521], device='cuda:2'), covar=tensor([0.0275, 0.0356, 0.0262, 0.0333, 0.0333, 0.0257, 0.0769, 0.0391], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0267, 0.0265, 0.0263, 0.0322, 0.0280, 0.0395, 0.0233], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 11:00:54,523 INFO [optim.py:368] (2/8) 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 11:01:10,429 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8482, 4.9479, 5.4756, 5.5126, 5.4928, 5.0506, 4.9752, 4.6835], device='cuda:2'), covar=tensor([0.0264, 0.0371, 0.0264, 0.0356, 0.0336, 0.0235, 0.0766, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0267, 0.0266, 0.0263, 0.0322, 0.0280, 0.0396, 0.0233], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 11:01:26,544 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1976, 4.0080, 4.2139, 4.4009, 4.5215, 4.0507, 4.1751, 4.4997], device='cuda:2'), covar=tensor([0.0973, 0.0790, 0.1232, 0.0555, 0.0469, 0.1038, 0.1376, 0.0472], device='cuda:2'), in_proj_covar=tensor([0.0446, 0.0549, 0.0699, 0.0564, 0.0420, 0.0415, 0.0439, 0.0472], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:01:30,568 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:01:36,632 INFO [train.py:904] (2/8) Epoch 6, batch 1950, loss[loss=0.2191, simple_loss=0.286, pruned_loss=0.07611, over 16400.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2863, pruned_loss=0.06544, over 3329581.98 frames. ], batch size: 146, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:01:57,098 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8076, 4.0296, 1.9252, 4.3022, 2.6244, 4.2834, 2.1309, 2.8812], device='cuda:2'), covar=tensor([0.0120, 0.0230, 0.1525, 0.0071, 0.0768, 0.0339, 0.1303, 0.0573], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0161, 0.0176, 0.0085, 0.0160, 0.0193, 0.0185, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:02:36,312 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52744.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:02:47,589 INFO [train.py:904] (2/8) Epoch 6, batch 2000, loss[loss=0.2216, simple_loss=0.2796, pruned_loss=0.08174, over 16873.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2857, pruned_loss=0.06466, over 3338530.06 frames. ], batch size: 116, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:02:49,195 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:03:15,694 INFO [optim.py:368] (2/8) 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 11:03:57,179 INFO [train.py:904] (2/8) Epoch 6, batch 2050, loss[loss=0.1643, simple_loss=0.2456, pruned_loss=0.04148, over 16738.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2859, pruned_loss=0.06527, over 3313572.50 frames. ], batch size: 39, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:04:14,392 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:05:06,133 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:05:08,343 INFO [train.py:904] (2/8) Epoch 6, batch 2100, loss[loss=0.1728, simple_loss=0.2682, pruned_loss=0.03871, over 17304.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2871, pruned_loss=0.06643, over 3315662.60 frames. ], batch size: 52, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:05:36,281 INFO [optim.py:368] (2/8) 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 11:05:50,143 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 11:06:12,924 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:06:18,011 INFO [train.py:904] (2/8) Epoch 6, batch 2150, loss[loss=0.2219, simple_loss=0.3028, pruned_loss=0.07052, over 17076.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2881, pruned_loss=0.06695, over 3323610.95 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:06:33,073 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:07:29,309 INFO [train.py:904] (2/8) Epoch 6, batch 2200, loss[loss=0.184, simple_loss=0.2759, pruned_loss=0.04603, over 17164.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2897, pruned_loss=0.06783, over 3323690.32 frames. ], batch size: 46, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:07:29,789 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8817, 1.8568, 2.3803, 2.9235, 2.5853, 3.4227, 1.7960, 3.3266], device='cuda:2'), covar=tensor([0.0109, 0.0236, 0.0152, 0.0122, 0.0151, 0.0081, 0.0226, 0.0052], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0151, 0.0134, 0.0137, 0.0141, 0.0101, 0.0145, 0.0085], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 11:07:30,849 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8406, 4.1944, 3.2633, 2.3717, 3.1204, 2.5442, 4.5929, 4.2073], device='cuda:2'), covar=tensor([0.2263, 0.0623, 0.1286, 0.1788, 0.2062, 0.1459, 0.0293, 0.0625], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0256, 0.0270, 0.0250, 0.0287, 0.0205, 0.0246, 0.0270], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:07:52,264 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:07:56,706 INFO [optim.py:368] (2/8) 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 11:08:38,587 INFO [train.py:904] (2/8) Epoch 6, batch 2250, loss[loss=0.2929, simple_loss=0.3481, pruned_loss=0.1188, over 11796.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2917, pruned_loss=0.06906, over 3317772.70 frames. ], batch size: 246, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:08:59,899 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:09:13,421 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.77 vs. limit=5.0 2023-04-28 11:09:21,291 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7087, 2.5959, 2.0874, 2.2924, 2.9898, 2.7163, 3.7244, 3.3053], device='cuda:2'), covar=tensor([0.0028, 0.0171, 0.0222, 0.0217, 0.0112, 0.0178, 0.0082, 0.0102], device='cuda:2'), in_proj_covar=tensor([0.0094, 0.0165, 0.0164, 0.0161, 0.0163, 0.0167, 0.0153, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:09:32,751 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0638, 5.0360, 4.8691, 4.2644, 4.8779, 1.9864, 4.6211, 4.8409], device='cuda:2'), covar=tensor([0.0061, 0.0050, 0.0092, 0.0309, 0.0061, 0.1598, 0.0091, 0.0115], device='cuda:2'), in_proj_covar=tensor([0.0102, 0.0090, 0.0138, 0.0136, 0.0104, 0.0148, 0.0120, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:09:47,159 INFO [train.py:904] (2/8) Epoch 6, batch 2300, loss[loss=0.2085, simple_loss=0.2961, pruned_loss=0.06039, over 17063.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.291, pruned_loss=0.06821, over 3321231.21 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:09:51,929 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 11:09:54,567 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:10:15,629 INFO [optim.py:368] (2/8) 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 11:10:57,605 INFO [train.py:904] (2/8) Epoch 6, batch 2350, loss[loss=0.222, simple_loss=0.3016, pruned_loss=0.07114, over 17064.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.291, pruned_loss=0.06878, over 3318656.55 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:11:08,090 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:11:20,188 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:12:08,722 INFO [train.py:904] (2/8) Epoch 6, batch 2400, loss[loss=0.2192, simple_loss=0.3049, pruned_loss=0.06678, over 17077.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2919, pruned_loss=0.06972, over 3314070.72 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:12:10,212 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5911, 4.5631, 4.6223, 4.7111, 4.6509, 5.2219, 4.8477, 4.6531], device='cuda:2'), covar=tensor([0.1158, 0.1727, 0.1879, 0.2005, 0.2869, 0.1003, 0.1256, 0.2334], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0429, 0.0425, 0.0367, 0.0488, 0.0453, 0.0349, 0.0491], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:12:36,945 INFO [optim.py:368] (2/8) 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 11:13:09,900 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3474, 4.0839, 4.2887, 4.5618, 4.6575, 4.1441, 4.4655, 4.5922], device='cuda:2'), covar=tensor([0.0915, 0.0793, 0.1371, 0.0531, 0.0496, 0.0943, 0.1000, 0.0542], device='cuda:2'), in_proj_covar=tensor([0.0438, 0.0542, 0.0700, 0.0557, 0.0422, 0.0414, 0.0441, 0.0472], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:13:19,222 INFO [train.py:904] (2/8) Epoch 6, batch 2450, loss[loss=0.1964, simple_loss=0.2807, pruned_loss=0.05605, over 16268.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2917, pruned_loss=0.06834, over 3319982.69 frames. ], batch size: 36, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:13:33,729 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:13:33,857 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7052, 2.3724, 1.7942, 2.0131, 2.7867, 2.6868, 2.9557, 2.9470], device='cuda:2'), covar=tensor([0.0078, 0.0191, 0.0258, 0.0246, 0.0120, 0.0145, 0.0119, 0.0110], device='cuda:2'), in_proj_covar=tensor([0.0094, 0.0166, 0.0165, 0.0163, 0.0163, 0.0168, 0.0155, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:14:15,994 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:14:29,168 INFO [train.py:904] (2/8) Epoch 6, batch 2500, loss[loss=0.2076, simple_loss=0.2769, pruned_loss=0.0691, over 16858.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2918, pruned_loss=0.06829, over 3304699.74 frames. ], batch size: 102, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:14:39,970 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:14:57,075 INFO [optim.py:368] (2/8) 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 11:15:38,466 INFO [train.py:904] (2/8) Epoch 6, batch 2550, loss[loss=0.2044, simple_loss=0.2754, pruned_loss=0.06673, over 16806.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2922, pruned_loss=0.06858, over 3292169.07 frames. ], batch size: 102, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:15:40,031 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:43,236 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:58,268 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:16:28,889 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-04-28 11:16:48,721 INFO [train.py:904] (2/8) Epoch 6, batch 2600, loss[loss=0.2227, simple_loss=0.291, pruned_loss=0.07719, over 15298.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2909, pruned_loss=0.06735, over 3293268.90 frames. ], batch size: 190, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:17:09,380 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:11,679 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:16,414 INFO [optim.py:368] (2/8) 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 11:17:24,080 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 6, batch 2650, loss[loss=0.2182, simple_loss=0.2977, pruned_loss=0.06933, over 16925.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2917, pruned_loss=0.06726, over 3305479.97 frames. ], batch size: 116, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:18:10,425 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:14,695 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:37,163 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:18:56,605 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-28 11:19:09,546 INFO [train.py:904] (2/8) Epoch 6, batch 2700, loss[loss=0.2154, simple_loss=0.2893, pruned_loss=0.07072, over 16693.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2915, pruned_loss=0.06613, over 3311664.89 frames. ], batch size: 89, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:19:16,987 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:19:38,362 INFO [optim.py:368] (2/8) 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 11:20:13,135 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9824, 5.3494, 5.0551, 5.2050, 4.7726, 4.5876, 4.8867, 5.4434], device='cuda:2'), covar=tensor([0.0801, 0.0723, 0.0893, 0.0451, 0.0749, 0.0845, 0.0693, 0.0785], device='cuda:2'), in_proj_covar=tensor([0.0409, 0.0555, 0.0459, 0.0353, 0.0339, 0.0349, 0.0447, 0.0389], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:20:19,909 INFO [train.py:904] (2/8) Epoch 6, batch 2750, loss[loss=0.2064, simple_loss=0.2785, pruned_loss=0.06717, over 16724.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2906, pruned_loss=0.06551, over 3314297.57 frames. ], batch size: 124, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:20:21,698 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5478, 4.3609, 3.8988, 2.1491, 3.2697, 2.5921, 3.9482, 4.0398], device='cuda:2'), covar=tensor([0.0297, 0.0444, 0.0442, 0.1502, 0.0632, 0.0841, 0.0638, 0.0987], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0137, 0.0155, 0.0140, 0.0130, 0.0124, 0.0140, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:20:24,533 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5351, 2.2034, 1.6002, 2.0248, 2.6816, 2.5551, 2.8413, 2.8280], device='cuda:2'), covar=tensor([0.0061, 0.0173, 0.0250, 0.0203, 0.0096, 0.0142, 0.0098, 0.0103], device='cuda:2'), in_proj_covar=tensor([0.0095, 0.0167, 0.0166, 0.0164, 0.0163, 0.0170, 0.0156, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:21:03,327 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 11:21:09,986 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8379, 4.9139, 5.4187, 5.4058, 5.3813, 4.9942, 4.9738, 4.6880], device='cuda:2'), covar=tensor([0.0250, 0.0359, 0.0258, 0.0340, 0.0340, 0.0249, 0.0764, 0.0316], device='cuda:2'), in_proj_covar=tensor([0.0267, 0.0269, 0.0264, 0.0261, 0.0317, 0.0278, 0.0395, 0.0236], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 11:21:31,892 INFO [train.py:904] (2/8) Epoch 6, batch 2800, loss[loss=0.2161, simple_loss=0.2841, pruned_loss=0.07408, over 15324.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2902, pruned_loss=0.06493, over 3322173.07 frames. ], batch size: 192, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:21:54,561 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-04-28 11:22:01,719 INFO [optim.py:368] (2/8) 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 11:22:18,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4042, 5.3729, 5.1498, 4.3822, 5.1763, 2.2669, 4.8978, 5.2518], device='cuda:2'), covar=tensor([0.0054, 0.0046, 0.0105, 0.0377, 0.0062, 0.1533, 0.0106, 0.0133], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0093, 0.0144, 0.0140, 0.0107, 0.0151, 0.0123, 0.0135], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:22:37,900 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:22:43,361 INFO [train.py:904] (2/8) Epoch 6, batch 2850, loss[loss=0.1948, simple_loss=0.2822, pruned_loss=0.05368, over 17116.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2896, pruned_loss=0.06488, over 3324735.89 frames. ], batch size: 47, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:23:34,773 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2819, 2.1668, 2.3302, 4.8825, 2.0327, 3.2923, 2.3414, 2.5115], device='cuda:2'), covar=tensor([0.0487, 0.2678, 0.1419, 0.0215, 0.3286, 0.1284, 0.2216, 0.2616], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0336, 0.0275, 0.0316, 0.0376, 0.0354, 0.0304, 0.0402], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:23:51,283 INFO [train.py:904] (2/8) Epoch 6, batch 2900, loss[loss=0.2013, simple_loss=0.2834, pruned_loss=0.05961, over 17220.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2891, pruned_loss=0.06607, over 3324269.56 frames. ], batch size: 46, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:24:04,494 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:19,363 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:20,247 INFO [optim.py:368] (2/8) 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 11:24:59,791 INFO [train.py:904] (2/8) Epoch 6, batch 2950, loss[loss=0.1652, simple_loss=0.2481, pruned_loss=0.04113, over 16840.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.289, pruned_loss=0.06683, over 3323825.31 frames. ], batch size: 42, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:25:10,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7914, 3.7892, 2.7453, 2.3226, 2.7273, 2.2845, 3.7600, 3.7660], device='cuda:2'), covar=tensor([0.1933, 0.0559, 0.1192, 0.1790, 0.1946, 0.1463, 0.0478, 0.0804], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0257, 0.0270, 0.0251, 0.0291, 0.0205, 0.0246, 0.0272], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:25:14,313 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1750, 5.8306, 5.9395, 5.7334, 5.7401, 6.1839, 5.7843, 5.5273], device='cuda:2'), covar=tensor([0.0735, 0.1431, 0.1538, 0.1479, 0.2577, 0.0923, 0.1057, 0.2055], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0423, 0.0415, 0.0363, 0.0487, 0.0446, 0.0344, 0.0491], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:25:15,506 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:25:30,127 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:26:08,566 INFO [train.py:904] (2/8) Epoch 6, batch 3000, loss[loss=0.1909, simple_loss=0.2823, pruned_loss=0.04972, over 17056.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2895, pruned_loss=0.06758, over 3318330.55 frames. ], batch size: 50, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:26:08,566 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 11:26:17,396 INFO [train.py:938] (2/8) Epoch 6, validation: loss=0.1514, simple_loss=0.258, pruned_loss=0.02246, over 944034.00 frames. 2023-04-28 11:26:17,398 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 11:26:29,470 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:26:38,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0341, 5.0070, 4.8533, 4.6224, 4.4262, 4.8956, 4.8420, 4.5594], device='cuda:2'), covar=tensor([0.0448, 0.0301, 0.0197, 0.0227, 0.0952, 0.0275, 0.0295, 0.0554], device='cuda:2'), in_proj_covar=tensor([0.0209, 0.0228, 0.0248, 0.0220, 0.0282, 0.0249, 0.0174, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:26:45,969 INFO [optim.py:368] (2/8) 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 11:26:47,715 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 11:27:25,829 INFO [train.py:904] (2/8) Epoch 6, batch 3050, loss[loss=0.2214, simple_loss=0.3086, pruned_loss=0.06707, over 16695.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2884, pruned_loss=0.06661, over 3315286.70 frames. ], batch size: 62, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:28:35,670 INFO [train.py:904] (2/8) Epoch 6, batch 3100, loss[loss=0.21, simple_loss=0.2939, pruned_loss=0.06302, over 16675.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2876, pruned_loss=0.06659, over 3320486.88 frames. ], batch size: 62, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:29:05,910 INFO [optim.py:368] (2/8) 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 11:29:12,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1465, 3.6461, 3.0760, 1.9822, 2.7445, 2.2929, 3.5361, 3.5188], device='cuda:2'), covar=tensor([0.0248, 0.0612, 0.0583, 0.1467, 0.0661, 0.0889, 0.0509, 0.0861], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0136, 0.0153, 0.0138, 0.0129, 0.0123, 0.0138, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:29:14,069 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1583, 1.8058, 2.4280, 2.9048, 2.8658, 3.3090, 1.9552, 3.2386], device='cuda:2'), covar=tensor([0.0088, 0.0235, 0.0155, 0.0138, 0.0125, 0.0083, 0.0226, 0.0090], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0153, 0.0137, 0.0137, 0.0142, 0.0103, 0.0147, 0.0089], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 11:29:40,276 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:29:45,137 INFO [train.py:904] (2/8) Epoch 6, batch 3150, loss[loss=0.223, simple_loss=0.289, pruned_loss=0.0785, over 16547.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2869, pruned_loss=0.06667, over 3305604.89 frames. ], batch size: 75, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:30:09,206 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0377, 5.0408, 4.8820, 3.8850, 4.8894, 1.6924, 4.6932, 4.8720], device='cuda:2'), covar=tensor([0.0092, 0.0071, 0.0124, 0.0529, 0.0096, 0.2145, 0.0122, 0.0201], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0093, 0.0144, 0.0141, 0.0108, 0.0151, 0.0125, 0.0137], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:30:35,934 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-28 11:30:49,056 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:30:56,670 INFO [train.py:904] (2/8) Epoch 6, batch 3200, loss[loss=0.1663, simple_loss=0.2564, pruned_loss=0.03811, over 17214.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2863, pruned_loss=0.06554, over 3320293.01 frames. ], batch size: 45, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:31:09,490 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:31:25,790 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 11:32:09,687 INFO [train.py:904] (2/8) Epoch 6, batch 3250, loss[loss=0.1945, simple_loss=0.2747, pruned_loss=0.05713, over 17089.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2869, pruned_loss=0.0661, over 3313094.19 frames. ], batch size: 47, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:32:19,694 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:35,404 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:40,060 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:33:17,041 INFO [train.py:904] (2/8) Epoch 6, batch 3300, loss[loss=0.1827, simple_loss=0.2651, pruned_loss=0.05021, over 16844.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2873, pruned_loss=0.06534, over 3321913.21 frames. ], batch size: 42, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:33:45,160 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:33:46,070 INFO [optim.py:368] (2/8) 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 11:33:53,601 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5559, 3.4700, 3.9769, 2.7596, 3.4776, 3.8935, 3.7255, 2.3444], device='cuda:2'), covar=tensor([0.0270, 0.0129, 0.0029, 0.0202, 0.0052, 0.0043, 0.0039, 0.0258], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0060, 0.0061, 0.0111, 0.0062, 0.0070, 0.0064, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:34:03,641 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 11:34:26,281 INFO [train.py:904] (2/8) Epoch 6, batch 3350, loss[loss=0.1677, simple_loss=0.2497, pruned_loss=0.04292, over 16743.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2871, pruned_loss=0.06477, over 3326265.06 frames. ], batch size: 39, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:34:59,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8801, 3.4917, 3.0786, 1.8291, 2.5890, 2.1124, 3.3574, 3.4377], device='cuda:2'), covar=tensor([0.0250, 0.0524, 0.0527, 0.1514, 0.0712, 0.0874, 0.0574, 0.0694], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0137, 0.0152, 0.0138, 0.0130, 0.0123, 0.0138, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:35:26,109 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-28 11:35:34,480 INFO [train.py:904] (2/8) Epoch 6, batch 3400, loss[loss=0.1852, simple_loss=0.2714, pruned_loss=0.04947, over 17123.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2868, pruned_loss=0.06445, over 3326021.68 frames. ], batch size: 49, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:36:05,330 INFO [optim.py:368] (2/8) 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 11:36:13,287 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1371, 5.1040, 4.9660, 4.8044, 4.5219, 5.0107, 4.9797, 4.6830], device='cuda:2'), covar=tensor([0.0438, 0.0281, 0.0202, 0.0185, 0.1032, 0.0275, 0.0224, 0.0559], device='cuda:2'), in_proj_covar=tensor([0.0215, 0.0233, 0.0255, 0.0227, 0.0289, 0.0256, 0.0181, 0.0285], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:36:33,076 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 11:36:46,918 INFO [train.py:904] (2/8) Epoch 6, batch 3450, loss[loss=0.162, simple_loss=0.2492, pruned_loss=0.03742, over 17213.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2854, pruned_loss=0.06379, over 3317732.30 frames. ], batch size: 44, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:37:06,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4475, 3.8229, 3.8051, 1.8720, 4.0152, 3.9867, 3.1063, 3.0010], device='cuda:2'), covar=tensor([0.0700, 0.0109, 0.0134, 0.1072, 0.0053, 0.0085, 0.0332, 0.0362], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0095, 0.0086, 0.0140, 0.0072, 0.0085, 0.0118, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 11:37:26,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7773, 4.1109, 2.9173, 2.3683, 3.0496, 2.2277, 4.2096, 4.0456], device='cuda:2'), covar=tensor([0.2180, 0.0596, 0.1381, 0.1696, 0.2148, 0.1734, 0.0380, 0.0688], device='cuda:2'), in_proj_covar=tensor([0.0281, 0.0258, 0.0272, 0.0255, 0.0294, 0.0207, 0.0249, 0.0280], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:37:48,858 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 11:37:58,612 INFO [train.py:904] (2/8) Epoch 6, batch 3500, loss[loss=0.2235, simple_loss=0.2815, pruned_loss=0.08276, over 16451.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2835, pruned_loss=0.06302, over 3309300.77 frames. ], batch size: 146, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:38:28,902 INFO [optim.py:368] (2/8) 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 11:38:34,527 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5028, 4.2862, 4.4314, 4.6885, 4.8336, 4.3525, 4.6288, 4.7120], device='cuda:2'), covar=tensor([0.1142, 0.0899, 0.1407, 0.0708, 0.0548, 0.0817, 0.1041, 0.0763], device='cuda:2'), in_proj_covar=tensor([0.0446, 0.0540, 0.0705, 0.0567, 0.0423, 0.0418, 0.0433, 0.0474], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:38:52,859 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8864, 1.6316, 2.1401, 2.7977, 2.7461, 2.6965, 1.6673, 2.8921], device='cuda:2'), covar=tensor([0.0081, 0.0250, 0.0197, 0.0117, 0.0112, 0.0121, 0.0260, 0.0070], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0152, 0.0138, 0.0137, 0.0143, 0.0103, 0.0145, 0.0090], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 11:39:10,638 INFO [train.py:904] (2/8) Epoch 6, batch 3550, loss[loss=0.1849, simple_loss=0.2743, pruned_loss=0.04773, over 17010.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2828, pruned_loss=0.06307, over 3306765.81 frames. ], batch size: 50, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:39:54,993 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9876, 5.5085, 5.5218, 5.3357, 5.5373, 6.0513, 5.6743, 5.3747], device='cuda:2'), covar=tensor([0.0785, 0.1537, 0.1636, 0.1726, 0.2509, 0.0924, 0.1145, 0.1944], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0431, 0.0426, 0.0372, 0.0496, 0.0457, 0.0351, 0.0497], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:40:04,951 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7484, 5.0391, 5.0728, 5.0332, 5.0183, 5.6220, 5.3040, 4.9668], device='cuda:2'), covar=tensor([0.0967, 0.1560, 0.1441, 0.1660, 0.2585, 0.0975, 0.1164, 0.2131], device='cuda:2'), in_proj_covar=tensor([0.0303, 0.0429, 0.0425, 0.0370, 0.0494, 0.0456, 0.0350, 0.0494], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:40:21,759 INFO [train.py:904] (2/8) Epoch 6, batch 3600, loss[loss=0.1905, simple_loss=0.2596, pruned_loss=0.06072, over 16763.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2815, pruned_loss=0.06217, over 3295319.66 frames. ], batch size: 102, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:40:35,087 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-28 11:40:51,131 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 6, batch 3650, loss[loss=0.2069, simple_loss=0.2728, pruned_loss=0.07045, over 16855.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2804, pruned_loss=0.06275, over 3293755.39 frames. ], batch size: 116, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:42:46,526 INFO [train.py:904] (2/8) Epoch 6, batch 3700, loss[loss=0.1975, simple_loss=0.2656, pruned_loss=0.06467, over 16724.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2801, pruned_loss=0.06472, over 3272057.94 frames. ], batch size: 134, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:43:16,559 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6634, 4.7762, 5.2385, 5.2033, 5.1402, 4.8460, 4.7553, 4.4633], device='cuda:2'), covar=tensor([0.0260, 0.0350, 0.0249, 0.0313, 0.0369, 0.0225, 0.0659, 0.0359], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0268, 0.0261, 0.0261, 0.0319, 0.0277, 0.0388, 0.0232], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 11:43:17,524 INFO [optim.py:368] (2/8) 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 11:43:59,569 INFO [train.py:904] (2/8) Epoch 6, batch 3750, loss[loss=0.2103, simple_loss=0.2796, pruned_loss=0.07052, over 16227.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2806, pruned_loss=0.06637, over 3268303.93 frames. ], batch size: 165, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:44:45,983 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 11:44:49,691 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7207, 3.8985, 2.7209, 2.3397, 2.9464, 2.2674, 3.7350, 3.8916], device='cuda:2'), covar=tensor([0.2048, 0.0518, 0.1330, 0.1641, 0.1744, 0.1394, 0.0509, 0.0661], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0255, 0.0269, 0.0252, 0.0291, 0.0205, 0.0248, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:45:13,202 INFO [train.py:904] (2/8) Epoch 6, batch 3800, loss[loss=0.2415, simple_loss=0.299, pruned_loss=0.09205, over 16879.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2819, pruned_loss=0.06819, over 3271623.35 frames. ], batch size: 109, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:45:19,898 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 11:45:46,020 INFO [optim.py:368] (2/8) 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 11:46:28,643 INFO [train.py:904] (2/8) Epoch 6, batch 3850, loss[loss=0.2216, simple_loss=0.2889, pruned_loss=0.07714, over 15432.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2827, pruned_loss=0.06912, over 3261620.23 frames. ], batch size: 190, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:03,779 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5966, 4.1549, 4.5167, 1.9029, 4.8478, 4.9478, 3.2601, 3.5497], device='cuda:2'), covar=tensor([0.0786, 0.0177, 0.0116, 0.1221, 0.0032, 0.0029, 0.0306, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0094, 0.0081, 0.0138, 0.0072, 0.0082, 0.0115, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 11:47:25,860 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 11:47:40,654 INFO [train.py:904] (2/8) Epoch 6, batch 3900, loss[loss=0.1932, simple_loss=0.276, pruned_loss=0.05523, over 16561.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2814, pruned_loss=0.06878, over 3279478.95 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:59,640 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:48:10,210 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3880, 4.1437, 4.3611, 4.5762, 4.6650, 4.1854, 4.4268, 4.6090], device='cuda:2'), covar=tensor([0.0860, 0.0852, 0.1272, 0.0521, 0.0480, 0.0930, 0.1195, 0.0588], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0515, 0.0664, 0.0536, 0.0402, 0.0396, 0.0416, 0.0450], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:48:10,861 INFO [optim.py:368] (2/8) 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 11:48:28,763 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6330, 1.6167, 2.0693, 2.5365, 2.6314, 2.4835, 1.6846, 2.7630], device='cuda:2'), covar=tensor([0.0087, 0.0252, 0.0181, 0.0115, 0.0104, 0.0129, 0.0218, 0.0059], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0151, 0.0136, 0.0136, 0.0141, 0.0103, 0.0145, 0.0089], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 11:48:51,859 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6202, 4.2542, 3.8492, 1.9992, 2.9421, 2.5870, 3.8423, 4.0575], device='cuda:2'), covar=tensor([0.0177, 0.0360, 0.0442, 0.1539, 0.0731, 0.0781, 0.0522, 0.0638], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0133, 0.0151, 0.0137, 0.0129, 0.0122, 0.0136, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:48:53,476 INFO [train.py:904] (2/8) Epoch 6, batch 3950, loss[loss=0.2104, simple_loss=0.2757, pruned_loss=0.07257, over 16541.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2805, pruned_loss=0.06942, over 3282204.76 frames. ], batch size: 75, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:49:17,395 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0354, 3.5420, 3.0351, 1.8790, 2.5797, 2.1022, 3.4696, 3.4328], device='cuda:2'), covar=tensor([0.0247, 0.0529, 0.0579, 0.1498, 0.0792, 0.0943, 0.0570, 0.0726], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0134, 0.0153, 0.0138, 0.0130, 0.0123, 0.0136, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:49:27,782 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:50:04,726 INFO [train.py:904] (2/8) Epoch 6, batch 4000, loss[loss=0.1774, simple_loss=0.2578, pruned_loss=0.04851, over 16626.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2804, pruned_loss=0.0696, over 3287449.41 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:50:36,654 INFO [optim.py:368] (2/8) 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 11:51:17,355 INFO [train.py:904] (2/8) Epoch 6, batch 4050, loss[loss=0.2015, simple_loss=0.2742, pruned_loss=0.06441, over 16724.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2808, pruned_loss=0.06833, over 3285004.26 frames. ], batch size: 57, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:16,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4167, 4.4177, 4.5495, 4.5603, 4.4979, 5.0206, 4.6222, 4.2701], device='cuda:2'), covar=tensor([0.1086, 0.1424, 0.1405, 0.1628, 0.2207, 0.0854, 0.1184, 0.2470], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0418, 0.0410, 0.0358, 0.0479, 0.0436, 0.0341, 0.0487], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:52:28,288 INFO [train.py:904] (2/8) Epoch 6, batch 4100, loss[loss=0.2339, simple_loss=0.3065, pruned_loss=0.08069, over 16749.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2813, pruned_loss=0.06708, over 3272439.67 frames. ], batch size: 124, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:45,818 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:52:53,974 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-28 11:53:01,850 INFO [optim.py:368] (2/8) 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 11:53:30,598 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5232, 3.5722, 2.7927, 2.1945, 2.7850, 2.1297, 3.8567, 3.6975], device='cuda:2'), covar=tensor([0.2508, 0.0875, 0.1484, 0.1751, 0.2086, 0.1603, 0.0484, 0.0696], device='cuda:2'), in_proj_covar=tensor([0.0281, 0.0251, 0.0272, 0.0255, 0.0298, 0.0207, 0.0249, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 11:53:45,451 INFO [train.py:904] (2/8) Epoch 6, batch 4150, loss[loss=0.2484, simple_loss=0.3308, pruned_loss=0.08297, over 16299.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.29, pruned_loss=0.0711, over 3223757.70 frames. ], batch size: 165, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:54:19,639 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 6, batch 4200, loss[loss=0.2904, simple_loss=0.3462, pruned_loss=0.1173, over 11564.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2981, pruned_loss=0.07362, over 3190980.81 frames. ], batch size: 248, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:55:33,715 INFO [optim.py:368] (2/8) 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 11:56:15,277 INFO [train.py:904] (2/8) Epoch 6, batch 4250, loss[loss=0.2098, simple_loss=0.3012, pruned_loss=0.05916, over 16679.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3014, pruned_loss=0.07368, over 3188879.11 frames. ], batch size: 134, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:56:34,414 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7362, 3.5576, 3.2269, 1.8861, 2.7474, 2.3347, 3.2293, 3.3412], device='cuda:2'), covar=tensor([0.0270, 0.0459, 0.0512, 0.1581, 0.0736, 0.0870, 0.0636, 0.0562], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0133, 0.0154, 0.0140, 0.0132, 0.0124, 0.0139, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:56:44,055 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:56:44,223 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4149, 3.5808, 1.7572, 3.7266, 2.4320, 3.6772, 2.0249, 2.6322], device='cuda:2'), covar=tensor([0.0141, 0.0239, 0.1550, 0.0047, 0.0773, 0.0260, 0.1352, 0.0627], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0157, 0.0178, 0.0084, 0.0162, 0.0189, 0.0184, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 11:57:29,394 INFO [train.py:904] (2/8) Epoch 6, batch 4300, loss[loss=0.2417, simple_loss=0.3201, pruned_loss=0.08166, over 16790.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3022, pruned_loss=0.07215, over 3205374.15 frames. ], batch size: 57, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:58:01,800 INFO [optim.py:368] (2/8) 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 11:58:30,631 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-28 11:58:43,134 INFO [train.py:904] (2/8) Epoch 6, batch 4350, loss[loss=0.2332, simple_loss=0.314, pruned_loss=0.07616, over 16313.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3058, pruned_loss=0.07377, over 3174546.48 frames. ], batch size: 165, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:59:47,982 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7953, 1.6950, 1.4144, 1.5121, 1.8379, 1.6691, 1.8349, 1.9579], device='cuda:2'), covar=tensor([0.0054, 0.0152, 0.0214, 0.0179, 0.0107, 0.0139, 0.0080, 0.0096], device='cuda:2'), in_proj_covar=tensor([0.0090, 0.0161, 0.0161, 0.0159, 0.0157, 0.0162, 0.0144, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 11:59:52,964 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:59:56,585 INFO [train.py:904] (2/8) Epoch 6, batch 4400, loss[loss=0.2221, simple_loss=0.3098, pruned_loss=0.06719, over 16642.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3079, pruned_loss=0.07484, over 3171259.64 frames. ], batch size: 57, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:00:27,149 INFO [optim.py:368] (2/8) 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 12:01:06,618 INFO [train.py:904] (2/8) Epoch 6, batch 4450, loss[loss=0.2157, simple_loss=0.3039, pruned_loss=0.06375, over 16921.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3106, pruned_loss=0.07563, over 3184031.46 frames. ], batch size: 96, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:01:19,208 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:21,612 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:31,360 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:02:16,914 INFO [train.py:904] (2/8) Epoch 6, batch 4500, loss[loss=0.2511, simple_loss=0.3299, pruned_loss=0.08615, over 16902.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3102, pruned_loss=0.07545, over 3175859.33 frames. ], batch size: 116, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:02:20,284 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4634, 2.9447, 2.5166, 4.4188, 3.4583, 4.1413, 1.6509, 2.9082], device='cuda:2'), covar=tensor([0.1444, 0.0559, 0.1110, 0.0068, 0.0246, 0.0289, 0.1336, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0143, 0.0168, 0.0095, 0.0193, 0.0187, 0.0163, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 12:02:24,548 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2931, 3.1632, 2.9849, 3.4254, 3.3838, 3.2367, 3.4232, 3.4571], device='cuda:2'), covar=tensor([0.0716, 0.0728, 0.1667, 0.0608, 0.0819, 0.1886, 0.0817, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0405, 0.0487, 0.0630, 0.0499, 0.0381, 0.0377, 0.0393, 0.0424], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:02:47,062 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:02:48,405 INFO [optim.py:368] (2/8) 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 12:02:59,260 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8068, 4.6351, 4.5445, 3.2618, 3.9006, 4.5414, 4.1433, 2.4607], device='cuda:2'), covar=tensor([0.0281, 0.0009, 0.0018, 0.0188, 0.0042, 0.0045, 0.0025, 0.0257], device='cuda:2'), in_proj_covar=tensor([0.0116, 0.0056, 0.0059, 0.0113, 0.0062, 0.0069, 0.0064, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 12:03:29,872 INFO [train.py:904] (2/8) Epoch 6, batch 4550, loss[loss=0.2516, simple_loss=0.3245, pruned_loss=0.08941, over 16307.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3104, pruned_loss=0.07557, over 3195653.33 frames. ], batch size: 35, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:03:33,386 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7966, 1.6569, 2.2411, 2.6393, 2.4569, 3.1287, 1.6432, 2.9183], device='cuda:2'), covar=tensor([0.0091, 0.0251, 0.0153, 0.0147, 0.0136, 0.0069, 0.0241, 0.0053], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0148, 0.0130, 0.0131, 0.0137, 0.0099, 0.0143, 0.0086], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 12:03:57,406 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:04:41,589 INFO [train.py:904] (2/8) Epoch 6, batch 4600, loss[loss=0.195, simple_loss=0.2812, pruned_loss=0.05443, over 16866.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3107, pruned_loss=0.07509, over 3204171.29 frames. ], batch size: 116, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:05:07,046 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:05:13,322 INFO [optim.py:368] (2/8) 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 12:05:52,752 INFO [train.py:904] (2/8) Epoch 6, batch 4650, loss[loss=0.2052, simple_loss=0.2883, pruned_loss=0.06107, over 16526.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3096, pruned_loss=0.075, over 3197947.99 frames. ], batch size: 68, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:06:12,015 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7031, 3.5302, 3.6536, 3.5659, 3.5902, 4.1137, 3.8643, 3.5708], device='cuda:2'), covar=tensor([0.1808, 0.1771, 0.1542, 0.2494, 0.3234, 0.1552, 0.1202, 0.2575], device='cuda:2'), in_proj_covar=tensor([0.0290, 0.0396, 0.0393, 0.0347, 0.0454, 0.0417, 0.0324, 0.0466], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 12:06:46,701 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-04-28 12:07:03,017 INFO [train.py:904] (2/8) Epoch 6, batch 4700, loss[loss=0.2163, simple_loss=0.3043, pruned_loss=0.06413, over 16879.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3066, pruned_loss=0.07336, over 3190977.52 frames. ], batch size: 96, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:07:34,099 INFO [optim.py:368] (2/8) 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 12:08:12,118 INFO [train.py:904] (2/8) Epoch 6, batch 4750, loss[loss=0.233, simple_loss=0.3102, pruned_loss=0.07787, over 15425.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3031, pruned_loss=0.07178, over 3193259.96 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:08:17,379 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:08:36,586 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:09:03,645 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1873, 1.8903, 1.5941, 1.7621, 2.2150, 2.0013, 2.2064, 2.3455], device='cuda:2'), covar=tensor([0.0048, 0.0187, 0.0226, 0.0217, 0.0101, 0.0178, 0.0076, 0.0106], device='cuda:2'), in_proj_covar=tensor([0.0090, 0.0160, 0.0162, 0.0159, 0.0155, 0.0164, 0.0145, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:09:05,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9210, 2.7021, 2.6279, 1.7934, 2.7742, 2.8286, 2.3673, 2.3700], device='cuda:2'), covar=tensor([0.0738, 0.0173, 0.0195, 0.0919, 0.0110, 0.0099, 0.0368, 0.0391], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0095, 0.0083, 0.0141, 0.0073, 0.0080, 0.0118, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 12:09:22,945 INFO [train.py:904] (2/8) Epoch 6, batch 4800, loss[loss=0.2064, simple_loss=0.2782, pruned_loss=0.06727, over 17207.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2991, pruned_loss=0.0695, over 3200563.14 frames. ], batch size: 43, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:09:45,467 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:46,596 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:53,990 INFO [optim.py:368] (2/8) 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 12:10:11,863 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 12:10:35,251 INFO [train.py:904] (2/8) Epoch 6, batch 4850, loss[loss=0.203, simple_loss=0.2796, pruned_loss=0.06324, over 16627.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3006, pruned_loss=0.06986, over 3177690.67 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:11:48,112 INFO [train.py:904] (2/8) Epoch 6, batch 4900, loss[loss=0.1986, simple_loss=0.2877, pruned_loss=0.05472, over 16954.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3004, pruned_loss=0.06885, over 3153537.35 frames. ], batch size: 109, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:11:58,087 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.39 vs. limit=5.0 2023-04-28 12:12:04,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3615, 3.5373, 1.6010, 3.7301, 2.4344, 3.6095, 1.8404, 2.5285], device='cuda:2'), covar=tensor([0.0148, 0.0230, 0.1612, 0.0045, 0.0708, 0.0346, 0.1467, 0.0680], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0156, 0.0177, 0.0082, 0.0162, 0.0187, 0.0186, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 12:12:19,090 INFO [optim.py:368] (2/8) 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 12:12:59,378 INFO [train.py:904] (2/8) Epoch 6, batch 4950, loss[loss=0.236, simple_loss=0.3186, pruned_loss=0.0767, over 16959.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2999, pruned_loss=0.06792, over 3176750.96 frames. ], batch size: 109, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:13:45,575 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 12:14:05,602 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:14:08,593 INFO [train.py:904] (2/8) Epoch 6, batch 5000, loss[loss=0.2004, simple_loss=0.2943, pruned_loss=0.05323, over 16461.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3021, pruned_loss=0.06796, over 3201946.75 frames. ], batch size: 146, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:28,551 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6347, 5.9034, 5.5376, 5.7768, 5.2819, 4.9208, 5.4436, 6.0218], device='cuda:2'), covar=tensor([0.0546, 0.0653, 0.0913, 0.0518, 0.0641, 0.0561, 0.0695, 0.0706], device='cuda:2'), in_proj_covar=tensor([0.0396, 0.0527, 0.0446, 0.0341, 0.0326, 0.0341, 0.0433, 0.0379], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:14:38,593 INFO [optim.py:368] (2/8) 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 12:15:13,067 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2989, 1.8938, 2.0399, 3.9044, 1.6919, 2.7000, 2.1011, 2.1010], device='cuda:2'), covar=tensor([0.0689, 0.2484, 0.1394, 0.0317, 0.3070, 0.1299, 0.2140, 0.2329], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0341, 0.0279, 0.0315, 0.0381, 0.0352, 0.0305, 0.0403], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:15:21,195 INFO [train.py:904] (2/8) Epoch 6, batch 5050, loss[loss=0.2085, simple_loss=0.3003, pruned_loss=0.05837, over 16726.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3022, pruned_loss=0.06744, over 3214033.81 frames. ], batch size: 83, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:15:25,452 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:15:33,304 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:28,564 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:31,556 INFO [train.py:904] (2/8) Epoch 6, batch 5100, loss[loss=0.2221, simple_loss=0.3028, pruned_loss=0.07069, over 16663.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.3001, pruned_loss=0.06647, over 3209280.19 frames. ], batch size: 134, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:16:32,963 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:54,848 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:17:04,478 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 6, batch 5150, loss[loss=0.1874, simple_loss=0.2873, pruned_loss=0.04368, over 16788.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2997, pruned_loss=0.06565, over 3205199.82 frames. ], batch size: 102, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:17:57,066 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:05,651 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:58,946 INFO [train.py:904] (2/8) Epoch 6, batch 5200, loss[loss=0.2175, simple_loss=0.2955, pruned_loss=0.06979, over 16392.00 frames. ], tot_loss[loss=0.214, simple_loss=0.298, pruned_loss=0.065, over 3196653.61 frames. ], batch size: 146, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:19:30,989 INFO [optim.py:368] (2/8) 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 12:20:15,819 INFO [train.py:904] (2/8) Epoch 6, batch 5250, loss[loss=0.1936, simple_loss=0.2645, pruned_loss=0.06138, over 17249.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2948, pruned_loss=0.06441, over 3208455.89 frames. ], batch size: 45, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:20:30,668 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7217, 3.5054, 3.7597, 3.5897, 3.7556, 4.1845, 3.8869, 3.5848], device='cuda:2'), covar=tensor([0.1760, 0.2021, 0.1591, 0.2197, 0.2560, 0.1438, 0.1252, 0.2582], device='cuda:2'), in_proj_covar=tensor([0.0288, 0.0393, 0.0391, 0.0340, 0.0455, 0.0421, 0.0324, 0.0466], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 12:21:28,014 INFO [train.py:904] (2/8) Epoch 6, batch 5300, loss[loss=0.1969, simple_loss=0.283, pruned_loss=0.05537, over 15354.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2919, pruned_loss=0.0635, over 3186653.95 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:21:59,802 INFO [optim.py:368] (2/8) 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 12:22:40,508 INFO [train.py:904] (2/8) Epoch 6, batch 5350, loss[loss=0.2007, simple_loss=0.2779, pruned_loss=0.06177, over 16181.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2903, pruned_loss=0.06255, over 3199014.20 frames. ], batch size: 35, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:22:45,577 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:44,956 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:52,038 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5433, 2.5009, 2.1803, 4.0613, 2.9455, 3.8751, 1.4000, 2.7554], device='cuda:2'), covar=tensor([0.1417, 0.0734, 0.1354, 0.0098, 0.0259, 0.0373, 0.1527, 0.0823], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0144, 0.0171, 0.0095, 0.0189, 0.0188, 0.0163, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 12:23:52,596 INFO [train.py:904] (2/8) Epoch 6, batch 5400, loss[loss=0.2256, simple_loss=0.3154, pruned_loss=0.06793, over 16949.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2938, pruned_loss=0.06394, over 3190742.39 frames. ], batch size: 116, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:24:26,633 INFO [optim.py:368] (2/8) 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 12:25:09,002 INFO [train.py:904] (2/8) Epoch 6, batch 5450, loss[loss=0.2513, simple_loss=0.3244, pruned_loss=0.08913, over 16866.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2981, pruned_loss=0.06659, over 3177232.67 frames. ], batch size: 116, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:25:14,382 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:17,575 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:22,892 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 12:26:21,090 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7223, 4.7665, 4.5353, 3.9497, 4.6196, 1.9556, 4.4254, 4.4678], device='cuda:2'), covar=tensor([0.0051, 0.0044, 0.0089, 0.0296, 0.0059, 0.1568, 0.0072, 0.0120], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0087, 0.0134, 0.0133, 0.0100, 0.0148, 0.0115, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:26:27,336 INFO [train.py:904] (2/8) Epoch 6, batch 5500, loss[loss=0.2463, simple_loss=0.322, pruned_loss=0.08533, over 16506.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3061, pruned_loss=0.07214, over 3165577.52 frames. ], batch size: 62, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:27:01,683 INFO [optim.py:368] (2/8) 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 12:27:46,057 INFO [train.py:904] (2/8) Epoch 6, batch 5550, loss[loss=0.2573, simple_loss=0.334, pruned_loss=0.09034, over 16484.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3141, pruned_loss=0.07856, over 3149951.57 frames. ], batch size: 68, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:28:13,623 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3124, 4.4258, 4.4992, 4.4536, 4.5169, 5.0518, 4.6493, 4.4472], device='cuda:2'), covar=tensor([0.1137, 0.1652, 0.1460, 0.1486, 0.2037, 0.0818, 0.1091, 0.1992], device='cuda:2'), in_proj_covar=tensor([0.0290, 0.0400, 0.0400, 0.0344, 0.0459, 0.0426, 0.0327, 0.0470], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 12:29:07,880 INFO [train.py:904] (2/8) Epoch 6, batch 5600, loss[loss=0.327, simple_loss=0.3715, pruned_loss=0.1412, over 10990.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3216, pruned_loss=0.0856, over 3091200.52 frames. ], batch size: 248, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:29:45,252 INFO [optim.py:368] (2/8) 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 12:30:04,437 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8976, 2.6643, 2.5940, 1.8053, 2.7583, 2.7310, 2.3655, 2.3367], device='cuda:2'), covar=tensor([0.0806, 0.0163, 0.0178, 0.0880, 0.0098, 0.0155, 0.0374, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0092, 0.0079, 0.0138, 0.0071, 0.0080, 0.0114, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 12:30:10,148 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6348, 4.6471, 4.4757, 4.3993, 4.0811, 4.5538, 4.4801, 4.2867], device='cuda:2'), covar=tensor([0.0522, 0.0248, 0.0238, 0.0198, 0.0824, 0.0308, 0.0312, 0.0546], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0214, 0.0231, 0.0202, 0.0258, 0.0232, 0.0161, 0.0257], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:30:22,194 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9279, 1.6719, 1.3657, 1.4851, 1.8710, 1.5116, 1.7392, 1.8789], device='cuda:2'), covar=tensor([0.0050, 0.0139, 0.0222, 0.0175, 0.0097, 0.0150, 0.0098, 0.0103], device='cuda:2'), in_proj_covar=tensor([0.0088, 0.0161, 0.0165, 0.0162, 0.0157, 0.0167, 0.0147, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:30:29,968 INFO [train.py:904] (2/8) Epoch 6, batch 5650, loss[loss=0.2603, simple_loss=0.3292, pruned_loss=0.09574, over 16679.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3282, pruned_loss=0.09141, over 3061820.02 frames. ], batch size: 134, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:30:34,884 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:31:48,514 INFO [train.py:904] (2/8) Epoch 6, batch 5700, loss[loss=0.2748, simple_loss=0.3435, pruned_loss=0.1031, over 16873.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3295, pruned_loss=0.09263, over 3079440.00 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:31:51,664 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:32:04,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6143, 3.3296, 2.8302, 1.6567, 2.4238, 2.1560, 3.0327, 3.2520], device='cuda:2'), covar=tensor([0.0334, 0.0536, 0.0674, 0.1817, 0.0923, 0.0950, 0.0798, 0.0722], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0126, 0.0150, 0.0138, 0.0130, 0.0123, 0.0136, 0.0137], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 12:32:25,511 INFO [optim.py:368] (2/8) 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 12:32:38,127 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6937, 1.1866, 1.6307, 1.7400, 1.7685, 1.8752, 1.3829, 1.6812], device='cuda:2'), covar=tensor([0.0108, 0.0180, 0.0106, 0.0122, 0.0108, 0.0070, 0.0197, 0.0055], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0147, 0.0127, 0.0131, 0.0135, 0.0097, 0.0143, 0.0086], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 12:33:08,901 INFO [train.py:904] (2/8) Epoch 6, batch 5750, loss[loss=0.2262, simple_loss=0.3111, pruned_loss=0.07065, over 16396.00 frames. ], tot_loss[loss=0.2601, simple_loss=0.3323, pruned_loss=0.09397, over 3063306.25 frames. ], batch size: 68, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:33:09,309 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:13,956 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:30,569 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-28 12:33:44,566 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:34:30,321 INFO [train.py:904] (2/8) Epoch 6, batch 5800, loss[loss=0.2286, simple_loss=0.3115, pruned_loss=0.07289, over 16398.00 frames. ], tot_loss[loss=0.2578, simple_loss=0.3313, pruned_loss=0.09216, over 3063266.11 frames. ], batch size: 68, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:34:32,752 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:35:05,968 INFO [optim.py:368] (2/8) 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 12:35:22,927 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:35:49,149 INFO [train.py:904] (2/8) Epoch 6, batch 5850, loss[loss=0.2444, simple_loss=0.321, pruned_loss=0.08393, over 15543.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.3289, pruned_loss=0.09017, over 3063007.78 frames. ], batch size: 191, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:36:54,898 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9968, 2.8722, 2.6617, 1.8624, 2.4748, 2.1018, 2.7095, 2.8531], device='cuda:2'), covar=tensor([0.0263, 0.0435, 0.0468, 0.1359, 0.0661, 0.0804, 0.0483, 0.0557], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0126, 0.0152, 0.0140, 0.0132, 0.0125, 0.0139, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 12:37:10,951 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8933, 4.0467, 4.3815, 1.7995, 4.6467, 4.6603, 3.2731, 3.3244], device='cuda:2'), covar=tensor([0.0677, 0.0150, 0.0123, 0.1251, 0.0045, 0.0056, 0.0250, 0.0377], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0092, 0.0080, 0.0139, 0.0072, 0.0080, 0.0114, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 12:37:11,643 INFO [train.py:904] (2/8) Epoch 6, batch 5900, loss[loss=0.255, simple_loss=0.3283, pruned_loss=0.09084, over 17144.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3282, pruned_loss=0.08969, over 3057691.17 frames. ], batch size: 48, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:37:13,025 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 12:37:52,137 INFO [optim.py:368] (2/8) 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 12:38:33,939 INFO [train.py:904] (2/8) Epoch 6, batch 5950, loss[loss=0.2697, simple_loss=0.356, pruned_loss=0.09168, over 16916.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3286, pruned_loss=0.08797, over 3056259.14 frames. ], batch size: 96, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:09,254 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4053, 2.3766, 1.7473, 2.0919, 2.8340, 2.4961, 3.2383, 3.0930], device='cuda:2'), covar=tensor([0.0029, 0.0202, 0.0294, 0.0241, 0.0120, 0.0192, 0.0084, 0.0100], device='cuda:2'), in_proj_covar=tensor([0.0086, 0.0159, 0.0164, 0.0161, 0.0156, 0.0165, 0.0146, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:39:52,035 INFO [train.py:904] (2/8) Epoch 6, batch 6000, loss[loss=0.321, simple_loss=0.3678, pruned_loss=0.1371, over 11407.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3272, pruned_loss=0.08688, over 3088439.33 frames. ], batch size: 247, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:52,036 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 12:40:01,514 INFO [train.py:938] (2/8) Epoch 6, validation: loss=0.18, simple_loss=0.2922, pruned_loss=0.03386, over 944034.00 frames. 2023-04-28 12:40:01,515 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 12:40:36,511 INFO [optim.py:368] (2/8) 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 12:41:19,686 INFO [train.py:904] (2/8) Epoch 6, batch 6050, loss[loss=0.2349, simple_loss=0.3292, pruned_loss=0.07031, over 16757.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3251, pruned_loss=0.08541, over 3105615.30 frames. ], batch size: 83, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:41:20,603 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:41:23,929 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:22,917 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8212, 3.2938, 3.1237, 5.0730, 3.9634, 4.7285, 2.0506, 3.4152], device='cuda:2'), covar=tensor([0.1374, 0.0599, 0.0967, 0.0080, 0.0433, 0.0233, 0.1253, 0.0746], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0146, 0.0171, 0.0096, 0.0200, 0.0191, 0.0165, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 12:42:25,245 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2661, 3.4090, 1.6848, 3.6215, 2.3744, 3.5341, 1.9390, 2.5271], device='cuda:2'), covar=tensor([0.0173, 0.0326, 0.1708, 0.0072, 0.0896, 0.0476, 0.1537, 0.0710], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0154, 0.0175, 0.0081, 0.0163, 0.0188, 0.0186, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 12:42:33,595 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:36,667 INFO [train.py:904] (2/8) Epoch 6, batch 6100, loss[loss=0.2083, simple_loss=0.2952, pruned_loss=0.06072, over 17114.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3244, pruned_loss=0.08403, over 3113033.06 frames. ], batch size: 49, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:42:58,730 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:58,826 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2872, 1.9385, 2.0846, 3.7411, 1.7890, 2.6528, 2.1585, 2.1225], device='cuda:2'), covar=tensor([0.0663, 0.2489, 0.1403, 0.0365, 0.3219, 0.1321, 0.2197, 0.2498], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0341, 0.0280, 0.0315, 0.0387, 0.0352, 0.0306, 0.0404], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:43:13,026 INFO [optim.py:368] (2/8) 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 12:43:20,793 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:43:50,253 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2711, 4.0178, 4.2019, 4.4501, 4.5747, 4.1544, 4.5059, 4.5023], device='cuda:2'), covar=tensor([0.0929, 0.0884, 0.1434, 0.0522, 0.0445, 0.0789, 0.0469, 0.0539], device='cuda:2'), in_proj_covar=tensor([0.0414, 0.0499, 0.0648, 0.0518, 0.0391, 0.0384, 0.0406, 0.0431], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:43:56,260 INFO [train.py:904] (2/8) Epoch 6, batch 6150, loss[loss=0.2258, simple_loss=0.3091, pruned_loss=0.07127, over 16507.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3226, pruned_loss=0.08321, over 3118210.44 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:44:08,594 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:44:26,136 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 12:45:05,675 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5987, 2.6675, 1.7584, 2.7605, 2.2036, 2.7585, 1.9875, 2.3461], device='cuda:2'), covar=tensor([0.0178, 0.0357, 0.1230, 0.0086, 0.0671, 0.0482, 0.1113, 0.0543], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0154, 0.0175, 0.0082, 0.0164, 0.0189, 0.0187, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 12:45:17,264 INFO [train.py:904] (2/8) Epoch 6, batch 6200, loss[loss=0.2568, simple_loss=0.33, pruned_loss=0.09179, over 16306.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3209, pruned_loss=0.08298, over 3113741.13 frames. ], batch size: 165, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:45:45,715 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:48,854 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1358, 1.4330, 1.8799, 2.0441, 2.2875, 2.4248, 1.5107, 2.2125], device='cuda:2'), covar=tensor([0.0098, 0.0237, 0.0144, 0.0172, 0.0115, 0.0083, 0.0247, 0.0054], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0148, 0.0129, 0.0130, 0.0134, 0.0098, 0.0145, 0.0086], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 12:45:53,732 INFO [optim.py:368] (2/8) 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 12:46:22,659 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:26,565 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:34,326 INFO [train.py:904] (2/8) Epoch 6, batch 6250, loss[loss=0.2226, simple_loss=0.3042, pruned_loss=0.07047, over 17129.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3201, pruned_loss=0.08275, over 3110056.02 frames. ], batch size: 48, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:50,596 INFO [train.py:904] (2/8) Epoch 6, batch 6300, loss[loss=0.2215, simple_loss=0.3038, pruned_loss=0.06959, over 16518.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3199, pruned_loss=0.0821, over 3121628.36 frames. ], batch size: 75, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:56,089 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:47:59,212 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:48:29,802 INFO [optim.py:368] (2/8) 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 12:49:09,382 INFO [train.py:904] (2/8) Epoch 6, batch 6350, loss[loss=0.2223, simple_loss=0.3104, pruned_loss=0.06707, over 16928.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3219, pruned_loss=0.08459, over 3086778.49 frames. ], batch size: 96, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:49:25,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2167, 4.2268, 4.7656, 4.7650, 4.7286, 4.2973, 4.3727, 4.1869], device='cuda:2'), covar=tensor([0.0252, 0.0452, 0.0312, 0.0334, 0.0394, 0.0308, 0.0704, 0.0398], device='cuda:2'), in_proj_covar=tensor([0.0256, 0.0249, 0.0252, 0.0249, 0.0301, 0.0269, 0.0371, 0.0222], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 12:49:37,818 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 12:50:26,696 INFO [train.py:904] (2/8) Epoch 6, batch 6400, loss[loss=0.2554, simple_loss=0.3251, pruned_loss=0.09288, over 16718.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3213, pruned_loss=0.08507, over 3108047.97 frames. ], batch size: 124, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:50:38,689 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:50:55,736 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 12:51:01,322 INFO [optim.py:368] (2/8) 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 12:51:07,869 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:51:19,370 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:35,125 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5793, 3.5184, 3.0091, 1.7577, 2.6525, 2.0845, 3.0355, 3.3980], device='cuda:2'), covar=tensor([0.0359, 0.0382, 0.0563, 0.1632, 0.0776, 0.0972, 0.0622, 0.0535], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0126, 0.0152, 0.0137, 0.0132, 0.0124, 0.0138, 0.0137], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 12:51:36,848 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:42,185 INFO [train.py:904] (2/8) Epoch 6, batch 6450, loss[loss=0.2141, simple_loss=0.3049, pruned_loss=0.06167, over 16535.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.3205, pruned_loss=0.08391, over 3102658.46 frames. ], batch size: 68, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:52:26,741 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:52:53,028 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:52:57,221 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:03,828 INFO [train.py:904] (2/8) Epoch 6, batch 6500, loss[loss=0.2382, simple_loss=0.3156, pruned_loss=0.08043, over 16722.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3181, pruned_loss=0.08257, over 3120288.44 frames. ], batch size: 76, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:53:15,574 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:53:25,363 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:41,776 INFO [optim.py:368] (2/8) 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 12:54:25,244 INFO [train.py:904] (2/8) Epoch 6, batch 6550, loss[loss=0.2184, simple_loss=0.3193, pruned_loss=0.05876, over 16823.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3209, pruned_loss=0.08331, over 3117281.38 frames. ], batch size: 102, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:54:31,740 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:54:37,910 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 12:55:40,097 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:55:42,171 INFO [train.py:904] (2/8) Epoch 6, batch 6600, loss[loss=0.327, simple_loss=0.3803, pruned_loss=0.1368, over 11683.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.323, pruned_loss=0.08366, over 3122815.36 frames. ], batch size: 248, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:55:43,245 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:56:18,767 INFO [optim.py:368] (2/8) 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 12:56:56,471 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:57:00,522 INFO [train.py:904] (2/8) Epoch 6, batch 6650, loss[loss=0.2271, simple_loss=0.3091, pruned_loss=0.07259, over 16394.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3234, pruned_loss=0.08432, over 3122805.15 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:57:38,852 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8731, 4.0234, 2.7356, 2.5444, 3.1455, 2.3376, 4.1642, 4.1473], device='cuda:2'), covar=tensor([0.2384, 0.0825, 0.1710, 0.1592, 0.1873, 0.1618, 0.0573, 0.0663], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0247, 0.0265, 0.0251, 0.0282, 0.0201, 0.0248, 0.0258], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 12:57:54,945 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 12:57:56,529 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-28 12:58:15,914 INFO [train.py:904] (2/8) Epoch 6, batch 6700, loss[loss=0.2227, simple_loss=0.3056, pruned_loss=0.0699, over 16563.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3222, pruned_loss=0.08428, over 3121144.03 frames. ], batch size: 68, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:58:29,535 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:29,572 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:34,613 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:54,093 INFO [optim.py:368] (2/8) 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 12:59:14,089 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 12:59:24,857 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:59:34,109 INFO [train.py:904] (2/8) Epoch 6, batch 6750, loss[loss=0.2275, simple_loss=0.3106, pruned_loss=0.07216, over 16801.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3215, pruned_loss=0.08469, over 3110776.12 frames. ], batch size: 124, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:59:38,731 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1275, 1.4828, 1.8494, 2.2917, 2.2613, 2.5037, 1.3823, 2.3368], device='cuda:2'), covar=tensor([0.0111, 0.0272, 0.0176, 0.0144, 0.0138, 0.0106, 0.0278, 0.0049], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0151, 0.0132, 0.0131, 0.0135, 0.0101, 0.0148, 0.0088], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 12:59:43,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:08,067 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:09,432 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7650, 4.7893, 4.7589, 3.2191, 4.3116, 4.7787, 4.2681, 2.7474], device='cuda:2'), covar=tensor([0.0319, 0.0011, 0.0021, 0.0212, 0.0031, 0.0044, 0.0026, 0.0254], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0055, 0.0059, 0.0115, 0.0062, 0.0073, 0.0066, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 13:00:36,603 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:42,249 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1473, 1.8673, 1.9744, 3.7027, 1.8303, 2.4925, 2.0843, 2.0459], device='cuda:2'), covar=tensor([0.0714, 0.2601, 0.1494, 0.0325, 0.3193, 0.1461, 0.2207, 0.2518], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0340, 0.0282, 0.0313, 0.0387, 0.0352, 0.0304, 0.0402], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:00:49,719 INFO [train.py:904] (2/8) Epoch 6, batch 6800, loss[loss=0.282, simple_loss=0.3445, pruned_loss=0.1097, over 11840.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3211, pruned_loss=0.08394, over 3120355.23 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:00:54,463 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 13:00:58,726 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:03,121 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5491, 4.5382, 4.3562, 3.7594, 4.3392, 1.7134, 4.1194, 4.2780], device='cuda:2'), covar=tensor([0.0065, 0.0061, 0.0095, 0.0320, 0.0072, 0.1825, 0.0095, 0.0133], device='cuda:2'), in_proj_covar=tensor([0.0097, 0.0085, 0.0130, 0.0129, 0.0098, 0.0149, 0.0114, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:01:09,878 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:27,504 INFO [optim.py:368] (2/8) 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 13:02:06,744 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:07,777 INFO [train.py:904] (2/8) Epoch 6, batch 6850, loss[loss=0.2142, simple_loss=0.3157, pruned_loss=0.05634, over 16796.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3221, pruned_loss=0.08458, over 3117180.30 frames. ], batch size: 102, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:02:21,577 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6229, 4.6975, 5.0891, 5.0258, 5.0513, 4.6623, 4.5990, 4.4530], device='cuda:2'), covar=tensor([0.0250, 0.0298, 0.0310, 0.0419, 0.0359, 0.0289, 0.0871, 0.0346], device='cuda:2'), in_proj_covar=tensor([0.0254, 0.0245, 0.0254, 0.0251, 0.0299, 0.0270, 0.0373, 0.0219], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 13:02:24,688 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:30,126 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0457, 3.1874, 3.1350, 1.5680, 3.3029, 3.3831, 2.7760, 2.6099], device='cuda:2'), covar=tensor([0.0829, 0.0144, 0.0188, 0.1205, 0.0076, 0.0099, 0.0328, 0.0429], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0091, 0.0082, 0.0139, 0.0072, 0.0081, 0.0116, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 13:03:10,585 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 13:03:12,914 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4929, 4.1574, 3.5987, 1.9144, 3.1848, 2.5698, 3.8492, 4.0845], device='cuda:2'), covar=tensor([0.0205, 0.0401, 0.0591, 0.1673, 0.0702, 0.0906, 0.0559, 0.0582], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0127, 0.0153, 0.0139, 0.0133, 0.0124, 0.0138, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 13:03:22,359 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:24,926 INFO [train.py:904] (2/8) Epoch 6, batch 6900, loss[loss=0.282, simple_loss=0.3513, pruned_loss=0.1064, over 16303.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3241, pruned_loss=0.08405, over 3121388.60 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:03:25,262 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:02,709 INFO [optim.py:368] (2/8) 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 13:04:06,677 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-28 13:04:37,887 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:40,786 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:43,530 INFO [train.py:904] (2/8) Epoch 6, batch 6950, loss[loss=0.2633, simple_loss=0.3322, pruned_loss=0.09722, over 16208.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3274, pruned_loss=0.08745, over 3099055.16 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:05:25,691 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0318, 3.1444, 1.6647, 3.3034, 2.2536, 3.2709, 1.8545, 2.5241], device='cuda:2'), covar=tensor([0.0173, 0.0340, 0.1561, 0.0077, 0.0789, 0.0480, 0.1421, 0.0660], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0154, 0.0177, 0.0083, 0.0161, 0.0190, 0.0187, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 13:05:34,124 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:05:46,052 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 13:06:01,324 INFO [train.py:904] (2/8) Epoch 6, batch 7000, loss[loss=0.2618, simple_loss=0.3264, pruned_loss=0.09858, over 11374.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3269, pruned_loss=0.08641, over 3093620.33 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:06:05,960 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:06:25,257 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 13:06:38,242 INFO [optim.py:368] (2/8) 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 13:07:07,531 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:07:17,559 INFO [train.py:904] (2/8) Epoch 6, batch 7050, loss[loss=0.2143, simple_loss=0.3022, pruned_loss=0.06323, over 16800.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3277, pruned_loss=0.08666, over 3080116.22 frames. ], batch size: 89, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:07:44,505 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:20,805 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:34,157 INFO [train.py:904] (2/8) Epoch 6, batch 7100, loss[loss=0.2762, simple_loss=0.3273, pruned_loss=0.1126, over 11183.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3274, pruned_loss=0.0875, over 3065343.12 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:08:34,651 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:37,906 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 13:09:10,525 INFO [optim.py:368] (2/8) 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 13:09:32,897 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:48,257 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:49,599 INFO [train.py:904] (2/8) Epoch 6, batch 7150, loss[loss=0.2451, simple_loss=0.324, pruned_loss=0.08308, over 16635.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3246, pruned_loss=0.08643, over 3059061.98 frames. ], batch size: 76, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:09:49,969 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:11:00,230 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:11:04,887 INFO [train.py:904] (2/8) Epoch 6, batch 7200, loss[loss=0.2369, simple_loss=0.3192, pruned_loss=0.07726, over 16287.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3227, pruned_loss=0.08526, over 3045646.52 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:11:14,378 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5915, 3.9413, 4.1821, 2.0090, 4.4076, 4.4142, 3.2310, 3.3232], device='cuda:2'), covar=tensor([0.0758, 0.0124, 0.0108, 0.1071, 0.0034, 0.0052, 0.0310, 0.0354], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0091, 0.0082, 0.0140, 0.0072, 0.0081, 0.0116, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 13:11:41,799 INFO [optim.py:368] (2/8) 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 13:12:28,070 INFO [train.py:904] (2/8) Epoch 6, batch 7250, loss[loss=0.2428, simple_loss=0.3201, pruned_loss=0.08275, over 16743.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3198, pruned_loss=0.08337, over 3051804.27 frames. ], batch size: 124, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:13:44,368 INFO [train.py:904] (2/8) Epoch 6, batch 7300, loss[loss=0.2822, simple_loss=0.3669, pruned_loss=0.09878, over 15300.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3188, pruned_loss=0.08285, over 3057093.73 frames. ], batch size: 190, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:13:49,470 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:13:59,218 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-28 13:14:21,942 INFO [optim.py:368] (2/8) 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 13:14:43,340 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:02,126 INFO [train.py:904] (2/8) Epoch 6, batch 7350, loss[loss=0.243, simple_loss=0.3162, pruned_loss=0.08487, over 15425.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3192, pruned_loss=0.08358, over 3031271.26 frames. ], batch size: 190, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:15:03,734 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:28,117 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:49,885 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3651, 4.1368, 4.1374, 2.7276, 3.7514, 4.0605, 3.8181, 2.3088], device='cuda:2'), covar=tensor([0.0306, 0.0015, 0.0024, 0.0220, 0.0043, 0.0076, 0.0034, 0.0274], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0055, 0.0058, 0.0115, 0.0062, 0.0074, 0.0064, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 13:16:04,440 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:09,229 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:19,318 INFO [train.py:904] (2/8) Epoch 6, batch 7400, loss[loss=0.2272, simple_loss=0.3067, pruned_loss=0.07378, over 16455.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3199, pruned_loss=0.08399, over 3036700.32 frames. ], batch size: 146, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:16:19,709 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:43,124 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:45,230 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6667, 3.5157, 3.2112, 5.1311, 4.2138, 4.5966, 1.8474, 3.4458], device='cuda:2'), covar=tensor([0.1397, 0.0508, 0.0912, 0.0105, 0.0360, 0.0300, 0.1349, 0.0703], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0145, 0.0171, 0.0095, 0.0198, 0.0192, 0.0166, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 13:16:46,463 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8087, 2.6074, 2.4695, 1.6522, 2.6531, 2.7271, 2.4251, 2.1898], device='cuda:2'), covar=tensor([0.0927, 0.0157, 0.0185, 0.1079, 0.0116, 0.0150, 0.0413, 0.0554], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0092, 0.0083, 0.0141, 0.0073, 0.0081, 0.0117, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 13:16:57,163 INFO [optim.py:368] (2/8) 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 13:17:34,945 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:37,742 INFO [train.py:904] (2/8) Epoch 6, batch 7450, loss[loss=0.2831, simple_loss=0.3343, pruned_loss=0.116, over 11336.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3207, pruned_loss=0.08473, over 3053082.98 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:17:39,822 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:45,769 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:18:59,296 INFO [train.py:904] (2/8) Epoch 6, batch 7500, loss[loss=0.2347, simple_loss=0.3141, pruned_loss=0.07768, over 16885.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.322, pruned_loss=0.08454, over 3048040.69 frames. ], batch size: 109, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:19:21,791 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-28 13:19:39,440 INFO [optim.py:368] (2/8) 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 13:20:17,495 INFO [train.py:904] (2/8) Epoch 6, batch 7550, loss[loss=0.2212, simple_loss=0.3041, pruned_loss=0.06919, over 17047.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3207, pruned_loss=0.08398, over 3060091.24 frames. ], batch size: 50, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:21:33,626 INFO [train.py:904] (2/8) Epoch 6, batch 7600, loss[loss=0.2145, simple_loss=0.2924, pruned_loss=0.06835, over 16624.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3198, pruned_loss=0.08411, over 3047059.04 frames. ], batch size: 62, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:22:14,310 INFO [optim.py:368] (2/8) 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 13:22:33,750 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:22:34,207 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 13:22:52,233 INFO [train.py:904] (2/8) Epoch 6, batch 7650, loss[loss=0.2588, simple_loss=0.3328, pruned_loss=0.09244, over 16230.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.3207, pruned_loss=0.08539, over 3042552.28 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:23:49,828 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:24:11,081 INFO [train.py:904] (2/8) Epoch 6, batch 7700, loss[loss=0.2923, simple_loss=0.3434, pruned_loss=0.1206, over 11763.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3222, pruned_loss=0.08669, over 3034495.13 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:24:51,994 INFO [optim.py:368] (2/8) 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 13:25:10,146 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3047, 3.9897, 3.9705, 4.5054, 4.6018, 4.1795, 4.6624, 4.5779], device='cuda:2'), covar=tensor([0.1154, 0.1255, 0.2604, 0.0898, 0.0809, 0.1083, 0.0721, 0.0842], device='cuda:2'), in_proj_covar=tensor([0.0413, 0.0501, 0.0645, 0.0513, 0.0395, 0.0387, 0.0410, 0.0437], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:25:23,657 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:28,187 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:29,685 INFO [train.py:904] (2/8) Epoch 6, batch 7750, loss[loss=0.2438, simple_loss=0.324, pruned_loss=0.08184, over 16942.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3221, pruned_loss=0.08642, over 3034646.12 frames. ], batch size: 109, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:26:34,902 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5587, 3.3603, 2.9081, 1.7933, 2.6925, 2.1976, 3.0118, 3.2863], device='cuda:2'), covar=tensor([0.0281, 0.0429, 0.0590, 0.1635, 0.0711, 0.0935, 0.0589, 0.0583], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0128, 0.0152, 0.0140, 0.0133, 0.0125, 0.0138, 0.0139], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 13:26:46,072 INFO [train.py:904] (2/8) Epoch 6, batch 7800, loss[loss=0.2273, simple_loss=0.3132, pruned_loss=0.0707, over 16753.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3232, pruned_loss=0.08703, over 3055453.16 frames. ], batch size: 89, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:26:48,202 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 13:27:26,211 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 6, batch 7850, loss[loss=0.2821, simple_loss=0.3405, pruned_loss=0.1119, over 11804.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3231, pruned_loss=0.08537, over 3081911.37 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:14,706 INFO [train.py:904] (2/8) Epoch 6, batch 7900, loss[loss=0.2336, simple_loss=0.3123, pruned_loss=0.07748, over 17117.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3213, pruned_loss=0.08458, over 3078034.05 frames. ], batch size: 49, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:53,308 INFO [optim.py:368] (2/8) 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 13:30:02,261 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:30:09,600 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 13:30:31,227 INFO [train.py:904] (2/8) Epoch 6, batch 7950, loss[loss=0.2912, simple_loss=0.3423, pruned_loss=0.1201, over 11712.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.322, pruned_loss=0.08491, over 3084577.32 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:31:33,525 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:31:46,443 INFO [train.py:904] (2/8) Epoch 6, batch 8000, loss[loss=0.241, simple_loss=0.3179, pruned_loss=0.08208, over 16685.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3216, pruned_loss=0.08457, over 3099516.24 frames. ], batch size: 134, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:32:25,039 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58775.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:32:27,145 INFO [optim.py:368] (2/8) 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 13:32:57,947 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:02,606 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:03,398 INFO [train.py:904] (2/8) Epoch 6, batch 8050, loss[loss=0.2557, simple_loss=0.3339, pruned_loss=0.0888, over 16389.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3222, pruned_loss=0.08494, over 3084678.17 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:33:26,046 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 13:33:27,640 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 13:33:58,005 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:12,439 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:16,876 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:20,744 INFO [train.py:904] (2/8) Epoch 6, batch 8100, loss[loss=0.2631, simple_loss=0.3372, pruned_loss=0.09451, over 16686.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3209, pruned_loss=0.08355, over 3100065.90 frames. ], batch size: 124, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:03,717 INFO [optim.py:368] (2/8) 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 13:35:15,023 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0290, 3.2888, 3.3458, 1.5915, 3.5595, 3.5592, 2.7100, 2.5707], device='cuda:2'), covar=tensor([0.0838, 0.0122, 0.0154, 0.1209, 0.0052, 0.0079, 0.0384, 0.0443], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0093, 0.0085, 0.0144, 0.0073, 0.0084, 0.0120, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 13:35:38,303 INFO [train.py:904] (2/8) Epoch 6, batch 8150, loss[loss=0.2119, simple_loss=0.2842, pruned_loss=0.06974, over 17238.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3187, pruned_loss=0.08301, over 3107587.91 frames. ], batch size: 52, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:46,400 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:36:16,940 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 13:36:54,139 INFO [train.py:904] (2/8) Epoch 6, batch 8200, loss[loss=0.211, simple_loss=0.3016, pruned_loss=0.06022, over 16867.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3159, pruned_loss=0.08195, over 3105720.27 frames. ], batch size: 96, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:37:19,603 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1100, 1.4111, 1.7344, 2.0295, 2.1439, 2.1646, 1.6056, 2.0050], device='cuda:2'), covar=tensor([0.0090, 0.0254, 0.0150, 0.0154, 0.0144, 0.0116, 0.0225, 0.0078], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0148, 0.0129, 0.0128, 0.0135, 0.0099, 0.0145, 0.0084], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 13:37:21,721 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:37:40,735 INFO [optim.py:368] (2/8) 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 13:38:17,119 INFO [train.py:904] (2/8) Epoch 6, batch 8250, loss[loss=0.2287, simple_loss=0.3188, pruned_loss=0.06933, over 16353.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3153, pruned_loss=0.08, over 3093142.42 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:39:04,716 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 13:39:16,123 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:39:37,601 INFO [train.py:904] (2/8) Epoch 6, batch 8300, loss[loss=0.1983, simple_loss=0.2966, pruned_loss=0.05001, over 16664.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3115, pruned_loss=0.07593, over 3098331.81 frames. ], batch size: 89, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:39:45,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8925, 3.7501, 3.9515, 4.0807, 4.2001, 3.8063, 4.1564, 4.1724], device='cuda:2'), covar=tensor([0.1027, 0.0789, 0.1176, 0.0604, 0.0509, 0.1135, 0.0467, 0.0474], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0485, 0.0617, 0.0498, 0.0383, 0.0373, 0.0396, 0.0425], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:40:22,347 INFO [optim.py:368] (2/8) 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 13:40:59,218 INFO [train.py:904] (2/8) Epoch 6, batch 8350, loss[loss=0.2142, simple_loss=0.3036, pruned_loss=0.06238, over 16108.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3102, pruned_loss=0.07358, over 3094459.36 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:41:04,424 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7595, 5.0072, 4.7878, 4.8178, 4.4184, 4.3148, 4.5290, 5.0665], device='cuda:2'), covar=tensor([0.0740, 0.0744, 0.0867, 0.0489, 0.0630, 0.0896, 0.0678, 0.0686], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0514, 0.0443, 0.0336, 0.0318, 0.0345, 0.0422, 0.0374], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:41:49,264 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:41:55,915 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 13:42:21,175 INFO [train.py:904] (2/8) Epoch 6, batch 8400, loss[loss=0.2251, simple_loss=0.3096, pruned_loss=0.07034, over 16212.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3067, pruned_loss=0.07106, over 3068177.31 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:42:42,282 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:42:51,398 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7210, 2.7237, 1.6297, 2.8473, 2.0835, 2.8172, 1.7698, 2.3455], device='cuda:2'), covar=tensor([0.0190, 0.0277, 0.1383, 0.0096, 0.0680, 0.0398, 0.1352, 0.0503], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0146, 0.0172, 0.0081, 0.0155, 0.0179, 0.0184, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 13:43:05,301 INFO [optim.py:368] (2/8) 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 13:43:40,726 INFO [train.py:904] (2/8) Epoch 6, batch 8450, loss[loss=0.1961, simple_loss=0.2864, pruned_loss=0.05294, over 15444.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3053, pruned_loss=0.06962, over 3057541.73 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:44:01,236 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5415, 1.4409, 2.0021, 2.4419, 2.4091, 2.5645, 1.5305, 2.4658], device='cuda:2'), covar=tensor([0.0095, 0.0306, 0.0169, 0.0141, 0.0135, 0.0123, 0.0301, 0.0089], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0150, 0.0131, 0.0130, 0.0137, 0.0099, 0.0148, 0.0085], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 13:44:19,040 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:45:00,570 INFO [train.py:904] (2/8) Epoch 6, batch 8500, loss[loss=0.1809, simple_loss=0.2543, pruned_loss=0.05376, over 11590.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3008, pruned_loss=0.06646, over 3054048.10 frames. ], batch size: 246, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:45:19,656 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:45:46,051 INFO [optim.py:368] (2/8) 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 13:45:49,789 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5410, 3.5057, 3.4814, 2.9919, 3.4144, 2.0353, 3.2275, 2.9844], device='cuda:2'), covar=tensor([0.0087, 0.0072, 0.0095, 0.0172, 0.0068, 0.1612, 0.0089, 0.0140], device='cuda:2'), in_proj_covar=tensor([0.0094, 0.0081, 0.0127, 0.0122, 0.0096, 0.0149, 0.0110, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:46:25,537 INFO [train.py:904] (2/8) Epoch 6, batch 8550, loss[loss=0.2107, simple_loss=0.2968, pruned_loss=0.06225, over 16877.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2974, pruned_loss=0.06451, over 3044216.25 frames. ], batch size: 109, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:47:25,455 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8843, 2.5104, 2.2866, 3.1734, 2.4202, 3.4140, 1.5383, 2.7275], device='cuda:2'), covar=tensor([0.1188, 0.0464, 0.0906, 0.0102, 0.0125, 0.0325, 0.1341, 0.0637], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0142, 0.0167, 0.0093, 0.0182, 0.0186, 0.0164, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 13:47:38,199 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:48:07,961 INFO [train.py:904] (2/8) Epoch 6, batch 8600, loss[loss=0.2153, simple_loss=0.3065, pruned_loss=0.06205, over 16429.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2977, pruned_loss=0.06363, over 3055064.33 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:49:03,383 INFO [optim.py:368] (2/8) 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 13:49:16,114 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:49:46,455 INFO [train.py:904] (2/8) Epoch 6, batch 8650, loss[loss=0.1885, simple_loss=0.2806, pruned_loss=0.04824, over 15250.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2948, pruned_loss=0.06147, over 3055477.77 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:50:55,034 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:51:31,167 INFO [train.py:904] (2/8) Epoch 6, batch 8700, loss[loss=0.2297, simple_loss=0.3232, pruned_loss=0.06809, over 15343.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2916, pruned_loss=0.05972, over 3057976.44 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:51:52,756 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-28 13:52:04,636 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 13:52:21,220 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:05,572 INFO [train.py:904] (2/8) Epoch 6, batch 8750, loss[loss=0.2213, simple_loss=0.3125, pruned_loss=0.06508, over 15335.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2919, pruned_loss=0.05927, over 3065448.63 frames. ], batch size: 191, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:53:08,811 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:16,829 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 13:53:54,431 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:54,618 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7355, 2.1121, 1.7453, 1.7942, 2.3746, 2.1635, 2.5580, 2.5624], device='cuda:2'), covar=tensor([0.0040, 0.0242, 0.0306, 0.0306, 0.0141, 0.0238, 0.0092, 0.0156], device='cuda:2'), in_proj_covar=tensor([0.0084, 0.0160, 0.0161, 0.0161, 0.0156, 0.0161, 0.0139, 0.0141], device='cuda:2'), out_proj_covar=tensor([9.7495e-05, 1.8630e-04, 1.8179e-04, 1.8332e-04, 1.8221e-04, 1.8641e-04, 1.5621e-04, 1.6268e-04], device='cuda:2') 2023-04-28 13:54:55,917 INFO [train.py:904] (2/8) Epoch 6, batch 8800, loss[loss=0.2045, simple_loss=0.2948, pruned_loss=0.05711, over 16894.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2909, pruned_loss=0.05861, over 3071888.35 frames. ], batch size: 116, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:55:17,760 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:19,837 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:52,197 INFO [optim.py:368] (2/8) 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 13:55:55,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9483, 4.2318, 4.0571, 4.0511, 3.7705, 3.7514, 3.8893, 4.1871], device='cuda:2'), covar=tensor([0.0767, 0.0758, 0.0786, 0.0447, 0.0551, 0.1426, 0.0657, 0.0914], device='cuda:2'), in_proj_covar=tensor([0.0383, 0.0500, 0.0416, 0.0325, 0.0307, 0.0331, 0.0408, 0.0362], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:56:37,975 INFO [train.py:904] (2/8) Epoch 6, batch 8850, loss[loss=0.1929, simple_loss=0.2918, pruned_loss=0.04701, over 17034.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2925, pruned_loss=0.05745, over 3050328.33 frames. ], batch size: 55, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:56:56,206 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:56:56,313 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.8273, 6.2262, 5.9108, 6.0433, 5.5069, 5.4420, 5.7512, 6.2928], device='cuda:2'), covar=tensor([0.0678, 0.0737, 0.0859, 0.0488, 0.0705, 0.0543, 0.0665, 0.0748], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0498, 0.0415, 0.0323, 0.0305, 0.0331, 0.0407, 0.0361], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 13:58:21,074 INFO [train.py:904] (2/8) Epoch 6, batch 8900, loss[loss=0.2187, simple_loss=0.3069, pruned_loss=0.06525, over 15325.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.293, pruned_loss=0.05663, over 3072984.49 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:59:05,516 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 13:59:22,403 INFO [optim.py:368] (2/8) 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 14:00:23,135 INFO [train.py:904] (2/8) Epoch 6, batch 8950, loss[loss=0.1902, simple_loss=0.281, pruned_loss=0.04971, over 15389.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2927, pruned_loss=0.05678, over 3090183.04 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:01:00,320 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7323, 3.3931, 3.1485, 1.7881, 2.7277, 2.1930, 3.1415, 3.3001], device='cuda:2'), covar=tensor([0.0244, 0.0481, 0.0520, 0.1675, 0.0727, 0.0953, 0.0677, 0.0621], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0120, 0.0149, 0.0139, 0.0129, 0.0123, 0.0134, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 14:01:15,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9896, 1.6002, 2.1459, 3.0168, 2.7231, 3.0465, 1.8793, 3.1220], device='cuda:2'), covar=tensor([0.0078, 0.0296, 0.0179, 0.0098, 0.0114, 0.0089, 0.0277, 0.0070], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0150, 0.0132, 0.0129, 0.0138, 0.0096, 0.0147, 0.0083], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 14:02:12,299 INFO [train.py:904] (2/8) Epoch 6, batch 9000, loss[loss=0.1863, simple_loss=0.2729, pruned_loss=0.04988, over 11781.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2886, pruned_loss=0.05469, over 3089039.03 frames. ], batch size: 246, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:02:12,300 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 14:02:22,185 INFO [train.py:938] (2/8) Epoch 6, validation: loss=0.1682, simple_loss=0.2716, pruned_loss=0.03235, over 944034.00 frames. 2023-04-28 14:02:22,186 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 14:02:48,558 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 14:03:21,581 INFO [optim.py:368] (2/8) 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 14:03:28,572 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-28 14:03:40,397 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3952, 1.8995, 1.6284, 1.5095, 2.1539, 1.8639, 2.1708, 2.3203], device='cuda:2'), covar=tensor([0.0036, 0.0217, 0.0257, 0.0269, 0.0125, 0.0211, 0.0099, 0.0115], device='cuda:2'), in_proj_covar=tensor([0.0084, 0.0167, 0.0167, 0.0166, 0.0162, 0.0167, 0.0145, 0.0146], device='cuda:2'), out_proj_covar=tensor([9.6808e-05, 1.9402e-04, 1.8800e-04, 1.8841e-04, 1.8874e-04, 1.9331e-04, 1.6194e-04, 1.6811e-04], device='cuda:2') 2023-04-28 14:03:50,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3675, 3.2238, 3.0982, 3.4620, 3.4547, 3.2123, 3.4841, 3.5248], device='cuda:2'), covar=tensor([0.0883, 0.0827, 0.1550, 0.0726, 0.0847, 0.2333, 0.0847, 0.0721], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0466, 0.0588, 0.0476, 0.0364, 0.0358, 0.0375, 0.0404], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:04:06,172 INFO [train.py:904] (2/8) Epoch 6, batch 9050, loss[loss=0.1862, simple_loss=0.2703, pruned_loss=0.05107, over 16707.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2894, pruned_loss=0.05535, over 3101459.28 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:04:44,581 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:03,802 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5341, 3.5205, 3.4806, 3.0904, 3.4333, 1.9509, 3.2436, 2.9851], device='cuda:2'), covar=tensor([0.0093, 0.0080, 0.0130, 0.0189, 0.0066, 0.1641, 0.0104, 0.0148], device='cuda:2'), in_proj_covar=tensor([0.0093, 0.0078, 0.0125, 0.0117, 0.0093, 0.0147, 0.0108, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:05:45,948 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:50,583 INFO [train.py:904] (2/8) Epoch 6, batch 9100, loss[loss=0.2032, simple_loss=0.2983, pruned_loss=0.05406, over 16204.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.29, pruned_loss=0.05643, over 3094870.41 frames. ], batch size: 166, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:06:04,213 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:24,268 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:55,730 INFO [optim.py:368] (2/8) 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 14:07:19,922 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:07:47,001 INFO [train.py:904] (2/8) Epoch 6, batch 9150, loss[loss=0.1884, simple_loss=0.2754, pruned_loss=0.05071, over 15341.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2904, pruned_loss=0.05636, over 3094645.13 frames. ], batch size: 190, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:08:06,763 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:09:29,926 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:09:31,017 INFO [train.py:904] (2/8) Epoch 6, batch 9200, loss[loss=0.1675, simple_loss=0.2563, pruned_loss=0.03934, over 16586.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2854, pruned_loss=0.05496, over 3077432.99 frames. ], batch size: 57, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:10:21,760 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 14:10:22,279 INFO [optim.py:368] (2/8) 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 14:11:09,067 INFO [train.py:904] (2/8) Epoch 6, batch 9250, loss[loss=0.1791, simple_loss=0.2624, pruned_loss=0.04787, over 12293.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2857, pruned_loss=0.05506, over 3089757.53 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:12:33,896 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:12:58,476 INFO [train.py:904] (2/8) Epoch 6, batch 9300, loss[loss=0.1665, simple_loss=0.2561, pruned_loss=0.03844, over 16702.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2841, pruned_loss=0.05416, over 3097982.24 frames. ], batch size: 83, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:13:01,066 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:10,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7556, 5.0692, 5.0872, 5.0333, 5.0994, 5.6044, 5.0052, 4.7533], device='cuda:2'), covar=tensor([0.0629, 0.1373, 0.1124, 0.1692, 0.2048, 0.0771, 0.1149, 0.2072], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0377, 0.0385, 0.0328, 0.0431, 0.0416, 0.0316, 0.0441], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:13:33,135 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:14:01,610 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 14:14:20,201 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 14:14:44,854 INFO [train.py:904] (2/8) Epoch 6, batch 9350, loss[loss=0.1964, simple_loss=0.2819, pruned_loss=0.05546, over 16707.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2833, pruned_loss=0.05392, over 3107268.07 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:14:45,699 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 14:15:12,248 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 14:15:36,716 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:15:42,083 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 14:16:03,911 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:16:16,684 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5111, 4.6646, 4.7960, 4.7581, 4.7631, 5.2741, 4.8867, 4.6188], device='cuda:2'), covar=tensor([0.0868, 0.1498, 0.1484, 0.1386, 0.2106, 0.0950, 0.1064, 0.1931], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0379, 0.0384, 0.0327, 0.0429, 0.0417, 0.0316, 0.0439], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:16:25,945 INFO [train.py:904] (2/8) Epoch 6, batch 9400, loss[loss=0.2041, simple_loss=0.2939, pruned_loss=0.05715, over 15249.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2836, pruned_loss=0.0538, over 3093871.91 frames. ], batch size: 190, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:16:34,279 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2250, 3.9880, 4.2751, 4.3785, 4.5270, 4.0552, 4.5212, 4.4917], device='cuda:2'), covar=tensor([0.0965, 0.0816, 0.1088, 0.0506, 0.0420, 0.0757, 0.0411, 0.0414], device='cuda:2'), in_proj_covar=tensor([0.0377, 0.0462, 0.0583, 0.0474, 0.0359, 0.0350, 0.0369, 0.0400], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:16:40,716 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 14:17:51,845 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9435, 1.9895, 2.2701, 3.2514, 2.0398, 2.3394, 2.2340, 1.9388], device='cuda:2'), covar=tensor([0.0690, 0.2628, 0.1373, 0.0393, 0.3224, 0.1541, 0.2185, 0.3063], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0326, 0.0278, 0.0303, 0.0377, 0.0342, 0.0300, 0.0382], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:18:05,620 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:09,620 INFO [train.py:904] (2/8) Epoch 6, batch 9450, loss[loss=0.1892, simple_loss=0.2806, pruned_loss=0.04891, over 15321.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2861, pruned_loss=0.05462, over 3096797.80 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:18:14,585 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:19,278 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:19:29,235 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 14:19:39,982 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:19:50,584 INFO [train.py:904] (2/8) Epoch 6, batch 9500, loss[loss=0.1979, simple_loss=0.2784, pruned_loss=0.05866, over 16940.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2849, pruned_loss=0.05429, over 3067636.51 frames. ], batch size: 109, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:20:11,545 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:20:44,379 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-28 14:20:51,250 INFO [optim.py:368] (2/8) 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 14:21:39,640 INFO [train.py:904] (2/8) Epoch 6, batch 9550, loss[loss=0.2024, simple_loss=0.2946, pruned_loss=0.05508, over 16714.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2845, pruned_loss=0.05395, over 3075830.05 frames. ], batch size: 76, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:22:11,252 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9357, 2.0253, 2.2641, 3.2765, 2.0371, 2.3853, 2.2520, 1.9694], device='cuda:2'), covar=tensor([0.0724, 0.2482, 0.1276, 0.0399, 0.3106, 0.1560, 0.2077, 0.3055], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0320, 0.0273, 0.0298, 0.0372, 0.0336, 0.0294, 0.0376], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:23:21,496 INFO [train.py:904] (2/8) Epoch 6, batch 9600, loss[loss=0.2329, simple_loss=0.3242, pruned_loss=0.07081, over 16819.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2865, pruned_loss=0.05509, over 3088941.98 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:24:17,216 INFO [optim.py:368] (2/8) 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 14:24:38,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9077, 3.7862, 3.8299, 3.1411, 3.7102, 1.6640, 3.5980, 3.5994], device='cuda:2'), covar=tensor([0.0096, 0.0093, 0.0117, 0.0300, 0.0103, 0.2247, 0.0128, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0093, 0.0080, 0.0124, 0.0115, 0.0094, 0.0150, 0.0109, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:24:57,310 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 14:25:03,049 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9076, 1.6583, 1.4344, 1.3045, 1.8393, 1.5288, 1.8083, 1.8337], device='cuda:2'), covar=tensor([0.0042, 0.0188, 0.0270, 0.0237, 0.0127, 0.0201, 0.0097, 0.0133], device='cuda:2'), in_proj_covar=tensor([0.0082, 0.0162, 0.0163, 0.0161, 0.0158, 0.0162, 0.0140, 0.0143], device='cuda:2'), out_proj_covar=tensor([9.3324e-05, 1.8785e-04, 1.8359e-04, 1.8216e-04, 1.8382e-04, 1.8748e-04, 1.5532e-04, 1.6388e-04], device='cuda:2') 2023-04-28 14:25:09,700 INFO [train.py:904] (2/8) Epoch 6, batch 9650, loss[loss=0.1817, simple_loss=0.2682, pruned_loss=0.04755, over 12212.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.289, pruned_loss=0.0556, over 3095771.97 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:25:28,107 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 14:25:57,582 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:15,577 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1679, 2.2686, 1.8979, 1.9409, 2.7775, 2.4821, 3.0057, 2.9142], device='cuda:2'), covar=tensor([0.0032, 0.0238, 0.0299, 0.0303, 0.0144, 0.0199, 0.0117, 0.0121], device='cuda:2'), in_proj_covar=tensor([0.0081, 0.0162, 0.0163, 0.0162, 0.0158, 0.0162, 0.0140, 0.0143], device='cuda:2'), out_proj_covar=tensor([9.3104e-05, 1.8741e-04, 1.8364e-04, 1.8292e-04, 1.8366e-04, 1.8752e-04, 1.5531e-04, 1.6452e-04], device='cuda:2') 2023-04-28 14:26:22,134 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:58,005 INFO [train.py:904] (2/8) Epoch 6, batch 9700, loss[loss=0.2016, simple_loss=0.2796, pruned_loss=0.06178, over 12145.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2875, pruned_loss=0.05503, over 3089175.14 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:27:59,429 INFO [optim.py:368] (2/8) 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 14:28:25,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4802, 4.7929, 4.5950, 4.6313, 4.2611, 4.1755, 4.2693, 4.8489], device='cuda:2'), covar=tensor([0.0829, 0.0807, 0.0839, 0.0459, 0.0646, 0.1091, 0.0848, 0.0769], device='cuda:2'), in_proj_covar=tensor([0.0382, 0.0497, 0.0410, 0.0324, 0.0310, 0.0331, 0.0411, 0.0364], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:28:41,228 INFO [train.py:904] (2/8) Epoch 6, batch 9750, loss[loss=0.1992, simple_loss=0.2899, pruned_loss=0.05426, over 16872.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2856, pruned_loss=0.05484, over 3084703.92 frames. ], batch size: 116, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:28:47,797 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:29:04,549 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 2023-04-28 14:30:11,480 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:19,566 INFO [train.py:904] (2/8) Epoch 6, batch 9800, loss[loss=0.2189, simple_loss=0.3221, pruned_loss=0.05784, over 16774.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.285, pruned_loss=0.05332, over 3101718.52 frames. ], batch size: 83, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:30:22,402 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:27,284 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:31:14,367 INFO [optim.py:368] (2/8) 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 14:31:48,143 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:31:53,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8831, 4.8636, 5.5353, 5.4385, 5.4388, 5.0626, 4.9963, 4.6346], device='cuda:2'), covar=tensor([0.0263, 0.0471, 0.0268, 0.0442, 0.0388, 0.0285, 0.0819, 0.0367], device='cuda:2'), in_proj_covar=tensor([0.0244, 0.0239, 0.0244, 0.0237, 0.0281, 0.0259, 0.0348, 0.0210], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 14:32:03,896 INFO [train.py:904] (2/8) Epoch 6, batch 9850, loss[loss=0.2013, simple_loss=0.2812, pruned_loss=0.06063, over 12416.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2862, pruned_loss=0.05323, over 3076705.74 frames. ], batch size: 247, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:33:56,848 INFO [train.py:904] (2/8) Epoch 6, batch 9900, loss[loss=0.1928, simple_loss=0.293, pruned_loss=0.04628, over 16494.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2856, pruned_loss=0.0527, over 3049830.46 frames. ], batch size: 147, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:35:04,552 INFO [optim.py:368] (2/8) 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 14:35:13,894 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9234, 3.8773, 4.3715, 4.3501, 4.2866, 4.0073, 4.0175, 3.9100], device='cuda:2'), covar=tensor([0.0277, 0.0453, 0.0303, 0.0326, 0.0390, 0.0318, 0.0733, 0.0369], device='cuda:2'), in_proj_covar=tensor([0.0245, 0.0239, 0.0246, 0.0236, 0.0282, 0.0260, 0.0349, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 14:35:16,617 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 14:35:42,389 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:35:52,850 INFO [train.py:904] (2/8) Epoch 6, batch 9950, loss[loss=0.1989, simple_loss=0.2924, pruned_loss=0.05264, over 15326.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2872, pruned_loss=0.05294, over 3045327.19 frames. ], batch size: 190, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:36:09,327 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:34,712 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:42,486 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:53,689 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9923, 3.9985, 4.3747, 4.3440, 4.4161, 4.1351, 3.9479, 3.9931], device='cuda:2'), covar=tensor([0.0361, 0.0622, 0.0476, 0.0560, 0.0522, 0.0432, 0.1061, 0.0434], device='cuda:2'), in_proj_covar=tensor([0.0244, 0.0239, 0.0246, 0.0236, 0.0280, 0.0260, 0.0348, 0.0213], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 14:37:10,323 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 14:37:13,207 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:38,497 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:54,650 INFO [train.py:904] (2/8) Epoch 6, batch 10000, loss[loss=0.2267, simple_loss=0.2992, pruned_loss=0.07705, over 12470.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2864, pruned_loss=0.05323, over 3048298.09 frames. ], batch size: 248, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:38:06,642 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:29,850 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:50,617 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:51,219 INFO [optim.py:368] (2/8) 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 14:38:56,774 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:39:33,417 INFO [train.py:904] (2/8) Epoch 6, batch 10050, loss[loss=0.1912, simple_loss=0.2836, pruned_loss=0.04934, over 16467.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2868, pruned_loss=0.05323, over 3055693.85 frames. ], batch size: 68, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:40:37,480 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0057, 5.1469, 5.6060, 5.5489, 5.5516, 5.2414, 5.1575, 4.8782], device='cuda:2'), covar=tensor([0.0241, 0.0345, 0.0286, 0.0339, 0.0301, 0.0253, 0.0564, 0.0341], device='cuda:2'), in_proj_covar=tensor([0.0244, 0.0238, 0.0245, 0.0235, 0.0278, 0.0258, 0.0345, 0.0214], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 14:40:38,969 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9650, 4.2349, 3.9997, 4.1011, 3.7297, 3.7910, 3.8933, 4.1878], device='cuda:2'), covar=tensor([0.0743, 0.0727, 0.0809, 0.0440, 0.0608, 0.1322, 0.0686, 0.0855], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0501, 0.0412, 0.0328, 0.0308, 0.0332, 0.0412, 0.0365], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:41:01,386 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7259, 2.8035, 1.7971, 2.9103, 2.1583, 2.9069, 2.0408, 2.4930], device='cuda:2'), covar=tensor([0.0172, 0.0348, 0.1234, 0.0112, 0.0701, 0.0399, 0.1171, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0146, 0.0176, 0.0082, 0.0155, 0.0177, 0.0186, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 14:41:05,273 INFO [train.py:904] (2/8) Epoch 6, batch 10100, loss[loss=0.224, simple_loss=0.2928, pruned_loss=0.07763, over 12637.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2875, pruned_loss=0.0537, over 3057144.38 frames. ], batch size: 246, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:41:10,985 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:42:01,260 INFO [optim.py:368] (2/8) 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 14:42:22,731 INFO [train.py:904] (2/8) Epoch 6, batch 10150, loss[loss=0.2027, simple_loss=0.2792, pruned_loss=0.06313, over 12404.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2856, pruned_loss=0.05375, over 3038506.94 frames. ], batch size: 248, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:42:48,322 INFO [train.py:904] (2/8) Epoch 7, batch 0, loss[loss=0.1814, simple_loss=0.263, pruned_loss=0.04991, over 16844.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.263, pruned_loss=0.04991, over 16844.00 frames. ], batch size: 42, lr: 1.02e-02, grad_scale: 8.0 2023-04-28 14:42:48,322 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 14:42:55,775 INFO [train.py:938] (2/8) Epoch 7, validation: loss=0.1665, simple_loss=0.2702, pruned_loss=0.03141, over 944034.00 frames. 2023-04-28 14:42:55,775 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 14:42:55,982 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:43:20,646 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7562, 5.1667, 5.2796, 5.1761, 5.1875, 5.7610, 5.3807, 5.0630], device='cuda:2'), covar=tensor([0.1068, 0.1998, 0.1840, 0.1970, 0.2664, 0.1141, 0.1391, 0.2584], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0393, 0.0398, 0.0337, 0.0447, 0.0432, 0.0330, 0.0456], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:44:05,497 INFO [train.py:904] (2/8) Epoch 7, batch 50, loss[loss=0.2056, simple_loss=0.2826, pruned_loss=0.06428, over 17214.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3017, pruned_loss=0.07367, over 751408.78 frames. ], batch size: 44, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:44:23,291 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1170, 4.0498, 4.0235, 3.5366, 4.0235, 1.6807, 3.8387, 3.6623], device='cuda:2'), covar=tensor([0.0081, 0.0069, 0.0115, 0.0214, 0.0070, 0.1986, 0.0097, 0.0140], device='cuda:2'), in_proj_covar=tensor([0.0098, 0.0083, 0.0131, 0.0119, 0.0097, 0.0155, 0.0114, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:44:25,385 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-28 14:44:49,441 INFO [optim.py:368] (2/8) 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 14:45:15,337 INFO [train.py:904] (2/8) Epoch 7, batch 100, loss[loss=0.2253, simple_loss=0.2829, pruned_loss=0.08382, over 16928.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.2982, pruned_loss=0.07249, over 1312629.14 frames. ], batch size: 109, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:46:24,664 INFO [train.py:904] (2/8) Epoch 7, batch 150, loss[loss=0.2451, simple_loss=0.3042, pruned_loss=0.09295, over 16866.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2949, pruned_loss=0.07098, over 1763237.70 frames. ], batch size: 116, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:46:25,065 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1487, 2.9826, 3.5558, 2.3574, 3.2944, 3.4789, 3.4597, 2.1988], device='cuda:2'), covar=tensor([0.0281, 0.0118, 0.0030, 0.0221, 0.0046, 0.0048, 0.0039, 0.0250], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0058, 0.0059, 0.0116, 0.0062, 0.0073, 0.0065, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 14:46:37,312 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:46:56,417 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:47:02,270 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9600, 3.5815, 3.5199, 4.0258, 4.1617, 3.8022, 3.9490, 4.1345], device='cuda:2'), covar=tensor([0.0970, 0.1056, 0.2301, 0.0961, 0.0717, 0.1347, 0.1282, 0.0787], device='cuda:2'), in_proj_covar=tensor([0.0417, 0.0511, 0.0641, 0.0512, 0.0389, 0.0386, 0.0407, 0.0439], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:47:07,162 INFO [optim.py:368] (2/8) 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 14:47:34,196 INFO [train.py:904] (2/8) Epoch 7, batch 200, loss[loss=0.2316, simple_loss=0.302, pruned_loss=0.08058, over 15697.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2958, pruned_loss=0.07271, over 2108530.28 frames. ], batch size: 191, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:01,676 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:48:44,071 INFO [train.py:904] (2/8) Epoch 7, batch 250, loss[loss=0.1823, simple_loss=0.2885, pruned_loss=0.03805, over 17037.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2932, pruned_loss=0.0709, over 2379207.00 frames. ], batch size: 50, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:50,842 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1035, 4.7442, 5.0515, 5.2697, 5.5107, 4.8137, 5.4287, 5.4212], device='cuda:2'), covar=tensor([0.1333, 0.0959, 0.1509, 0.0583, 0.0487, 0.0628, 0.0459, 0.0409], device='cuda:2'), in_proj_covar=tensor([0.0428, 0.0525, 0.0657, 0.0523, 0.0398, 0.0396, 0.0417, 0.0449], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:49:24,839 INFO [optim.py:368] (2/8) 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 14:49:52,015 INFO [train.py:904] (2/8) Epoch 7, batch 300, loss[loss=0.2285, simple_loss=0.2852, pruned_loss=0.08596, over 16716.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2894, pruned_loss=0.06874, over 2591401.91 frames. ], batch size: 83, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:50:59,340 INFO [train.py:904] (2/8) Epoch 7, batch 350, loss[loss=0.2052, simple_loss=0.2954, pruned_loss=0.05752, over 17128.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2869, pruned_loss=0.06678, over 2749923.53 frames. ], batch size: 48, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:51:03,823 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8625, 4.7951, 4.6750, 4.4830, 4.2425, 4.7384, 4.6795, 4.3957], device='cuda:2'), covar=tensor([0.0477, 0.0341, 0.0232, 0.0203, 0.0843, 0.0344, 0.0311, 0.0488], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0228, 0.0239, 0.0211, 0.0273, 0.0244, 0.0169, 0.0274], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 14:51:05,693 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6337, 3.2527, 2.5400, 4.9773, 4.1808, 4.5208, 1.6931, 2.8993], device='cuda:2'), covar=tensor([0.1499, 0.0648, 0.1380, 0.0137, 0.0335, 0.0343, 0.1521, 0.0929], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0143, 0.0169, 0.0097, 0.0176, 0.0193, 0.0164, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 14:51:20,127 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-28 14:51:41,729 INFO [optim.py:368] (2/8) 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 14:52:07,328 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5234, 2.0652, 2.2710, 4.1641, 1.9529, 2.7938, 2.1601, 2.2905], device='cuda:2'), covar=tensor([0.0695, 0.2604, 0.1429, 0.0344, 0.3109, 0.1489, 0.2410, 0.2286], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0345, 0.0288, 0.0317, 0.0386, 0.0368, 0.0311, 0.0408], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:52:08,571 INFO [train.py:904] (2/8) Epoch 7, batch 400, loss[loss=0.2299, simple_loss=0.291, pruned_loss=0.08443, over 16869.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2855, pruned_loss=0.06594, over 2875615.37 frames. ], batch size: 109, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:53:17,211 INFO [train.py:904] (2/8) Epoch 7, batch 450, loss[loss=0.2196, simple_loss=0.2862, pruned_loss=0.07648, over 16593.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2841, pruned_loss=0.0655, over 2983430.55 frames. ], batch size: 75, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:53:48,649 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:54:00,435 INFO [optim.py:368] (2/8) 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 14:54:27,748 INFO [train.py:904] (2/8) Epoch 7, batch 500, loss[loss=0.2048, simple_loss=0.2941, pruned_loss=0.05772, over 17146.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2823, pruned_loss=0.0647, over 3066810.25 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:54:47,691 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:55:31,952 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 14:55:35,821 INFO [train.py:904] (2/8) Epoch 7, batch 550, loss[loss=0.214, simple_loss=0.2919, pruned_loss=0.06803, over 17103.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2814, pruned_loss=0.06431, over 3117092.56 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:56:17,396 INFO [optim.py:368] (2/8) 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 14:56:43,946 INFO [train.py:904] (2/8) Epoch 7, batch 600, loss[loss=0.2185, simple_loss=0.2862, pruned_loss=0.07543, over 12301.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2805, pruned_loss=0.0636, over 3157378.00 frames. ], batch size: 247, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:57:27,440 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.47 vs. limit=5.0 2023-04-28 14:57:39,353 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 14:57:53,638 INFO [train.py:904] (2/8) Epoch 7, batch 650, loss[loss=0.2056, simple_loss=0.2707, pruned_loss=0.07023, over 11981.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2783, pruned_loss=0.06276, over 3195187.82 frames. ], batch size: 246, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:58:02,479 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2050, 3.5931, 3.3871, 1.9870, 2.8660, 2.1920, 3.5927, 3.5674], device='cuda:2'), covar=tensor([0.0209, 0.0578, 0.0520, 0.1548, 0.0691, 0.0980, 0.0429, 0.0742], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0129, 0.0154, 0.0140, 0.0132, 0.0124, 0.0137, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 14:58:30,435 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2649, 4.2279, 4.1710, 3.6370, 4.2003, 1.6930, 3.9638, 3.8923], device='cuda:2'), covar=tensor([0.0070, 0.0062, 0.0100, 0.0230, 0.0065, 0.1987, 0.0092, 0.0139], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0089, 0.0140, 0.0131, 0.0105, 0.0158, 0.0122, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 14:58:30,608 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 14:58:35,761 INFO [optim.py:368] (2/8) 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 14:59:01,942 INFO [train.py:904] (2/8) Epoch 7, batch 700, loss[loss=0.2228, simple_loss=0.2915, pruned_loss=0.07702, over 16268.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2792, pruned_loss=0.06289, over 3219293.19 frames. ], batch size: 164, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:59:47,304 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6598, 3.9696, 3.0311, 2.4234, 2.8995, 2.3273, 4.0845, 3.8432], device='cuda:2'), covar=tensor([0.2243, 0.0688, 0.1336, 0.1793, 0.2139, 0.1513, 0.0432, 0.0800], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0255, 0.0273, 0.0256, 0.0275, 0.0211, 0.0252, 0.0272], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:00:08,122 INFO [train.py:904] (2/8) Epoch 7, batch 750, loss[loss=0.1946, simple_loss=0.2773, pruned_loss=0.05599, over 17182.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.28, pruned_loss=0.06315, over 3246521.44 frames. ], batch size: 46, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 15:00:51,555 INFO [optim.py:368] (2/8) 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 15:01:16,033 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2047, 5.6148, 5.2023, 5.5417, 5.0015, 4.5062, 5.2842, 5.7717], device='cuda:2'), covar=tensor([0.1861, 0.1470, 0.2377, 0.0953, 0.1359, 0.1485, 0.1566, 0.1346], device='cuda:2'), in_proj_covar=tensor([0.0442, 0.0570, 0.0479, 0.0382, 0.0357, 0.0374, 0.0476, 0.0419], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:01:16,767 INFO [train.py:904] (2/8) Epoch 7, batch 800, loss[loss=0.1997, simple_loss=0.2828, pruned_loss=0.05824, over 17139.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.279, pruned_loss=0.06222, over 3257457.77 frames. ], batch size: 48, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:01:37,344 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:01:57,063 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:01:59,519 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 15:02:13,246 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:24,038 INFO [train.py:904] (2/8) Epoch 7, batch 850, loss[loss=0.1854, simple_loss=0.2614, pruned_loss=0.05465, over 16728.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.278, pruned_loss=0.06145, over 3277194.42 frames. ], batch size: 134, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:02:42,572 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:51,035 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3112, 2.3123, 1.8561, 2.1131, 2.7768, 2.5816, 2.7907, 2.8731], device='cuda:2'), covar=tensor([0.0089, 0.0208, 0.0268, 0.0262, 0.0110, 0.0186, 0.0135, 0.0119], device='cuda:2'), in_proj_covar=tensor([0.0099, 0.0173, 0.0173, 0.0172, 0.0168, 0.0175, 0.0163, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:03:06,806 INFO [optim.py:368] (2/8) 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 15:03:12,411 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5693, 3.1187, 2.9225, 1.8893, 2.6993, 2.0916, 3.0358, 3.1175], device='cuda:2'), covar=tensor([0.0306, 0.0604, 0.0534, 0.1610, 0.0735, 0.0988, 0.0615, 0.0866], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0130, 0.0153, 0.0138, 0.0131, 0.0124, 0.0137, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 15:03:19,374 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:33,686 INFO [train.py:904] (2/8) Epoch 7, batch 900, loss[loss=0.1892, simple_loss=0.2652, pruned_loss=0.05661, over 16801.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2773, pruned_loss=0.06151, over 3276708.86 frames. ], batch size: 102, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:03:36,336 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:04:11,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5280, 4.2563, 4.4477, 4.7081, 4.8157, 4.3434, 4.6628, 4.8079], device='cuda:2'), covar=tensor([0.0885, 0.0876, 0.1250, 0.0486, 0.0494, 0.0836, 0.1081, 0.0434], device='cuda:2'), in_proj_covar=tensor([0.0457, 0.0561, 0.0705, 0.0561, 0.0428, 0.0420, 0.0447, 0.0479], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:04:40,613 INFO [train.py:904] (2/8) Epoch 7, batch 950, loss[loss=0.1665, simple_loss=0.2473, pruned_loss=0.04285, over 16810.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2761, pruned_loss=0.06066, over 3292921.79 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:05:20,755 INFO [optim.py:368] (2/8) 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 15:05:21,240 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6678, 3.9797, 4.2379, 2.0124, 4.4682, 4.4429, 3.1788, 3.4070], device='cuda:2'), covar=tensor([0.0771, 0.0155, 0.0180, 0.1174, 0.0061, 0.0108, 0.0354, 0.0375], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0095, 0.0084, 0.0140, 0.0071, 0.0088, 0.0119, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 15:05:46,847 INFO [train.py:904] (2/8) Epoch 7, batch 1000, loss[loss=0.161, simple_loss=0.2552, pruned_loss=0.03342, over 17218.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2747, pruned_loss=0.06032, over 3298281.34 frames. ], batch size: 45, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:06:05,338 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:06:31,391 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 15:06:56,255 INFO [train.py:904] (2/8) Epoch 7, batch 1050, loss[loss=0.1968, simple_loss=0.2785, pruned_loss=0.05759, over 16541.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2749, pruned_loss=0.06028, over 3296833.89 frames. ], batch size: 68, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:07:28,181 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:07:37,640 INFO [optim.py:368] (2/8) 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 15:07:47,793 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0149, 4.2727, 3.3697, 2.5349, 3.1980, 2.6043, 4.7841, 3.9609], device='cuda:2'), covar=tensor([0.2294, 0.0794, 0.1260, 0.1829, 0.2464, 0.1544, 0.0309, 0.0959], device='cuda:2'), in_proj_covar=tensor([0.0286, 0.0258, 0.0273, 0.0255, 0.0280, 0.0212, 0.0251, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:07:59,303 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4605, 2.3505, 2.0213, 2.1692, 2.8126, 2.7865, 3.5297, 3.1378], device='cuda:2'), covar=tensor([0.0044, 0.0241, 0.0295, 0.0261, 0.0159, 0.0193, 0.0115, 0.0136], device='cuda:2'), in_proj_covar=tensor([0.0102, 0.0174, 0.0173, 0.0173, 0.0170, 0.0176, 0.0165, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:08:07,086 INFO [train.py:904] (2/8) Epoch 7, batch 1100, loss[loss=0.2106, simple_loss=0.287, pruned_loss=0.06715, over 17083.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2741, pruned_loss=0.05953, over 3309588.68 frames. ], batch size: 53, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:08:39,732 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7875, 2.2094, 2.3264, 4.5301, 2.0056, 2.9321, 2.2793, 2.4381], device='cuda:2'), covar=tensor([0.0619, 0.2714, 0.1607, 0.0291, 0.3384, 0.1591, 0.2463, 0.2809], device='cuda:2'), in_proj_covar=tensor([0.0337, 0.0347, 0.0291, 0.0321, 0.0387, 0.0377, 0.0316, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:09:15,539 INFO [train.py:904] (2/8) Epoch 7, batch 1150, loss[loss=0.2101, simple_loss=0.2815, pruned_loss=0.06931, over 16439.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2727, pruned_loss=0.05869, over 3314034.76 frames. ], batch size: 146, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:09:57,383 INFO [optim.py:368] (2/8) 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 15:10:04,072 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:18,762 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:22,593 INFO [train.py:904] (2/8) Epoch 7, batch 1200, loss[loss=0.1868, simple_loss=0.2549, pruned_loss=0.05934, over 16927.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2718, pruned_loss=0.05805, over 3318116.00 frames. ], batch size: 116, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:10:51,371 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:11:30,498 INFO [train.py:904] (2/8) Epoch 7, batch 1250, loss[loss=0.2242, simple_loss=0.2799, pruned_loss=0.08421, over 16785.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2726, pruned_loss=0.05871, over 3316860.86 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:11:31,246 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-04-28 15:12:13,503 INFO [optim.py:368] (2/8) 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 15:12:13,927 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:12:17,112 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8829, 5.0827, 5.1881, 5.0677, 4.9537, 5.6556, 5.2206, 4.9203], device='cuda:2'), covar=tensor([0.0860, 0.1769, 0.1652, 0.1844, 0.2840, 0.1036, 0.1122, 0.2325], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0453, 0.0453, 0.0384, 0.0517, 0.0484, 0.0360, 0.0514], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:12:39,181 INFO [train.py:904] (2/8) Epoch 7, batch 1300, loss[loss=0.2288, simple_loss=0.2876, pruned_loss=0.08504, over 16781.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2737, pruned_loss=0.05943, over 3324495.62 frames. ], batch size: 83, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:13:17,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5227, 3.6194, 3.9995, 2.7824, 3.7172, 3.9643, 3.8059, 2.3573], device='cuda:2'), covar=tensor([0.0288, 0.0126, 0.0034, 0.0226, 0.0039, 0.0055, 0.0036, 0.0281], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0063, 0.0062, 0.0116, 0.0064, 0.0077, 0.0067, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:13:27,287 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2010, 3.7156, 3.3315, 2.0605, 2.8681, 2.3009, 3.5505, 3.6228], device='cuda:2'), covar=tensor([0.0208, 0.0569, 0.0535, 0.1447, 0.0643, 0.0955, 0.0482, 0.0799], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0134, 0.0154, 0.0139, 0.0132, 0.0124, 0.0137, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 15:13:49,052 INFO [train.py:904] (2/8) Epoch 7, batch 1350, loss[loss=0.1621, simple_loss=0.2442, pruned_loss=0.03997, over 16999.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2733, pruned_loss=0.05885, over 3322298.03 frames. ], batch size: 41, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:14:15,364 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:19,007 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:31,591 INFO [optim.py:368] (2/8) 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 15:14:56,760 INFO [train.py:904] (2/8) Epoch 7, batch 1400, loss[loss=0.1965, simple_loss=0.2682, pruned_loss=0.06242, over 16704.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2732, pruned_loss=0.05881, over 3311566.55 frames. ], batch size: 134, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:15:30,603 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0159, 5.3797, 5.1129, 5.2096, 4.7330, 4.6859, 4.8807, 5.4637], device='cuda:2'), covar=tensor([0.0877, 0.0763, 0.0933, 0.0590, 0.0673, 0.0765, 0.0734, 0.0738], device='cuda:2'), in_proj_covar=tensor([0.0459, 0.0594, 0.0497, 0.0396, 0.0371, 0.0385, 0.0490, 0.0431], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:15:42,028 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:16:06,306 INFO [train.py:904] (2/8) Epoch 7, batch 1450, loss[loss=0.2176, simple_loss=0.2791, pruned_loss=0.07802, over 16795.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2722, pruned_loss=0.05851, over 3313603.27 frames. ], batch size: 102, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:16:42,543 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 15:16:47,645 INFO [optim.py:368] (2/8) 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 15:16:56,074 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:16:59,996 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:10,944 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:14,669 INFO [train.py:904] (2/8) Epoch 7, batch 1500, loss[loss=0.2168, simple_loss=0.2924, pruned_loss=0.07062, over 16765.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.272, pruned_loss=0.05886, over 3312372.63 frames. ], batch size: 124, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:17:15,127 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5640, 4.0750, 4.3585, 1.9445, 4.5955, 4.6464, 3.2768, 3.5781], device='cuda:2'), covar=tensor([0.0817, 0.0149, 0.0199, 0.1280, 0.0069, 0.0059, 0.0348, 0.0341], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0098, 0.0086, 0.0143, 0.0072, 0.0091, 0.0122, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 15:17:41,047 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 15:18:01,110 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:12,612 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3632, 2.9087, 2.4896, 2.2798, 2.1567, 2.1060, 2.7766, 2.9049], device='cuda:2'), covar=tensor([0.1839, 0.0800, 0.1124, 0.1494, 0.1759, 0.1566, 0.0435, 0.0816], device='cuda:2'), in_proj_covar=tensor([0.0284, 0.0257, 0.0271, 0.0255, 0.0282, 0.0209, 0.0249, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:18:15,913 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:23,649 INFO [train.py:904] (2/8) Epoch 7, batch 1550, loss[loss=0.1858, simple_loss=0.2741, pruned_loss=0.04871, over 17096.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2744, pruned_loss=0.06091, over 3313498.53 frames. ], batch size: 47, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:18:24,177 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:28,391 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9353, 4.2810, 4.5245, 3.4350, 3.9294, 4.4981, 4.0888, 3.1086], device='cuda:2'), covar=tensor([0.0266, 0.0031, 0.0019, 0.0171, 0.0048, 0.0030, 0.0032, 0.0217], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0063, 0.0062, 0.0115, 0.0065, 0.0076, 0.0067, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:18:59,836 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:19:04,152 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7331, 2.4017, 1.9033, 2.2317, 2.8674, 2.6814, 3.0824, 2.9858], device='cuda:2'), covar=tensor([0.0093, 0.0219, 0.0292, 0.0250, 0.0119, 0.0174, 0.0147, 0.0132], device='cuda:2'), in_proj_covar=tensor([0.0106, 0.0175, 0.0174, 0.0176, 0.0172, 0.0178, 0.0168, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:19:06,019 INFO [optim.py:368] (2/8) 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 15:19:16,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1582, 5.6237, 5.6414, 5.4577, 5.5787, 6.0724, 5.7899, 5.4597], device='cuda:2'), covar=tensor([0.0643, 0.1614, 0.1585, 0.1809, 0.2337, 0.0908, 0.1088, 0.2139], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0446, 0.0450, 0.0377, 0.0503, 0.0478, 0.0356, 0.0508], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:19:30,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9883, 4.3859, 4.5512, 2.2583, 4.8885, 4.8477, 3.4307, 3.8168], device='cuda:2'), covar=tensor([0.0740, 0.0116, 0.0162, 0.1109, 0.0040, 0.0068, 0.0297, 0.0315], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0097, 0.0085, 0.0141, 0.0072, 0.0090, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 15:19:32,092 INFO [train.py:904] (2/8) Epoch 7, batch 1600, loss[loss=0.1735, simple_loss=0.2603, pruned_loss=0.04337, over 17183.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2758, pruned_loss=0.06139, over 3308906.76 frames. ], batch size: 46, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:19:37,849 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3242, 5.8341, 5.5193, 5.5864, 5.1232, 4.9076, 5.3075, 5.8356], device='cuda:2'), covar=tensor([0.0932, 0.0757, 0.0953, 0.0528, 0.0740, 0.0646, 0.0769, 0.0819], device='cuda:2'), in_proj_covar=tensor([0.0455, 0.0591, 0.0494, 0.0391, 0.0368, 0.0381, 0.0482, 0.0429], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:20:25,553 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 15:20:29,593 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4064, 4.1437, 4.4082, 4.5984, 4.7128, 4.2379, 4.5653, 4.7081], device='cuda:2'), covar=tensor([0.1192, 0.0968, 0.1262, 0.0714, 0.0607, 0.0875, 0.1012, 0.0732], device='cuda:2'), in_proj_covar=tensor([0.0469, 0.0570, 0.0720, 0.0578, 0.0438, 0.0432, 0.0460, 0.0494], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:20:31,597 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:20:40,967 INFO [train.py:904] (2/8) Epoch 7, batch 1650, loss[loss=0.2106, simple_loss=0.3012, pruned_loss=0.06003, over 17078.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2778, pruned_loss=0.06225, over 3316826.36 frames. ], batch size: 55, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:20:56,060 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5231, 4.4778, 4.3904, 4.2587, 4.1038, 4.3646, 4.2871, 4.1307], device='cuda:2'), covar=tensor([0.0462, 0.0323, 0.0197, 0.0203, 0.0677, 0.0374, 0.0405, 0.0496], device='cuda:2'), in_proj_covar=tensor([0.0218, 0.0252, 0.0257, 0.0232, 0.0295, 0.0261, 0.0183, 0.0298], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:21:04,841 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:21:22,701 INFO [optim.py:368] (2/8) 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 15:21:25,408 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 1700, loss[loss=0.2273, simple_loss=0.3094, pruned_loss=0.07258, over 17058.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2803, pruned_loss=0.06282, over 3322123.51 frames. ], batch size: 53, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:21:55,129 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:12,978 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:27,849 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:51,308 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:58,544 INFO [train.py:904] (2/8) Epoch 7, batch 1750, loss[loss=0.2062, simple_loss=0.293, pruned_loss=0.05967, over 17217.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2802, pruned_loss=0.06228, over 3328647.66 frames. ], batch size: 46, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:23:18,099 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:23:34,264 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4264, 2.2344, 1.6925, 1.9942, 2.7382, 2.5837, 2.8569, 2.8680], device='cuda:2'), covar=tensor([0.0085, 0.0232, 0.0292, 0.0276, 0.0116, 0.0155, 0.0137, 0.0127], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0177, 0.0175, 0.0178, 0.0173, 0.0180, 0.0172, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:23:41,547 INFO [optim.py:368] (2/8) 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 15:24:08,264 INFO [train.py:904] (2/8) Epoch 7, batch 1800, loss[loss=0.1719, simple_loss=0.2532, pruned_loss=0.04527, over 16798.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.282, pruned_loss=0.06253, over 3318500.64 frames. ], batch size: 39, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:24:27,564 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7263, 3.7504, 2.8717, 2.3533, 2.8292, 2.3366, 3.7478, 3.7632], device='cuda:2'), covar=tensor([0.1977, 0.0621, 0.1198, 0.1687, 0.1819, 0.1422, 0.0474, 0.0764], device='cuda:2'), in_proj_covar=tensor([0.0281, 0.0253, 0.0270, 0.0254, 0.0279, 0.0207, 0.0248, 0.0274], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:24:42,426 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:10,912 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:17,218 INFO [train.py:904] (2/8) Epoch 7, batch 1850, loss[loss=0.2009, simple_loss=0.2899, pruned_loss=0.05597, over 17253.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2826, pruned_loss=0.06206, over 3324804.33 frames. ], batch size: 52, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:25:45,822 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:52,511 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:59,964 INFO [optim.py:368] (2/8) 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 15:26:09,634 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 15:26:26,217 INFO [train.py:904] (2/8) Epoch 7, batch 1900, loss[loss=0.1931, simple_loss=0.2823, pruned_loss=0.052, over 16569.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2812, pruned_loss=0.06054, over 3334110.61 frames. ], batch size: 68, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:26:55,340 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4843, 3.9436, 4.1669, 2.0479, 4.3001, 4.3291, 3.1336, 3.0914], device='cuda:2'), covar=tensor([0.0808, 0.0122, 0.0127, 0.1096, 0.0050, 0.0086, 0.0348, 0.0401], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0096, 0.0084, 0.0141, 0.0072, 0.0090, 0.0120, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 15:27:00,818 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:11,573 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:36,784 INFO [train.py:904] (2/8) Epoch 7, batch 1950, loss[loss=0.2101, simple_loss=0.2853, pruned_loss=0.06747, over 16481.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2806, pruned_loss=0.06016, over 3329152.92 frames. ], batch size: 146, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:27:39,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4016, 4.7803, 4.4890, 4.4925, 4.1759, 4.1300, 4.2616, 4.7775], device='cuda:2'), covar=tensor([0.0948, 0.0732, 0.0976, 0.0630, 0.0735, 0.1141, 0.0870, 0.0808], device='cuda:2'), in_proj_covar=tensor([0.0452, 0.0585, 0.0493, 0.0390, 0.0366, 0.0375, 0.0481, 0.0424], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:28:19,259 INFO [optim.py:368] (2/8) 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 15:28:44,374 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:28:45,196 INFO [train.py:904] (2/8) Epoch 7, batch 2000, loss[loss=0.2195, simple_loss=0.2867, pruned_loss=0.07613, over 16768.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2813, pruned_loss=0.0605, over 3317250.94 frames. ], batch size: 124, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:28:53,757 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 15:29:24,482 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:39,596 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:55,690 INFO [train.py:904] (2/8) Epoch 7, batch 2050, loss[loss=0.2113, simple_loss=0.2952, pruned_loss=0.06368, over 16792.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2814, pruned_loss=0.06103, over 3315318.03 frames. ], batch size: 83, lr: 9.99e-03, grad_scale: 16.0 2023-04-28 15:30:31,319 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:30:39,208 INFO [optim.py:368] (2/8) 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 15:31:05,139 INFO [train.py:904] (2/8) Epoch 7, batch 2100, loss[loss=0.1985, simple_loss=0.2793, pruned_loss=0.05887, over 17268.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2828, pruned_loss=0.06231, over 3297775.71 frames. ], batch size: 52, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:31:33,156 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:31:47,523 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2805, 4.1207, 3.8954, 1.9779, 3.0241, 2.4432, 3.7872, 3.9321], device='cuda:2'), covar=tensor([0.0280, 0.0569, 0.0479, 0.1715, 0.0727, 0.0944, 0.0590, 0.0804], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0139, 0.0156, 0.0141, 0.0134, 0.0124, 0.0139, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 15:32:09,450 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:16,867 INFO [train.py:904] (2/8) Epoch 7, batch 2150, loss[loss=0.2006, simple_loss=0.2903, pruned_loss=0.05546, over 16682.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2833, pruned_loss=0.06259, over 3309672.36 frames. ], batch size: 62, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:33:00,561 INFO [optim.py:368] (2/8) 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 15:33:02,204 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9142, 4.6519, 4.7530, 5.0864, 5.2209, 4.6297, 5.1828, 5.2454], device='cuda:2'), covar=tensor([0.1111, 0.1061, 0.1741, 0.0644, 0.0658, 0.0702, 0.0648, 0.0568], device='cuda:2'), in_proj_covar=tensor([0.0464, 0.0567, 0.0718, 0.0571, 0.0437, 0.0431, 0.0450, 0.0489], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:33:17,098 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:33:25,757 INFO [train.py:904] (2/8) Epoch 7, batch 2200, loss[loss=0.2132, simple_loss=0.2972, pruned_loss=0.06464, over 16493.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2844, pruned_loss=0.0631, over 3306855.81 frames. ], batch size: 68, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:33:28,124 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0868, 3.6974, 3.2267, 2.0345, 2.7014, 2.3580, 3.3366, 3.5718], device='cuda:2'), covar=tensor([0.0303, 0.0568, 0.0514, 0.1402, 0.0713, 0.0823, 0.0671, 0.0754], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0139, 0.0156, 0.0139, 0.0133, 0.0124, 0.0139, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 15:33:31,600 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7131, 2.0920, 2.2107, 4.3175, 2.0949, 2.8346, 2.2291, 2.3303], device='cuda:2'), covar=tensor([0.0687, 0.2586, 0.1422, 0.0327, 0.3001, 0.1428, 0.2383, 0.2306], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0351, 0.0291, 0.0319, 0.0386, 0.0383, 0.0318, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:34:03,308 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:34:33,869 INFO [train.py:904] (2/8) Epoch 7, batch 2250, loss[loss=0.1748, simple_loss=0.2637, pruned_loss=0.04296, over 17190.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.284, pruned_loss=0.0628, over 3310718.91 frames. ], batch size: 44, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:35:18,521 INFO [optim.py:368] (2/8) 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 15:35:42,282 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:35:43,089 INFO [train.py:904] (2/8) Epoch 7, batch 2300, loss[loss=0.2283, simple_loss=0.2986, pruned_loss=0.07904, over 16843.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2847, pruned_loss=0.06304, over 3297663.08 frames. ], batch size: 96, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:36:10,189 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:36,800 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:48,217 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 2350, loss[loss=0.2347, simple_loss=0.2958, pruned_loss=0.08681, over 16898.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2853, pruned_loss=0.06395, over 3298841.78 frames. ], batch size: 109, lr: 9.96e-03, grad_scale: 4.0 2023-04-28 15:37:01,783 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2707, 5.1808, 5.0711, 4.4724, 5.1127, 1.8586, 4.7908, 5.0998], device='cuda:2'), covar=tensor([0.0051, 0.0051, 0.0104, 0.0294, 0.0057, 0.1953, 0.0098, 0.0128], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0096, 0.0147, 0.0141, 0.0113, 0.0158, 0.0130, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:37:33,977 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:37:38,003 INFO [optim.py:368] (2/8) 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 15:37:43,611 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:38:00,178 INFO [train.py:904] (2/8) Epoch 7, batch 2400, loss[loss=0.2438, simple_loss=0.3254, pruned_loss=0.08112, over 17065.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2856, pruned_loss=0.06369, over 3307901.07 frames. ], batch size: 53, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:38:27,636 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:10,358 INFO [train.py:904] (2/8) Epoch 7, batch 2450, loss[loss=0.2184, simple_loss=0.2979, pruned_loss=0.06946, over 16707.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2856, pruned_loss=0.06303, over 3313205.16 frames. ], batch size: 134, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:39:27,279 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7266, 4.7067, 5.2679, 5.2319, 5.2621, 4.8200, 4.7869, 4.5532], device='cuda:2'), covar=tensor([0.0281, 0.0458, 0.0344, 0.0408, 0.0414, 0.0292, 0.0847, 0.0404], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0287, 0.0290, 0.0274, 0.0333, 0.0306, 0.0409, 0.0247], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 15:39:34,202 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:36,259 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9673, 3.9886, 3.8139, 3.6606, 3.5588, 3.9475, 3.6415, 3.6498], device='cuda:2'), covar=tensor([0.0517, 0.0397, 0.0233, 0.0240, 0.0698, 0.0328, 0.0858, 0.0519], device='cuda:2'), in_proj_covar=tensor([0.0221, 0.0256, 0.0262, 0.0236, 0.0295, 0.0262, 0.0185, 0.0300], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:39:42,810 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1076, 2.2049, 2.4383, 4.7417, 2.1166, 2.9548, 2.4679, 2.4290], device='cuda:2'), covar=tensor([0.0554, 0.3021, 0.1574, 0.0258, 0.3381, 0.1756, 0.2315, 0.3131], device='cuda:2'), in_proj_covar=tensor([0.0339, 0.0351, 0.0293, 0.0320, 0.0386, 0.0386, 0.0318, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:39:55,096 INFO [optim.py:368] (2/8) 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 15:40:17,255 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0239, 5.4082, 5.5368, 5.3228, 5.2938, 5.9076, 5.4165, 5.1765], device='cuda:2'), covar=tensor([0.0797, 0.1534, 0.1481, 0.1707, 0.2559, 0.0953, 0.1112, 0.2149], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0461, 0.0457, 0.0386, 0.0524, 0.0493, 0.0372, 0.0521], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:40:19,245 INFO [train.py:904] (2/8) Epoch 7, batch 2500, loss[loss=0.192, simple_loss=0.2774, pruned_loss=0.05334, over 17053.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2856, pruned_loss=0.06323, over 3313155.86 frames. ], batch size: 50, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:40:33,793 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 15:40:44,222 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:46,527 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:55,009 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:41:26,924 INFO [train.py:904] (2/8) Epoch 7, batch 2550, loss[loss=0.2024, simple_loss=0.2694, pruned_loss=0.06775, over 16837.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2862, pruned_loss=0.06362, over 3309457.68 frames. ], batch size: 109, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:41:42,854 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 15:41:43,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8641, 4.8947, 5.4560, 5.4274, 5.4003, 5.0138, 4.9872, 4.7086], device='cuda:2'), covar=tensor([0.0245, 0.0373, 0.0248, 0.0300, 0.0361, 0.0261, 0.0738, 0.0358], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0289, 0.0291, 0.0275, 0.0334, 0.0307, 0.0413, 0.0249], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 15:42:01,822 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:08,270 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:11,087 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 15:42:18,506 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 15:42:35,898 INFO [train.py:904] (2/8) Epoch 7, batch 2600, loss[loss=0.2025, simple_loss=0.2873, pruned_loss=0.05885, over 17244.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2861, pruned_loss=0.06312, over 3316098.15 frames. ], batch size: 45, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:43:42,047 INFO [train.py:904] (2/8) Epoch 7, batch 2650, loss[loss=0.1944, simple_loss=0.2877, pruned_loss=0.0506, over 17079.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2863, pruned_loss=0.0624, over 3323935.09 frames. ], batch size: 55, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:44:06,153 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:15,589 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:27,534 INFO [optim.py:368] (2/8) 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 15:44:27,956 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:31,776 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4647, 2.3246, 1.9836, 2.1357, 2.8573, 2.5685, 3.3514, 3.1230], device='cuda:2'), covar=tensor([0.0046, 0.0251, 0.0285, 0.0283, 0.0149, 0.0219, 0.0135, 0.0119], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0175, 0.0172, 0.0175, 0.0172, 0.0177, 0.0173, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:44:48,752 INFO [train.py:904] (2/8) Epoch 7, batch 2700, loss[loss=0.2223, simple_loss=0.2899, pruned_loss=0.07738, over 16913.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2867, pruned_loss=0.06191, over 3326813.60 frames. ], batch size: 109, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:45:06,202 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9609, 3.7608, 3.9787, 4.1687, 4.2297, 3.8162, 3.9532, 4.2379], device='cuda:2'), covar=tensor([0.1063, 0.0941, 0.1186, 0.0534, 0.0565, 0.1395, 0.1290, 0.0537], device='cuda:2'), in_proj_covar=tensor([0.0476, 0.0583, 0.0740, 0.0581, 0.0448, 0.0447, 0.0458, 0.0509], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:45:14,489 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1624, 3.1925, 3.5068, 2.3251, 3.2346, 3.5056, 3.3187, 1.7872], device='cuda:2'), covar=tensor([0.0315, 0.0081, 0.0039, 0.0256, 0.0067, 0.0064, 0.0058, 0.0370], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0063, 0.0061, 0.0116, 0.0065, 0.0076, 0.0067, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:45:28,046 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6162, 4.0204, 4.1652, 2.9745, 3.6602, 4.1059, 3.9169, 2.3890], device='cuda:2'), covar=tensor([0.0313, 0.0039, 0.0031, 0.0242, 0.0059, 0.0055, 0.0041, 0.0321], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0063, 0.0061, 0.0116, 0.0065, 0.0076, 0.0067, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:45:28,088 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:44,444 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 15:45:51,362 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 2750, loss[loss=0.1909, simple_loss=0.2722, pruned_loss=0.0548, over 17185.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2857, pruned_loss=0.06124, over 3329489.41 frames. ], batch size: 46, lr: 9.93e-03, grad_scale: 4.0 2023-04-28 15:46:39,345 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9355, 1.6796, 2.3931, 2.8651, 2.6845, 3.3393, 1.8175, 3.2768], device='cuda:2'), covar=tensor([0.0118, 0.0274, 0.0166, 0.0146, 0.0140, 0.0091, 0.0267, 0.0071], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0158, 0.0140, 0.0143, 0.0147, 0.0104, 0.0149, 0.0094], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 15:46:45,844 INFO [optim.py:368] (2/8) 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 15:46:51,987 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-28 15:47:08,528 INFO [train.py:904] (2/8) Epoch 7, batch 2800, loss[loss=0.2124, simple_loss=0.2872, pruned_loss=0.06876, over 15384.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2863, pruned_loss=0.06176, over 3323339.35 frames. ], batch size: 190, lr: 9.93e-03, grad_scale: 8.0 2023-04-28 15:47:32,018 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-28 15:48:05,144 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8462, 3.2747, 2.7442, 4.3633, 3.8697, 4.2703, 1.6804, 3.1053], device='cuda:2'), covar=tensor([0.1134, 0.0397, 0.0826, 0.0119, 0.0185, 0.0285, 0.1143, 0.0587], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0145, 0.0166, 0.0105, 0.0197, 0.0198, 0.0164, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 15:48:12,843 INFO [train.py:904] (2/8) Epoch 7, batch 2850, loss[loss=0.1862, simple_loss=0.2619, pruned_loss=0.05525, over 16788.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2857, pruned_loss=0.06164, over 3327933.86 frames. ], batch size: 102, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:48:17,213 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-28 15:48:47,331 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:48:50,119 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:49:02,034 INFO [optim.py:368] (2/8) 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 15:49:23,198 INFO [train.py:904] (2/8) Epoch 7, batch 2900, loss[loss=0.1973, simple_loss=0.2811, pruned_loss=0.0568, over 17118.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2853, pruned_loss=0.06308, over 3319145.07 frames. ], batch size: 48, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:49:31,591 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0137, 4.3975, 2.3851, 4.6950, 3.0099, 4.6488, 2.2968, 3.2553], device='cuda:2'), covar=tensor([0.0150, 0.0199, 0.1315, 0.0107, 0.0652, 0.0339, 0.1442, 0.0555], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0162, 0.0177, 0.0101, 0.0161, 0.0205, 0.0189, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 15:49:35,868 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:49:53,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7593, 4.9230, 5.0785, 4.9734, 4.9771, 5.5568, 5.1601, 4.9458], device='cuda:2'), covar=tensor([0.1249, 0.1880, 0.1672, 0.1964, 0.2520, 0.1096, 0.1286, 0.2256], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0458, 0.0456, 0.0385, 0.0523, 0.0488, 0.0369, 0.0521], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 15:50:22,267 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:50:32,271 INFO [train.py:904] (2/8) Epoch 7, batch 2950, loss[loss=0.206, simple_loss=0.2711, pruned_loss=0.07045, over 16374.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2844, pruned_loss=0.06363, over 3318054.62 frames. ], batch size: 146, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:50:58,344 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:05,372 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:17,102 INFO [optim.py:368] (2/8) 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 15:51:38,107 INFO [train.py:904] (2/8) Epoch 7, batch 3000, loss[loss=0.2358, simple_loss=0.2993, pruned_loss=0.08617, over 16916.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2849, pruned_loss=0.06415, over 3315791.10 frames. ], batch size: 109, lr: 9.91e-03, grad_scale: 8.0 2023-04-28 15:51:38,107 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 15:51:46,836 INFO [train.py:938] (2/8) Epoch 7, validation: loss=0.1489, simple_loss=0.2553, pruned_loss=0.02124, over 944034.00 frames. 2023-04-28 15:51:46,837 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 15:51:54,307 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:06,756 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:16,452 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1151, 4.0621, 4.4962, 4.4858, 4.5092, 4.1501, 4.2177, 4.0610], device='cuda:2'), covar=tensor([0.0278, 0.0514, 0.0282, 0.0358, 0.0402, 0.0314, 0.0721, 0.0477], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0296, 0.0294, 0.0282, 0.0345, 0.0313, 0.0425, 0.0253], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 15:52:19,937 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:19,951 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:42,089 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:56,533 INFO [train.py:904] (2/8) Epoch 7, batch 3050, loss[loss=0.1881, simple_loss=0.2793, pruned_loss=0.04847, over 17064.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2841, pruned_loss=0.0634, over 3324191.50 frames. ], batch size: 53, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:53:30,203 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:53:43,706 INFO [optim.py:368] (2/8) 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 15:54:06,475 INFO [train.py:904] (2/8) Epoch 7, batch 3100, loss[loss=0.2053, simple_loss=0.2866, pruned_loss=0.06196, over 16577.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2837, pruned_loss=0.06367, over 3321785.71 frames. ], batch size: 62, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:54:29,442 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1493, 1.9347, 1.5588, 1.7475, 2.2517, 1.9870, 2.2350, 2.3716], device='cuda:2'), covar=tensor([0.0088, 0.0188, 0.0235, 0.0236, 0.0101, 0.0195, 0.0134, 0.0109], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0175, 0.0171, 0.0173, 0.0172, 0.0176, 0.0173, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:54:36,561 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:16,375 INFO [train.py:904] (2/8) Epoch 7, batch 3150, loss[loss=0.2281, simple_loss=0.2944, pruned_loss=0.08089, over 16707.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2833, pruned_loss=0.06402, over 3315205.50 frames. ], batch size: 134, lr: 9.90e-03, grad_scale: 4.0 2023-04-28 15:55:16,875 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5499, 2.3420, 2.0703, 2.3700, 2.8477, 2.6584, 3.5155, 3.0753], device='cuda:2'), covar=tensor([0.0045, 0.0268, 0.0273, 0.0254, 0.0150, 0.0228, 0.0120, 0.0151], device='cuda:2'), in_proj_covar=tensor([0.0108, 0.0175, 0.0171, 0.0173, 0.0171, 0.0176, 0.0173, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:55:26,751 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-28 15:55:45,663 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9309, 4.8932, 4.7397, 4.1996, 4.7730, 2.0880, 4.6024, 4.7635], device='cuda:2'), covar=tensor([0.0070, 0.0058, 0.0123, 0.0303, 0.0071, 0.1827, 0.0094, 0.0125], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0100, 0.0152, 0.0148, 0.0117, 0.0160, 0.0134, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 15:55:49,774 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:52,128 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:00,109 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 15:56:19,954 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2570, 4.3645, 4.6219, 2.3146, 4.9253, 4.8957, 3.3271, 3.8737], device='cuda:2'), covar=tensor([0.0609, 0.0132, 0.0140, 0.1003, 0.0047, 0.0071, 0.0310, 0.0287], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0099, 0.0086, 0.0140, 0.0073, 0.0093, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 15:56:24,282 INFO [train.py:904] (2/8) Epoch 7, batch 3200, loss[loss=0.17, simple_loss=0.2473, pruned_loss=0.04637, over 16838.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2809, pruned_loss=0.06216, over 3326084.39 frames. ], batch size: 42, lr: 9.90e-03, grad_scale: 8.0 2023-04-28 15:56:54,782 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:57,814 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:57:33,480 INFO [train.py:904] (2/8) Epoch 7, batch 3250, loss[loss=0.2242, simple_loss=0.2915, pruned_loss=0.07839, over 16494.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2801, pruned_loss=0.06148, over 3324824.23 frames. ], batch size: 75, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:57:50,158 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 15:57:53,162 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:21,283 INFO [optim.py:368] (2/8) 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 15:58:42,126 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:43,055 INFO [train.py:904] (2/8) Epoch 7, batch 3300, loss[loss=0.1742, simple_loss=0.2535, pruned_loss=0.04745, over 17060.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2813, pruned_loss=0.06178, over 3329356.42 frames. ], batch size: 41, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:59:16,292 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:25,373 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:26,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3174, 3.7039, 3.6946, 1.9062, 3.9226, 3.9581, 3.0578, 3.0560], device='cuda:2'), covar=tensor([0.0741, 0.0116, 0.0120, 0.0987, 0.0072, 0.0083, 0.0335, 0.0314], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0097, 0.0085, 0.0138, 0.0072, 0.0091, 0.0120, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 15:59:40,452 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:51,971 INFO [train.py:904] (2/8) Epoch 7, batch 3350, loss[loss=0.1742, simple_loss=0.2586, pruned_loss=0.0449, over 17168.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2828, pruned_loss=0.06265, over 3319656.45 frames. ], batch size: 46, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 16:00:14,924 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1742, 1.9331, 2.1216, 3.5765, 1.9503, 2.4555, 2.1010, 2.0738], device='cuda:2'), covar=tensor([0.0718, 0.2594, 0.1616, 0.0399, 0.3009, 0.1677, 0.2469, 0.2666], device='cuda:2'), in_proj_covar=tensor([0.0347, 0.0359, 0.0300, 0.0332, 0.0392, 0.0394, 0.0324, 0.0426], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:00:20,484 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:42,046 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:49,528 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:01:02,164 INFO [train.py:904] (2/8) Epoch 7, batch 3400, loss[loss=0.1686, simple_loss=0.2444, pruned_loss=0.04636, over 17009.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2828, pruned_loss=0.06237, over 3325633.17 frames. ], batch size: 41, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:01:09,362 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1542, 4.4011, 4.5515, 2.1298, 4.9210, 4.8607, 3.4054, 3.9226], device='cuda:2'), covar=tensor([0.0614, 0.0121, 0.0144, 0.1024, 0.0044, 0.0057, 0.0315, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0097, 0.0085, 0.0138, 0.0073, 0.0091, 0.0119, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 16:02:03,017 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:02:09,232 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 16:02:10,853 INFO [train.py:904] (2/8) Epoch 7, batch 3450, loss[loss=0.1925, simple_loss=0.2855, pruned_loss=0.04973, over 17027.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2818, pruned_loss=0.06195, over 3325630.72 frames. ], batch size: 55, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:02:49,994 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:03:01,780 INFO [optim.py:368] (2/8) 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 16:03:21,642 INFO [train.py:904] (2/8) Epoch 7, batch 3500, loss[loss=0.2597, simple_loss=0.3214, pruned_loss=0.099, over 11788.00 frames. ], tot_loss[loss=0.201, simple_loss=0.28, pruned_loss=0.06101, over 3324927.40 frames. ], batch size: 248, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:03:28,794 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:03:59,586 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9758, 3.9356, 4.3694, 4.3143, 4.3717, 4.0032, 4.0989, 3.9648], device='cuda:2'), covar=tensor([0.0278, 0.0501, 0.0311, 0.0415, 0.0337, 0.0323, 0.0648, 0.0483], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0296, 0.0297, 0.0287, 0.0344, 0.0314, 0.0424, 0.0257], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 16:04:30,074 INFO [train.py:904] (2/8) Epoch 7, batch 3550, loss[loss=0.1787, simple_loss=0.263, pruned_loss=0.04718, over 17200.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2786, pruned_loss=0.06043, over 3318368.41 frames. ], batch size: 46, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:04:48,132 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:49,886 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:18,414 INFO [optim.py:368] (2/8) 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 16:05:38,460 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:39,299 INFO [train.py:904] (2/8) Epoch 7, batch 3600, loss[loss=0.1826, simple_loss=0.2548, pruned_loss=0.05521, over 16833.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.277, pruned_loss=0.05983, over 3325836.39 frames. ], batch size: 102, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:05:56,510 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:14,272 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:15,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1157, 3.5852, 3.2305, 2.0222, 2.7986, 2.2423, 3.5714, 3.6302], device='cuda:2'), covar=tensor([0.0206, 0.0563, 0.0516, 0.1472, 0.0689, 0.0898, 0.0456, 0.0651], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0140, 0.0155, 0.0139, 0.0132, 0.0123, 0.0138, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 16:06:45,518 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:45,920 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 16:06:48,841 INFO [train.py:904] (2/8) Epoch 7, batch 3650, loss[loss=0.1835, simple_loss=0.2432, pruned_loss=0.06189, over 16824.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2768, pruned_loss=0.06028, over 3307413.88 frames. ], batch size: 83, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:07:19,287 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:40,347 INFO [optim.py:368] (2/8) 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 16:07:42,462 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:42,634 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2929, 3.4282, 1.9855, 3.5650, 2.4895, 3.6140, 1.9853, 2.7926], device='cuda:2'), covar=tensor([0.0213, 0.0373, 0.1378, 0.0154, 0.0765, 0.0523, 0.1377, 0.0603], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0164, 0.0179, 0.0103, 0.0163, 0.0207, 0.0192, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 16:07:44,476 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:08:01,998 INFO [train.py:904] (2/8) Epoch 7, batch 3700, loss[loss=0.1933, simple_loss=0.2543, pruned_loss=0.0661, over 16727.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.275, pruned_loss=0.06163, over 3287594.55 frames. ], batch size: 83, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:08:27,963 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:08:36,352 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 16:09:09,919 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:12,301 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:14,069 INFO [train.py:904] (2/8) Epoch 7, batch 3750, loss[loss=0.2448, simple_loss=0.3098, pruned_loss=0.08984, over 16857.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2758, pruned_loss=0.06353, over 3265331.99 frames. ], batch size: 116, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:09:28,413 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0325, 3.0829, 3.1410, 1.6289, 3.3611, 3.3077, 2.6728, 2.5597], device='cuda:2'), covar=tensor([0.0809, 0.0174, 0.0154, 0.1176, 0.0077, 0.0110, 0.0401, 0.0442], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0097, 0.0085, 0.0139, 0.0073, 0.0092, 0.0120, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 16:09:55,030 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:04,721 INFO [optim.py:368] (2/8) 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 16:10:24,877 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:25,829 INFO [train.py:904] (2/8) Epoch 7, batch 3800, loss[loss=0.1897, simple_loss=0.2623, pruned_loss=0.05858, over 16740.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2769, pruned_loss=0.06513, over 3278087.50 frames. ], batch size: 124, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:10:27,925 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:38,632 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:05,637 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:31,443 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:39,992 INFO [train.py:904] (2/8) Epoch 7, batch 3850, loss[loss=0.1908, simple_loss=0.2698, pruned_loss=0.05584, over 16497.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2772, pruned_loss=0.06558, over 3274761.81 frames. ], batch size: 62, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:11:40,481 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:51,067 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-28 16:11:58,643 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:12:26,436 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 16:12:33,098 INFO [optim.py:368] (2/8) 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 16:12:34,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4586, 3.8979, 4.1846, 2.8074, 3.7441, 4.0833, 4.0220, 2.5361], device='cuda:2'), covar=tensor([0.0325, 0.0102, 0.0024, 0.0221, 0.0036, 0.0046, 0.0030, 0.0272], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0060, 0.0060, 0.0114, 0.0064, 0.0074, 0.0066, 0.0110], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:12:52,613 INFO [train.py:904] (2/8) Epoch 7, batch 3900, loss[loss=0.1938, simple_loss=0.2657, pruned_loss=0.06094, over 16295.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2769, pruned_loss=0.06607, over 3273523.40 frames. ], batch size: 165, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:13:00,361 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4052, 3.5818, 1.8136, 3.7130, 2.6124, 3.7216, 1.9747, 2.7182], device='cuda:2'), covar=tensor([0.0143, 0.0286, 0.1473, 0.0104, 0.0658, 0.0417, 0.1381, 0.0618], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0164, 0.0180, 0.0101, 0.0162, 0.0205, 0.0193, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 16:13:00,378 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:08,259 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:20,516 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:29,561 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5906, 3.6562, 2.7753, 2.1999, 2.4621, 2.1247, 3.5393, 3.3660], device='cuda:2'), covar=tensor([0.2082, 0.0622, 0.1291, 0.2012, 0.2168, 0.1723, 0.0482, 0.1000], device='cuda:2'), in_proj_covar=tensor([0.0287, 0.0259, 0.0276, 0.0263, 0.0295, 0.0215, 0.0256, 0.0283], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:14:02,898 INFO [train.py:904] (2/8) Epoch 7, batch 3950, loss[loss=0.2094, simple_loss=0.2743, pruned_loss=0.07222, over 16746.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2763, pruned_loss=0.06676, over 3273072.09 frames. ], batch size: 124, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:14:53,035 INFO [optim.py:368] (2/8) 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 16:14:53,496 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:15:12,235 INFO [train.py:904] (2/8) Epoch 7, batch 4000, loss[loss=0.1975, simple_loss=0.2746, pruned_loss=0.0602, over 17172.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2769, pruned_loss=0.06724, over 3274971.09 frames. ], batch size: 46, lr: 9.84e-03, grad_scale: 8.0 2023-04-28 16:16:02,047 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:16,961 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:24,726 INFO [train.py:904] (2/8) Epoch 7, batch 4050, loss[loss=0.1929, simple_loss=0.2731, pruned_loss=0.05639, over 16822.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2767, pruned_loss=0.06541, over 3276537.91 frames. ], batch size: 83, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:16:58,883 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 16:16:59,765 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:17:16,149 INFO [optim.py:368] (2/8) 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 16:17:35,727 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 4100, loss[loss=0.2129, simple_loss=0.2934, pruned_loss=0.06621, over 16836.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2773, pruned_loss=0.06392, over 3268305.56 frames. ], batch size: 116, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:17:39,308 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:29,716 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:18:47,166 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:51,079 INFO [train.py:904] (2/8) Epoch 7, batch 4150, loss[loss=0.224, simple_loss=0.3054, pruned_loss=0.07129, over 17057.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2858, pruned_loss=0.06725, over 3229847.23 frames. ], batch size: 41, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:18:56,794 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:01,895 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:18,695 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8545, 4.6999, 5.2493, 5.2088, 5.2642, 4.7264, 4.7957, 4.4293], device='cuda:2'), covar=tensor([0.0214, 0.0402, 0.0286, 0.0319, 0.0329, 0.0292, 0.0772, 0.0370], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0280, 0.0280, 0.0271, 0.0327, 0.0296, 0.0397, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 16:19:22,451 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9503, 2.4509, 2.3213, 3.0217, 2.4359, 3.2461, 1.7050, 2.7051], device='cuda:2'), covar=tensor([0.1092, 0.0501, 0.0970, 0.0103, 0.0195, 0.0372, 0.1280, 0.0653], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0147, 0.0169, 0.0107, 0.0202, 0.0196, 0.0167, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 16:19:44,054 INFO [optim.py:368] (2/8) 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 16:20:03,713 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:04,510 INFO [train.py:904] (2/8) Epoch 7, batch 4200, loss[loss=0.2342, simple_loss=0.3203, pruned_loss=0.07409, over 16876.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2928, pruned_loss=0.06912, over 3210647.88 frames. ], batch size: 109, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:20:13,268 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:24,647 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:32,655 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:21:16,726 INFO [train.py:904] (2/8) Epoch 7, batch 4250, loss[loss=0.2342, simple_loss=0.3193, pruned_loss=0.0745, over 16473.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2961, pruned_loss=0.0692, over 3202013.26 frames. ], batch size: 35, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:21:42,348 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:22:08,030 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 7, batch 4300, loss[loss=0.2441, simple_loss=0.3279, pruned_loss=0.08018, over 16800.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2972, pruned_loss=0.06809, over 3198645.35 frames. ], batch size: 124, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:23:09,285 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2974, 2.3171, 1.8102, 2.1071, 2.7517, 2.4128, 3.2143, 3.0482], device='cuda:2'), covar=tensor([0.0037, 0.0244, 0.0322, 0.0265, 0.0150, 0.0219, 0.0092, 0.0110], device='cuda:2'), in_proj_covar=tensor([0.0105, 0.0172, 0.0172, 0.0172, 0.0167, 0.0174, 0.0166, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:23:31,789 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:23:41,836 INFO [train.py:904] (2/8) Epoch 7, batch 4350, loss[loss=0.2394, simple_loss=0.3178, pruned_loss=0.08046, over 16832.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3004, pruned_loss=0.06904, over 3206587.19 frames. ], batch size: 116, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:30,808 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:33,878 INFO [optim.py:368] (2/8) 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 16:24:42,765 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:54,958 INFO [train.py:904] (2/8) Epoch 7, batch 4400, loss[loss=0.283, simple_loss=0.3438, pruned_loss=0.1111, over 11830.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3023, pruned_loss=0.07, over 3192015.76 frames. ], batch size: 246, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:57,768 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:25:39,691 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:25:49,570 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1258, 2.4047, 2.5676, 4.9743, 2.1490, 3.1435, 2.5913, 2.5895], device='cuda:2'), covar=tensor([0.0611, 0.2501, 0.1439, 0.0209, 0.3200, 0.1475, 0.2015, 0.2517], device='cuda:2'), in_proj_covar=tensor([0.0343, 0.0360, 0.0298, 0.0323, 0.0391, 0.0392, 0.0321, 0.0427], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:25:59,717 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:07,813 INFO [train.py:904] (2/8) Epoch 7, batch 4450, loss[loss=0.2261, simple_loss=0.3119, pruned_loss=0.0702, over 16467.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3062, pruned_loss=0.07109, over 3200066.37 frames. ], batch size: 75, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:26:08,114 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:17,459 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:42,434 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6010, 4.8522, 4.9599, 4.8536, 4.9139, 5.4478, 4.9711, 4.7069], device='cuda:2'), covar=tensor([0.0900, 0.1255, 0.1140, 0.1333, 0.2118, 0.0781, 0.0929, 0.2098], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0441, 0.0435, 0.0367, 0.0498, 0.0466, 0.0352, 0.0503], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:26:58,767 INFO [optim.py:368] (2/8) 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 16:27:19,263 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:20,139 INFO [train.py:904] (2/8) Epoch 7, batch 4500, loss[loss=0.2025, simple_loss=0.2941, pruned_loss=0.05548, over 16669.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3061, pruned_loss=0.07115, over 3199959.17 frames. ], batch size: 76, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:27:26,480 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:28,723 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:32,337 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:27,210 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:31,302 INFO [train.py:904] (2/8) Epoch 7, batch 4550, loss[loss=0.2133, simple_loss=0.2839, pruned_loss=0.07134, over 17118.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.306, pruned_loss=0.07113, over 3215206.54 frames. ], batch size: 55, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:28:38,293 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:29:10,183 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-28 16:29:15,453 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0506, 3.2954, 3.5371, 3.5097, 3.4969, 3.2578, 3.3105, 3.3412], device='cuda:2'), covar=tensor([0.0376, 0.0512, 0.0366, 0.0397, 0.0463, 0.0428, 0.0771, 0.0488], device='cuda:2'), in_proj_covar=tensor([0.0272, 0.0268, 0.0268, 0.0264, 0.0317, 0.0286, 0.0384, 0.0229], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 16:29:21,410 INFO [optim.py:368] (2/8) 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 16:29:41,307 INFO [train.py:904] (2/8) Epoch 7, batch 4600, loss[loss=0.2324, simple_loss=0.3076, pruned_loss=0.07858, over 16940.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3066, pruned_loss=0.0712, over 3223451.66 frames. ], batch size: 109, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:12,415 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 16:30:24,191 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:30:50,580 INFO [train.py:904] (2/8) Epoch 7, batch 4650, loss[loss=0.2204, simple_loss=0.2959, pruned_loss=0.07251, over 16643.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3051, pruned_loss=0.0707, over 3221551.66 frames. ], batch size: 62, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:52,125 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9151, 5.4507, 5.5776, 5.4277, 5.3708, 6.0541, 5.5713, 5.2912], device='cuda:2'), covar=tensor([0.0782, 0.1589, 0.1205, 0.1629, 0.2484, 0.0914, 0.0986, 0.2136], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0435, 0.0433, 0.0363, 0.0498, 0.0468, 0.0349, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:31:41,970 INFO [optim.py:368] (2/8) 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 16:31:49,823 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:32:03,152 INFO [train.py:904] (2/8) Epoch 7, batch 4700, loss[loss=0.2051, simple_loss=0.2899, pruned_loss=0.06015, over 16732.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3013, pruned_loss=0.06865, over 3231587.43 frames. ], batch size: 89, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:32:26,229 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4566, 4.4620, 4.4998, 3.1479, 4.4143, 1.4996, 4.1735, 4.3250], device='cuda:2'), covar=tensor([0.0208, 0.0117, 0.0132, 0.0926, 0.0163, 0.2683, 0.0167, 0.0319], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0091, 0.0139, 0.0138, 0.0107, 0.0151, 0.0123, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:32:45,689 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:32:56,238 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:33:12,630 INFO [train.py:904] (2/8) Epoch 7, batch 4750, loss[loss=0.174, simple_loss=0.2604, pruned_loss=0.04378, over 17122.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2983, pruned_loss=0.06744, over 3199702.58 frames. ], batch size: 49, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:33:53,098 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:34:03,333 INFO [optim.py:368] (2/8) 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 16:34:22,790 INFO [train.py:904] (2/8) Epoch 7, batch 4800, loss[loss=0.2142, simple_loss=0.2842, pruned_loss=0.07213, over 11748.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.295, pruned_loss=0.06573, over 3202248.27 frames. ], batch size: 246, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:34:36,333 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:34:47,942 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8018, 5.0428, 5.1873, 5.1128, 5.0920, 5.6151, 5.1639, 5.0197], device='cuda:2'), covar=tensor([0.0716, 0.1350, 0.1232, 0.1346, 0.1927, 0.0781, 0.0957, 0.1866], device='cuda:2'), in_proj_covar=tensor([0.0304, 0.0432, 0.0426, 0.0363, 0.0493, 0.0465, 0.0346, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:35:02,962 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6340, 4.6986, 4.8496, 4.8211, 4.7443, 5.3168, 4.9040, 4.6691], device='cuda:2'), covar=tensor([0.0878, 0.1400, 0.1183, 0.1562, 0.2305, 0.0883, 0.0975, 0.2103], device='cuda:2'), in_proj_covar=tensor([0.0304, 0.0432, 0.0428, 0.0364, 0.0494, 0.0467, 0.0345, 0.0502], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:35:36,917 INFO [train.py:904] (2/8) Epoch 7, batch 4850, loss[loss=0.2092, simple_loss=0.2986, pruned_loss=0.05991, over 16646.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2955, pruned_loss=0.06463, over 3218214.99 frames. ], batch size: 134, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:35:45,498 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:36:03,180 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9674, 3.0822, 3.0969, 1.5729, 3.3401, 3.3355, 2.7297, 2.5784], device='cuda:2'), covar=tensor([0.0886, 0.0157, 0.0156, 0.1177, 0.0067, 0.0072, 0.0325, 0.0437], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0095, 0.0082, 0.0137, 0.0069, 0.0085, 0.0118, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 16:36:28,554 INFO [optim.py:368] (2/8) 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 16:36:44,039 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5620, 4.5495, 5.0140, 4.9444, 4.9223, 4.5988, 4.5492, 4.2932], device='cuda:2'), covar=tensor([0.0238, 0.0351, 0.0240, 0.0372, 0.0333, 0.0239, 0.0711, 0.0378], device='cuda:2'), in_proj_covar=tensor([0.0267, 0.0265, 0.0264, 0.0262, 0.0314, 0.0284, 0.0383, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-28 16:36:48,886 INFO [train.py:904] (2/8) Epoch 7, batch 4900, loss[loss=0.2231, simple_loss=0.3062, pruned_loss=0.07004, over 16503.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2952, pruned_loss=0.06364, over 3220329.15 frames. ], batch size: 146, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:37:43,350 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:38:02,372 INFO [train.py:904] (2/8) Epoch 7, batch 4950, loss[loss=0.2147, simple_loss=0.298, pruned_loss=0.06568, over 17144.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2949, pruned_loss=0.06315, over 3214599.34 frames. ], batch size: 47, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:38:05,779 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4044, 2.5917, 2.3828, 4.0525, 3.0148, 3.9457, 1.2772, 2.7284], device='cuda:2'), covar=tensor([0.1467, 0.0679, 0.1149, 0.0083, 0.0223, 0.0319, 0.1602, 0.0853], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0147, 0.0169, 0.0102, 0.0196, 0.0195, 0.0167, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 16:38:53,490 INFO [optim.py:368] (2/8) 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 16:38:53,896 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:39:08,334 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:39:10,120 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2958, 1.9549, 2.0880, 4.0302, 1.8069, 2.5644, 2.0969, 2.2811], device='cuda:2'), covar=tensor([0.0758, 0.2641, 0.1622, 0.0330, 0.3290, 0.1570, 0.2428, 0.2300], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0350, 0.0295, 0.0318, 0.0386, 0.0382, 0.0316, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:39:12,681 INFO [train.py:904] (2/8) Epoch 7, batch 5000, loss[loss=0.2199, simple_loss=0.3174, pruned_loss=0.06125, over 16768.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.297, pruned_loss=0.06375, over 3218460.33 frames. ], batch size: 83, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:40:09,105 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:40:23,701 INFO [train.py:904] (2/8) Epoch 7, batch 5050, loss[loss=0.2169, simple_loss=0.292, pruned_loss=0.07094, over 16361.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2969, pruned_loss=0.06335, over 3226783.47 frames. ], batch size: 35, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:41:14,576 INFO [optim.py:368] (2/8) 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 16:41:16,133 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:41:35,930 INFO [train.py:904] (2/8) Epoch 7, batch 5100, loss[loss=0.1897, simple_loss=0.2765, pruned_loss=0.05144, over 16586.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2948, pruned_loss=0.06256, over 3216806.84 frames. ], batch size: 75, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:42:00,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6353, 2.0121, 1.6294, 1.9295, 2.4955, 2.2233, 2.6762, 2.7430], device='cuda:2'), covar=tensor([0.0064, 0.0298, 0.0349, 0.0295, 0.0144, 0.0239, 0.0099, 0.0140], device='cuda:2'), in_proj_covar=tensor([0.0103, 0.0175, 0.0175, 0.0173, 0.0169, 0.0176, 0.0163, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:42:26,808 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 16:42:46,394 INFO [train.py:904] (2/8) Epoch 7, batch 5150, loss[loss=0.2116, simple_loss=0.3068, pruned_loss=0.05827, over 16775.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2949, pruned_loss=0.06171, over 3209128.91 frames. ], batch size: 102, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:43:37,177 INFO [optim.py:368] (2/8) 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 16:43:55,948 INFO [train.py:904] (2/8) Epoch 7, batch 5200, loss[loss=0.1884, simple_loss=0.2696, pruned_loss=0.05361, over 16507.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2929, pruned_loss=0.06137, over 3225417.72 frames. ], batch size: 75, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:44:03,835 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 16:44:13,021 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 16:44:16,988 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4869, 3.5834, 3.2481, 3.1920, 3.0941, 3.4643, 3.2159, 3.2188], device='cuda:2'), covar=tensor([0.0486, 0.0386, 0.0232, 0.0213, 0.0581, 0.0318, 0.1279, 0.0442], device='cuda:2'), in_proj_covar=tensor([0.0207, 0.0242, 0.0246, 0.0219, 0.0277, 0.0248, 0.0171, 0.0279], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:44:25,574 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1794, 4.8496, 5.1566, 5.3903, 5.5782, 4.9571, 5.4867, 5.4732], device='cuda:2'), covar=tensor([0.1232, 0.0967, 0.1505, 0.0514, 0.0441, 0.0562, 0.0425, 0.0499], device='cuda:2'), in_proj_covar=tensor([0.0440, 0.0538, 0.0673, 0.0548, 0.0410, 0.0408, 0.0416, 0.0465], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:44:56,177 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:45:06,328 INFO [train.py:904] (2/8) Epoch 7, batch 5250, loss[loss=0.2145, simple_loss=0.2956, pruned_loss=0.06673, over 16761.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2899, pruned_loss=0.06063, over 3230024.12 frames. ], batch size: 124, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:45:14,736 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 16:45:19,010 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:45:59,720 INFO [optim.py:368] (2/8) 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 16:46:00,121 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:46:02,350 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1140, 4.1357, 3.9426, 3.8119, 3.6381, 4.0732, 3.8352, 3.7616], device='cuda:2'), covar=tensor([0.0450, 0.0397, 0.0236, 0.0218, 0.0847, 0.0320, 0.0570, 0.0519], device='cuda:2'), in_proj_covar=tensor([0.0212, 0.0248, 0.0252, 0.0223, 0.0284, 0.0255, 0.0174, 0.0287], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:46:06,943 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:46:18,393 INFO [train.py:904] (2/8) Epoch 7, batch 5300, loss[loss=0.2208, simple_loss=0.2913, pruned_loss=0.07519, over 12292.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2874, pruned_loss=0.06006, over 3220169.64 frames. ], batch size: 247, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:46:20,631 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4675, 2.0196, 2.1639, 4.1705, 1.8800, 2.6749, 2.1911, 2.2667], device='cuda:2'), covar=tensor([0.0734, 0.2794, 0.1714, 0.0287, 0.3359, 0.1616, 0.2532, 0.2493], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0348, 0.0293, 0.0315, 0.0386, 0.0379, 0.0313, 0.0412], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:46:23,372 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:46:36,405 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 16:46:45,741 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:46:46,923 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6158, 2.1085, 2.3740, 4.3114, 1.9744, 2.7781, 2.2920, 2.3427], device='cuda:2'), covar=tensor([0.0674, 0.2535, 0.1576, 0.0278, 0.3268, 0.1542, 0.2298, 0.2367], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0348, 0.0294, 0.0316, 0.0386, 0.0379, 0.0313, 0.0413], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:47:06,139 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:47:28,949 INFO [train.py:904] (2/8) Epoch 7, batch 5350, loss[loss=0.241, simple_loss=0.3145, pruned_loss=0.08376, over 12198.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2858, pruned_loss=0.05934, over 3205715.86 frames. ], batch size: 247, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:47:36,966 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:48:20,487 INFO [optim.py:368] (2/8) 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 16:48:40,523 INFO [train.py:904] (2/8) Epoch 7, batch 5400, loss[loss=0.2239, simple_loss=0.3072, pruned_loss=0.07034, over 16280.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.288, pruned_loss=0.05991, over 3202691.26 frames. ], batch size: 165, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:49:05,990 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:49:55,998 INFO [train.py:904] (2/8) Epoch 7, batch 5450, loss[loss=0.2765, simple_loss=0.3468, pruned_loss=0.103, over 16272.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2912, pruned_loss=0.06165, over 3206873.99 frames. ], batch size: 165, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:50:50,980 INFO [optim.py:368] (2/8) 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 16:51:12,555 INFO [train.py:904] (2/8) Epoch 7, batch 5500, loss[loss=0.2446, simple_loss=0.3233, pruned_loss=0.08298, over 16776.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.301, pruned_loss=0.06816, over 3167132.72 frames. ], batch size: 89, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:51:56,567 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9090, 2.2299, 1.7349, 1.9425, 2.6533, 2.3669, 2.9372, 2.8868], device='cuda:2'), covar=tensor([0.0059, 0.0217, 0.0295, 0.0277, 0.0125, 0.0209, 0.0108, 0.0142], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0174, 0.0174, 0.0172, 0.0168, 0.0176, 0.0166, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:52:27,822 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8338, 2.1886, 2.5196, 4.5387, 2.0950, 2.9549, 2.4320, 2.5153], device='cuda:2'), covar=tensor([0.0716, 0.2735, 0.1438, 0.0300, 0.3265, 0.1545, 0.2171, 0.2670], device='cuda:2'), in_proj_covar=tensor([0.0337, 0.0352, 0.0296, 0.0319, 0.0388, 0.0384, 0.0317, 0.0416], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 16:52:29,857 INFO [train.py:904] (2/8) Epoch 7, batch 5550, loss[loss=0.273, simple_loss=0.3353, pruned_loss=0.1053, over 15289.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3088, pruned_loss=0.07381, over 3161622.50 frames. ], batch size: 190, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:53:27,056 INFO [optim.py:368] (2/8) 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 16:53:30,394 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8435, 2.4768, 2.2538, 3.4922, 2.3866, 3.6307, 1.4562, 2.6841], device='cuda:2'), covar=tensor([0.1291, 0.0605, 0.1258, 0.0135, 0.0249, 0.0434, 0.1482, 0.0837], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0148, 0.0171, 0.0104, 0.0198, 0.0196, 0.0168, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 16:53:35,848 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:53:47,082 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:53:49,162 INFO [train.py:904] (2/8) Epoch 7, batch 5600, loss[loss=0.3484, simple_loss=0.3899, pruned_loss=0.1534, over 11257.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3157, pruned_loss=0.08037, over 3103638.65 frames. ], batch size: 248, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:54:11,812 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:54:20,984 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:54:54,203 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:55:08,336 INFO [train.py:904] (2/8) Epoch 7, batch 5650, loss[loss=0.3178, simple_loss=0.3716, pruned_loss=0.132, over 11101.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.323, pruned_loss=0.08699, over 3061416.40 frames. ], batch size: 248, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:55:15,701 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 16:55:51,761 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 16:55:53,656 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:56:03,089 INFO [optim.py:368] (2/8) 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 16:56:23,902 INFO [train.py:904] (2/8) Epoch 7, batch 5700, loss[loss=0.2365, simple_loss=0.324, pruned_loss=0.07451, over 16294.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3247, pruned_loss=0.08853, over 3050055.05 frames. ], batch size: 165, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:56:42,486 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:56:52,874 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6036, 4.9255, 5.1530, 4.9800, 4.9485, 5.5601, 5.0746, 4.8603], device='cuda:2'), covar=tensor([0.0948, 0.1494, 0.1405, 0.1423, 0.2142, 0.0839, 0.1095, 0.2104], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0437, 0.0440, 0.0372, 0.0500, 0.0473, 0.0351, 0.0511], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 16:57:41,028 INFO [train.py:904] (2/8) Epoch 7, batch 5750, loss[loss=0.2899, simple_loss=0.342, pruned_loss=0.1189, over 11204.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3272, pruned_loss=0.08959, over 3057479.54 frames. ], batch size: 247, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:58:39,453 INFO [optim.py:368] (2/8) 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 16:59:02,126 INFO [train.py:904] (2/8) Epoch 7, batch 5800, loss[loss=0.2517, simple_loss=0.3313, pruned_loss=0.08602, over 17204.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.326, pruned_loss=0.08728, over 3082955.30 frames. ], batch size: 45, lr: 9.70e-03, grad_scale: 16.0 2023-04-28 16:59:48,399 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:00:19,938 INFO [train.py:904] (2/8) Epoch 7, batch 5850, loss[loss=0.2494, simple_loss=0.3244, pruned_loss=0.08724, over 16323.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3241, pruned_loss=0.08603, over 3061908.41 frames. ], batch size: 165, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:01:20,001 INFO [optim.py:368] (2/8) 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 17:01:25,690 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:01:38,960 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:01:41,100 INFO [train.py:904] (2/8) Epoch 7, batch 5900, loss[loss=0.2188, simple_loss=0.3011, pruned_loss=0.06819, over 16794.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3233, pruned_loss=0.08535, over 3071363.03 frames. ], batch size: 124, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:02:07,232 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:02:09,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3720, 2.8674, 2.6374, 2.3363, 2.2917, 2.0992, 2.8491, 2.8928], device='cuda:2'), covar=tensor([0.1900, 0.0710, 0.1185, 0.1509, 0.1766, 0.1594, 0.0441, 0.0783], device='cuda:2'), in_proj_covar=tensor([0.0287, 0.0252, 0.0274, 0.0259, 0.0280, 0.0208, 0.0252, 0.0271], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:02:55,659 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5198, 4.5539, 4.3619, 4.1982, 3.9765, 4.4472, 4.2712, 4.0637], device='cuda:2'), covar=tensor([0.0554, 0.0466, 0.0263, 0.0221, 0.0943, 0.0351, 0.0488, 0.0635], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0236, 0.0239, 0.0213, 0.0269, 0.0243, 0.0167, 0.0275], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:02:56,754 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:03:02,860 INFO [train.py:904] (2/8) Epoch 7, batch 5950, loss[loss=0.2573, simple_loss=0.333, pruned_loss=0.09078, over 16179.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3235, pruned_loss=0.08339, over 3075398.23 frames. ], batch size: 35, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:03:23,047 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:03:41,702 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:04:00,334 INFO [optim.py:368] (2/8) 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 17:04:14,735 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3493, 5.7364, 5.1842, 5.6693, 5.2719, 4.7749, 5.4239, 5.8471], device='cuda:2'), covar=tensor([0.1481, 0.1125, 0.1894, 0.0757, 0.1028, 0.1086, 0.1261, 0.1114], device='cuda:2'), in_proj_covar=tensor([0.0437, 0.0554, 0.0473, 0.0363, 0.0345, 0.0374, 0.0462, 0.0407], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:04:22,260 INFO [train.py:904] (2/8) Epoch 7, batch 6000, loss[loss=0.2445, simple_loss=0.3217, pruned_loss=0.0836, over 16923.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3224, pruned_loss=0.08281, over 3102681.79 frames. ], batch size: 116, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:04:22,260 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 17:04:32,870 INFO [train.py:938] (2/8) Epoch 7, validation: loss=0.1758, simple_loss=0.2891, pruned_loss=0.03127, over 944034.00 frames. 2023-04-28 17:04:32,871 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17418MB 2023-04-28 17:04:51,806 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:05:53,514 INFO [train.py:904] (2/8) Epoch 7, batch 6050, loss[loss=0.2165, simple_loss=0.3126, pruned_loss=0.06014, over 16656.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3205, pruned_loss=0.08136, over 3115100.53 frames. ], batch size: 62, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:06:02,626 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:09,558 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:45,834 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-28 17:06:50,891 INFO [optim.py:368] (2/8) 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 17:07:12,265 INFO [train.py:904] (2/8) Epoch 7, batch 6100, loss[loss=0.2983, simple_loss=0.3424, pruned_loss=0.1271, over 11573.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3193, pruned_loss=0.0799, over 3127081.93 frames. ], batch size: 247, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:07:39,902 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:08:32,437 INFO [train.py:904] (2/8) Epoch 7, batch 6150, loss[loss=0.2127, simple_loss=0.2938, pruned_loss=0.06583, over 16844.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.317, pruned_loss=0.07941, over 3121861.20 frames. ], batch size: 116, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:08:38,322 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-04-28 17:09:23,190 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:09:29,158 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:09:31,359 INFO [optim.py:368] (2/8) 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 17:09:51,780 INFO [train.py:904] (2/8) Epoch 7, batch 6200, loss[loss=0.2159, simple_loss=0.2921, pruned_loss=0.06986, over 17225.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3155, pruned_loss=0.07893, over 3130337.78 frames. ], batch size: 52, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:02,507 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 6250, loss[loss=0.2211, simple_loss=0.3054, pruned_loss=0.06836, over 16670.00 frames. ], tot_loss[loss=0.237, simple_loss=0.316, pruned_loss=0.07897, over 3130291.00 frames. ], batch size: 57, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:27,062 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 17:11:52,535 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:12:11,782 INFO [optim.py:368] (2/8) 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 17:12:30,898 INFO [train.py:904] (2/8) Epoch 7, batch 6300, loss[loss=0.2258, simple_loss=0.3097, pruned_loss=0.07096, over 16844.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3157, pruned_loss=0.07823, over 3143864.47 frames. ], batch size: 102, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:12:40,278 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1182, 4.4096, 4.6953, 4.6655, 4.7311, 4.3423, 3.8897, 4.2139], device='cuda:2'), covar=tensor([0.0676, 0.0535, 0.0550, 0.0788, 0.0655, 0.0546, 0.1653, 0.0539], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0279, 0.0283, 0.0274, 0.0331, 0.0301, 0.0404, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 17:13:08,169 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:13:11,282 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 17:13:17,087 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 17:13:49,162 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2807, 2.5066, 1.9034, 2.2443, 2.8669, 2.5516, 3.1586, 3.1200], device='cuda:2'), covar=tensor([0.0050, 0.0229, 0.0341, 0.0271, 0.0140, 0.0207, 0.0123, 0.0132], device='cuda:2'), in_proj_covar=tensor([0.0101, 0.0174, 0.0173, 0.0170, 0.0167, 0.0173, 0.0166, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:13:49,867 INFO [train.py:904] (2/8) Epoch 7, batch 6350, loss[loss=0.2212, simple_loss=0.3038, pruned_loss=0.06925, over 16470.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3166, pruned_loss=0.07979, over 3131731.29 frames. ], batch size: 68, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:14:47,872 INFO [optim.py:368] (2/8) 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 17:15:05,416 INFO [train.py:904] (2/8) Epoch 7, batch 6400, loss[loss=0.2282, simple_loss=0.3056, pruned_loss=0.07538, over 16661.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3171, pruned_loss=0.08117, over 3116172.44 frames. ], batch size: 62, lr: 9.66e-03, grad_scale: 8.0 2023-04-28 17:15:11,979 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8488, 4.8278, 4.8922, 3.3993, 4.2785, 4.6854, 4.2906, 2.7578], device='cuda:2'), covar=tensor([0.0315, 0.0016, 0.0015, 0.0211, 0.0040, 0.0057, 0.0030, 0.0255], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0057, 0.0060, 0.0118, 0.0065, 0.0077, 0.0067, 0.0111], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 17:15:22,341 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:16:15,629 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8685, 2.6589, 2.0778, 2.2637, 3.1512, 2.8138, 3.6906, 3.4808], device='cuda:2'), covar=tensor([0.0031, 0.0214, 0.0286, 0.0266, 0.0121, 0.0187, 0.0111, 0.0110], device='cuda:2'), in_proj_covar=tensor([0.0099, 0.0170, 0.0170, 0.0168, 0.0163, 0.0172, 0.0163, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:16:20,298 INFO [train.py:904] (2/8) Epoch 7, batch 6450, loss[loss=0.2164, simple_loss=0.2942, pruned_loss=0.06933, over 15443.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3158, pruned_loss=0.07949, over 3132665.39 frames. ], batch size: 190, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:16:24,874 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-28 17:16:33,533 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7384, 2.5431, 2.2252, 3.6265, 2.8011, 3.6726, 1.3638, 2.7560], device='cuda:2'), covar=tensor([0.1252, 0.0553, 0.1142, 0.0117, 0.0235, 0.0402, 0.1505, 0.0709], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0147, 0.0169, 0.0104, 0.0196, 0.0197, 0.0167, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 17:17:16,532 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:17:22,055 INFO [optim.py:368] (2/8) 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 17:17:28,507 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6660, 2.9801, 2.5355, 4.6942, 3.7768, 4.4402, 1.6191, 3.0819], device='cuda:2'), covar=tensor([0.1500, 0.0629, 0.1214, 0.0120, 0.0399, 0.0307, 0.1558, 0.0829], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0148, 0.0168, 0.0104, 0.0197, 0.0196, 0.0167, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 17:17:38,435 INFO [train.py:904] (2/8) Epoch 7, batch 6500, loss[loss=0.2387, simple_loss=0.3107, pruned_loss=0.08334, over 15461.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3134, pruned_loss=0.07881, over 3119508.72 frames. ], batch size: 191, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:18:18,229 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7113, 2.5228, 2.2573, 3.4695, 2.7037, 3.6769, 1.3588, 2.7373], device='cuda:2'), covar=tensor([0.1293, 0.0579, 0.1095, 0.0123, 0.0252, 0.0341, 0.1508, 0.0706], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0147, 0.0168, 0.0103, 0.0198, 0.0196, 0.0167, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 17:18:29,446 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:18:36,064 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:18:57,738 INFO [train.py:904] (2/8) Epoch 7, batch 6550, loss[loss=0.2518, simple_loss=0.3353, pruned_loss=0.08418, over 16735.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3166, pruned_loss=0.08006, over 3119518.79 frames. ], batch size: 124, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:19:56,001 INFO [optim.py:368] (2/8) 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 17:20:13,101 INFO [train.py:904] (2/8) Epoch 7, batch 6600, loss[loss=0.2128, simple_loss=0.2975, pruned_loss=0.06411, over 16635.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3198, pruned_loss=0.08176, over 3085396.95 frames. ], batch size: 57, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:20:30,787 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:20:44,150 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 17:21:29,933 INFO [train.py:904] (2/8) Epoch 7, batch 6650, loss[loss=0.2096, simple_loss=0.2956, pruned_loss=0.06184, over 16552.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3205, pruned_loss=0.08257, over 3090461.37 frames. ], batch size: 75, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:21:40,563 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9065, 1.7023, 1.5039, 1.5379, 1.8888, 1.5954, 1.7996, 1.9232], device='cuda:2'), covar=tensor([0.0059, 0.0158, 0.0222, 0.0197, 0.0109, 0.0146, 0.0104, 0.0112], device='cuda:2'), in_proj_covar=tensor([0.0099, 0.0174, 0.0173, 0.0171, 0.0167, 0.0173, 0.0165, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:21:48,637 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.42 vs. limit=5.0 2023-04-28 17:22:02,864 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:22:24,997 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4937, 2.6501, 2.2661, 3.8389, 2.8910, 3.7799, 1.4090, 2.7529], device='cuda:2'), covar=tensor([0.1435, 0.0625, 0.1282, 0.0137, 0.0276, 0.0410, 0.1524, 0.0820], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0148, 0.0171, 0.0104, 0.0198, 0.0197, 0.0167, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 17:22:27,318 INFO [optim.py:368] (2/8) 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 17:22:36,036 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 17:22:43,252 INFO [train.py:904] (2/8) Epoch 7, batch 6700, loss[loss=0.202, simple_loss=0.285, pruned_loss=0.05949, over 16537.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3195, pruned_loss=0.08291, over 3078573.38 frames. ], batch size: 75, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:22:50,792 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 17:22:59,780 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:23:56,589 INFO [train.py:904] (2/8) Epoch 7, batch 6750, loss[loss=0.2145, simple_loss=0.2907, pruned_loss=0.06914, over 16714.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3183, pruned_loss=0.0825, over 3081051.38 frames. ], batch size: 134, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:24:07,649 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6527, 2.8227, 2.5473, 4.2755, 3.2958, 4.0875, 1.3839, 3.0303], device='cuda:2'), covar=tensor([0.1356, 0.0604, 0.1128, 0.0093, 0.0266, 0.0330, 0.1503, 0.0703], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0148, 0.0171, 0.0105, 0.0199, 0.0198, 0.0169, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 17:24:11,054 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:37,892 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:54,244 INFO [optim.py:368] (2/8) 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 17:24:55,933 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67692.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:25:08,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1245, 3.2566, 3.5620, 3.5355, 3.5030, 3.2713, 3.3545, 3.4038], device='cuda:2'), covar=tensor([0.0371, 0.0555, 0.0385, 0.0441, 0.0504, 0.0423, 0.0784, 0.0461], device='cuda:2'), in_proj_covar=tensor([0.0277, 0.0273, 0.0279, 0.0272, 0.0330, 0.0296, 0.0397, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 17:25:10,438 INFO [train.py:904] (2/8) Epoch 7, batch 6800, loss[loss=0.2584, simple_loss=0.3191, pruned_loss=0.09883, over 11815.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3175, pruned_loss=0.08174, over 3088668.51 frames. ], batch size: 248, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:06,144 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:26:10,106 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 6850, loss[loss=0.2674, simple_loss=0.3316, pruned_loss=0.1015, over 11436.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3191, pruned_loss=0.08225, over 3093979.66 frames. ], batch size: 246, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:25,717 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:27:14,419 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:27:20,977 INFO [optim.py:368] (2/8) 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 17:27:35,614 INFO [train.py:904] (2/8) Epoch 7, batch 6900, loss[loss=0.2941, simple_loss=0.3549, pruned_loss=0.1167, over 15426.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.321, pruned_loss=0.08194, over 3102068.45 frames. ], batch size: 191, lr: 9.63e-03, grad_scale: 2.0 2023-04-28 17:28:46,544 INFO [train.py:904] (2/8) Epoch 7, batch 6950, loss[loss=0.2175, simple_loss=0.2999, pruned_loss=0.06754, over 16723.00 frames. ], tot_loss[loss=0.2459, simple_loss=0.3232, pruned_loss=0.08429, over 3084693.17 frames. ], batch size: 89, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:29:13,145 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:29:46,384 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 7, batch 7000, loss[loss=0.2438, simple_loss=0.3294, pruned_loss=0.07908, over 15324.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3233, pruned_loss=0.08373, over 3073963.03 frames. ], batch size: 190, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:31:13,270 INFO [train.py:904] (2/8) Epoch 7, batch 7050, loss[loss=0.3064, simple_loss=0.3549, pruned_loss=0.1289, over 11604.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3231, pruned_loss=0.08236, over 3100497.83 frames. ], batch size: 246, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:31:57,453 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 7, batch 7100, loss[loss=0.2759, simple_loss=0.3301, pruned_loss=0.1108, over 11251.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3217, pruned_loss=0.08261, over 3071984.19 frames. ], batch size: 247, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:32:50,760 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0452, 4.1194, 3.9570, 3.8174, 3.5766, 3.9952, 3.7314, 3.7552], device='cuda:2'), covar=tensor([0.0580, 0.0353, 0.0240, 0.0223, 0.0912, 0.0369, 0.0705, 0.0608], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0232, 0.0233, 0.0207, 0.0264, 0.0238, 0.0166, 0.0271], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:33:20,754 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:30,246 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:39,225 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:33:44,552 INFO [train.py:904] (2/8) Epoch 7, batch 7150, loss[loss=0.2659, simple_loss=0.3341, pruned_loss=0.09891, over 15320.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3206, pruned_loss=0.08268, over 3074362.44 frames. ], batch size: 191, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:33:53,114 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 17:33:54,611 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3868, 4.3579, 4.9391, 4.8705, 4.8561, 4.4574, 4.5461, 4.3229], device='cuda:2'), covar=tensor([0.0273, 0.0432, 0.0234, 0.0358, 0.0397, 0.0292, 0.0755, 0.0384], device='cuda:2'), in_proj_covar=tensor([0.0273, 0.0275, 0.0275, 0.0270, 0.0324, 0.0296, 0.0392, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 17:34:36,442 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:34:44,522 INFO [optim.py:368] (2/8) 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 17:34:48,831 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 17:34:50,394 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4616, 3.7681, 3.8915, 1.4589, 4.1327, 4.3014, 3.0851, 3.0243], device='cuda:2'), covar=tensor([0.0842, 0.0162, 0.0242, 0.1458, 0.0079, 0.0060, 0.0363, 0.0480], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0095, 0.0083, 0.0139, 0.0070, 0.0085, 0.0118, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 17:34:51,460 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7549, 4.8082, 4.6257, 3.9720, 4.6016, 1.8781, 4.4078, 4.5150], device='cuda:2'), covar=tensor([0.0059, 0.0044, 0.0097, 0.0301, 0.0066, 0.1876, 0.0092, 0.0122], device='cuda:2'), in_proj_covar=tensor([0.0103, 0.0090, 0.0138, 0.0134, 0.0105, 0.0155, 0.0122, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:34:58,221 INFO [train.py:904] (2/8) Epoch 7, batch 7200, loss[loss=0.2022, simple_loss=0.3094, pruned_loss=0.04752, over 16768.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3175, pruned_loss=0.08016, over 3084305.43 frames. ], batch size: 102, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:36:11,157 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:36:16,039 INFO [train.py:904] (2/8) Epoch 7, batch 7250, loss[loss=0.2275, simple_loss=0.2978, pruned_loss=0.0786, over 16854.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3148, pruned_loss=0.07852, over 3084906.95 frames. ], batch size: 116, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:36:34,450 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5008, 2.6126, 1.7003, 2.6627, 2.1340, 2.7303, 1.8889, 2.3218], device='cuda:2'), covar=tensor([0.0244, 0.0457, 0.1519, 0.0139, 0.0737, 0.0493, 0.1396, 0.0626], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0155, 0.0180, 0.0094, 0.0163, 0.0192, 0.0189, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 17:36:41,117 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:37:10,521 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-04-28 17:37:16,280 INFO [optim.py:368] (2/8) 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 17:37:25,860 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5553, 2.0976, 2.2776, 4.2935, 2.0295, 2.6094, 2.2624, 2.2796], device='cuda:2'), covar=tensor([0.0723, 0.2692, 0.1630, 0.0306, 0.3221, 0.1686, 0.2300, 0.2804], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0354, 0.0294, 0.0320, 0.0391, 0.0383, 0.0315, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:37:29,955 INFO [train.py:904] (2/8) Epoch 7, batch 7300, loss[loss=0.2255, simple_loss=0.3104, pruned_loss=0.07026, over 16415.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3138, pruned_loss=0.07833, over 3064069.14 frames. ], batch size: 146, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:37:52,949 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:38:44,624 INFO [train.py:904] (2/8) Epoch 7, batch 7350, loss[loss=0.2274, simple_loss=0.3089, pruned_loss=0.07289, over 16221.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3139, pruned_loss=0.07867, over 3046761.42 frames. ], batch size: 165, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:39:02,428 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 17:39:32,959 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4813, 3.3938, 2.7660, 2.1800, 2.4152, 2.2079, 3.4951, 3.3736], device='cuda:2'), covar=tensor([0.2506, 0.0737, 0.1388, 0.1933, 0.1966, 0.1606, 0.0487, 0.0815], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0253, 0.0278, 0.0263, 0.0285, 0.0212, 0.0256, 0.0273], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:39:48,567 INFO [optim.py:368] (2/8) 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 17:39:59,754 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9826, 3.3318, 3.4746, 2.1600, 3.1868, 3.4045, 3.2180, 1.7327], device='cuda:2'), covar=tensor([0.0381, 0.0029, 0.0027, 0.0270, 0.0051, 0.0073, 0.0047, 0.0359], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0056, 0.0059, 0.0116, 0.0065, 0.0077, 0.0067, 0.0111], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 17:40:02,136 INFO [train.py:904] (2/8) Epoch 7, batch 7400, loss[loss=0.2306, simple_loss=0.3206, pruned_loss=0.07029, over 16940.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3154, pruned_loss=0.07958, over 3048281.25 frames. ], batch size: 90, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:40:53,655 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:40:54,928 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:14,131 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:19,088 INFO [train.py:904] (2/8) Epoch 7, batch 7450, loss[loss=0.2554, simple_loss=0.339, pruned_loss=0.08593, over 15374.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3169, pruned_loss=0.081, over 3040757.20 frames. ], batch size: 190, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:42:11,320 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 17:42:30,794 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:35,742 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:39,839 INFO [train.py:904] (2/8) Epoch 7, batch 7500, loss[loss=0.2246, simple_loss=0.3074, pruned_loss=0.07087, over 16897.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3174, pruned_loss=0.08028, over 3050645.42 frames. ], batch size: 109, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:43:43,733 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:43:57,919 INFO [train.py:904] (2/8) Epoch 7, batch 7550, loss[loss=0.2241, simple_loss=0.3044, pruned_loss=0.07192, over 16848.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3165, pruned_loss=0.08001, over 3070585.96 frames. ], batch size: 102, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:44:10,860 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:44:41,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4325, 4.5612, 4.6648, 4.6164, 4.5606, 5.0893, 4.6507, 4.4483], device='cuda:2'), covar=tensor([0.1132, 0.1581, 0.1619, 0.1600, 0.2238, 0.0936, 0.1373, 0.2241], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0427, 0.0440, 0.0367, 0.0491, 0.0467, 0.0352, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 17:44:48,176 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9849, 2.3345, 2.6588, 4.7551, 2.1718, 3.0478, 2.5007, 2.6114], device='cuda:2'), covar=tensor([0.0711, 0.3037, 0.1552, 0.0292, 0.3558, 0.1679, 0.2552, 0.2854], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0357, 0.0296, 0.0322, 0.0394, 0.0382, 0.0316, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:44:59,735 INFO [optim.py:368] (2/8) 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 17:45:12,771 INFO [train.py:904] (2/8) Epoch 7, batch 7600, loss[loss=0.243, simple_loss=0.3105, pruned_loss=0.08779, over 11707.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3158, pruned_loss=0.08026, over 3056649.63 frames. ], batch size: 247, lr: 9.58e-03, grad_scale: 8.0 2023-04-28 17:46:00,315 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 17:46:28,266 INFO [train.py:904] (2/8) Epoch 7, batch 7650, loss[loss=0.2192, simple_loss=0.3067, pruned_loss=0.06588, over 16814.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3158, pruned_loss=0.08005, over 3077261.29 frames. ], batch size: 83, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:47:01,507 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.55 vs. limit=5.0 2023-04-28 17:47:08,463 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:47:26,484 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8246, 1.9491, 2.2256, 3.1514, 2.0531, 2.2791, 2.2140, 2.0000], device='cuda:2'), covar=tensor([0.0743, 0.2539, 0.1475, 0.0443, 0.3076, 0.1611, 0.2133, 0.2632], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0355, 0.0296, 0.0323, 0.0394, 0.0383, 0.0315, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:47:29,673 INFO [optim.py:368] (2/8) 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 17:47:42,702 INFO [train.py:904] (2/8) Epoch 7, batch 7700, loss[loss=0.2452, simple_loss=0.3186, pruned_loss=0.08593, over 16763.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3157, pruned_loss=0.08032, over 3085015.94 frames. ], batch size: 124, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:47:58,951 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3675, 2.9364, 2.6061, 2.3299, 2.2740, 2.1433, 2.9155, 2.9416], device='cuda:2'), covar=tensor([0.1899, 0.0620, 0.1121, 0.1492, 0.1744, 0.1515, 0.0413, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0255, 0.0279, 0.0262, 0.0283, 0.0212, 0.0254, 0.0273], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:48:34,674 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:38,418 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:56,281 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 17:48:57,123 INFO [train.py:904] (2/8) Epoch 7, batch 7750, loss[loss=0.2357, simple_loss=0.3234, pruned_loss=0.07402, over 16813.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3162, pruned_loss=0.08023, over 3088714.73 frames. ], batch size: 83, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:49:34,232 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-28 17:49:44,494 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:49:44,611 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 17:50:11,067 INFO [train.py:904] (2/8) Epoch 7, batch 7800, loss[loss=0.2601, simple_loss=0.3259, pruned_loss=0.09713, over 15243.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3179, pruned_loss=0.08158, over 3083520.66 frames. ], batch size: 190, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:11,943 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 17:51:12,887 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:51:17,563 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 7850, loss[loss=0.2391, simple_loss=0.3272, pruned_loss=0.07546, over 16842.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3183, pruned_loss=0.08132, over 3080087.44 frames. ], batch size: 39, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:30,857 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:52:26,873 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 7, batch 7900, loss[loss=0.249, simple_loss=0.3297, pruned_loss=0.08419, over 16625.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3177, pruned_loss=0.0807, over 3078611.96 frames. ], batch size: 134, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:53:58,015 INFO [train.py:904] (2/8) Epoch 7, batch 7950, loss[loss=0.2316, simple_loss=0.3023, pruned_loss=0.08049, over 17076.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3175, pruned_loss=0.08083, over 3070773.55 frames. ], batch size: 53, lr: 9.55e-03, grad_scale: 2.0 2023-04-28 17:55:03,363 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 7, batch 8000, loss[loss=0.2159, simple_loss=0.307, pruned_loss=0.06239, over 16872.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3174, pruned_loss=0.08095, over 3073515.22 frames. ], batch size: 96, lr: 9.55e-03, grad_scale: 4.0 2023-04-28 17:56:00,096 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 8050, loss[loss=0.2509, simple_loss=0.3374, pruned_loss=0.0822, over 16409.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3174, pruned_loss=0.08071, over 3088637.55 frames. ], batch size: 146, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:57:00,532 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7735, 4.4229, 4.7389, 4.9267, 5.1011, 4.5432, 5.0588, 5.0171], device='cuda:2'), covar=tensor([0.1052, 0.0988, 0.1175, 0.0528, 0.0393, 0.0661, 0.0428, 0.0517], device='cuda:2'), in_proj_covar=tensor([0.0441, 0.0545, 0.0677, 0.0544, 0.0413, 0.0418, 0.0438, 0.0473], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 17:57:30,024 INFO [optim.py:368] (2/8) 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 17:57:41,502 INFO [train.py:904] (2/8) Epoch 7, batch 8100, loss[loss=0.2259, simple_loss=0.2942, pruned_loss=0.07876, over 16545.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3166, pruned_loss=0.0798, over 3099698.38 frames. ], batch size: 62, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:58:11,794 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0960, 3.1278, 3.1477, 1.6986, 3.3484, 3.3606, 2.6034, 2.6073], device='cuda:2'), covar=tensor([0.0833, 0.0149, 0.0173, 0.1081, 0.0052, 0.0105, 0.0416, 0.0421], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0094, 0.0082, 0.0138, 0.0068, 0.0085, 0.0119, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 17:58:40,377 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:43,509 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:55,987 INFO [train.py:904] (2/8) Epoch 7, batch 8150, loss[loss=0.2091, simple_loss=0.2825, pruned_loss=0.06783, over 17044.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3145, pruned_loss=0.07904, over 3124735.05 frames. ], batch size: 50, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:59:00,806 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:59:03,731 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8099, 4.1100, 4.3096, 1.9110, 4.4796, 4.5519, 3.1749, 3.4546], device='cuda:2'), covar=tensor([0.0684, 0.0129, 0.0142, 0.1158, 0.0035, 0.0058, 0.0327, 0.0333], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0093, 0.0081, 0.0137, 0.0067, 0.0084, 0.0117, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 17:59:59,698 INFO [optim.py:368] (2/8) 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 18:00:10,163 INFO [train.py:904] (2/8) Epoch 7, batch 8200, loss[loss=0.2074, simple_loss=0.2917, pruned_loss=0.06153, over 16527.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3116, pruned_loss=0.07832, over 3120613.90 frames. ], batch size: 68, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:00:12,049 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:00:13,541 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:20,641 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0516, 3.3490, 3.5951, 3.5805, 3.5757, 3.3634, 3.3726, 3.4368], device='cuda:2'), covar=tensor([0.0385, 0.0539, 0.0386, 0.0415, 0.0426, 0.0420, 0.0840, 0.0406], device='cuda:2'), in_proj_covar=tensor([0.0273, 0.0275, 0.0277, 0.0268, 0.0321, 0.0294, 0.0392, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 18:01:30,736 INFO [train.py:904] (2/8) Epoch 7, batch 8250, loss[loss=0.199, simple_loss=0.287, pruned_loss=0.05546, over 16777.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3102, pruned_loss=0.07559, over 3105273.95 frames. ], batch size: 124, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:01:47,268 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:58,208 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:02:28,786 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-28 18:02:40,829 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 7, batch 8300, loss[loss=0.189, simple_loss=0.2846, pruned_loss=0.04672, over 16498.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3068, pruned_loss=0.07256, over 3081523.39 frames. ], batch size: 68, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:02:56,333 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0218, 2.5009, 2.2932, 2.9871, 2.1503, 3.3054, 1.6749, 2.8091], device='cuda:2'), covar=tensor([0.1158, 0.0453, 0.0867, 0.0098, 0.0122, 0.0375, 0.1248, 0.0616], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0146, 0.0168, 0.0105, 0.0195, 0.0193, 0.0167, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 18:03:21,446 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 18:03:25,495 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:29,145 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:34,954 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:42,606 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:04:00,582 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:04:11,882 INFO [train.py:904] (2/8) Epoch 7, batch 8350, loss[loss=0.2515, simple_loss=0.317, pruned_loss=0.09302, over 11919.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3059, pruned_loss=0.07018, over 3084929.12 frames. ], batch size: 247, lr: 9.52e-03, grad_scale: 4.0 2023-04-28 18:04:44,397 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2995, 5.8240, 5.9733, 5.7473, 5.8205, 6.3140, 5.7859, 5.6085], device='cuda:2'), covar=tensor([0.0617, 0.1340, 0.1512, 0.1992, 0.2325, 0.0935, 0.1239, 0.2232], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0405, 0.0425, 0.0352, 0.0472, 0.0450, 0.0340, 0.0483], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 18:04:46,443 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:04:59,338 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:06,532 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:20,698 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 7, batch 8400, loss[loss=0.2099, simple_loss=0.298, pruned_loss=0.06094, over 16757.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3027, pruned_loss=0.06796, over 3065403.50 frames. ], batch size: 124, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:05:39,358 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:06:02,569 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:23,771 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:06:33,340 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 18:06:35,286 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:52,099 INFO [train.py:904] (2/8) Epoch 7, batch 8450, loss[loss=0.209, simple_loss=0.2922, pruned_loss=0.06288, over 15224.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.3009, pruned_loss=0.06614, over 3067379.18 frames. ], batch size: 190, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:06:54,345 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8001, 3.6696, 3.8434, 3.9625, 4.0240, 3.6027, 3.9872, 4.0644], device='cuda:2'), covar=tensor([0.1076, 0.0768, 0.1063, 0.0546, 0.0472, 0.1475, 0.0512, 0.0461], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0527, 0.0655, 0.0533, 0.0403, 0.0407, 0.0423, 0.0461], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:07:18,693 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:38,250 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:58,567 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0550, 4.6760, 4.8223, 5.1957, 5.3289, 4.7743, 5.3295, 5.3573], device='cuda:2'), covar=tensor([0.1269, 0.1097, 0.1834, 0.0803, 0.0620, 0.0623, 0.0659, 0.0636], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0527, 0.0653, 0.0532, 0.0399, 0.0408, 0.0422, 0.0460], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:07:58,605 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:59,251 INFO [optim.py:368] (2/8) 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 18:08:06,558 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 7, batch 8500, loss[loss=0.1854, simple_loss=0.2713, pruned_loss=0.04974, over 15300.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2961, pruned_loss=0.06307, over 3044555.20 frames. ], batch size: 190, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:08:31,457 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3812, 4.3583, 4.1667, 3.6616, 4.1947, 1.7817, 4.0106, 4.0792], device='cuda:2'), covar=tensor([0.0065, 0.0068, 0.0127, 0.0268, 0.0075, 0.1973, 0.0109, 0.0152], device='cuda:2'), in_proj_covar=tensor([0.0102, 0.0090, 0.0138, 0.0130, 0.0104, 0.0156, 0.0122, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:08:55,052 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3720, 3.4827, 1.8548, 3.7497, 2.3147, 3.6881, 1.9574, 2.8425], device='cuda:2'), covar=tensor([0.0161, 0.0259, 0.1400, 0.0087, 0.0808, 0.0409, 0.1463, 0.0576], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0148, 0.0173, 0.0090, 0.0158, 0.0182, 0.0184, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 18:08:55,075 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:09:33,177 INFO [train.py:904] (2/8) Epoch 7, batch 8550, loss[loss=0.2224, simple_loss=0.3074, pruned_loss=0.06872, over 15287.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2935, pruned_loss=0.06202, over 3023907.00 frames. ], batch size: 191, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:09:41,233 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:10:58,765 INFO [optim.py:368] (2/8) 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 18:11:12,034 INFO [train.py:904] (2/8) Epoch 7, batch 8600, loss[loss=0.2046, simple_loss=0.2799, pruned_loss=0.06461, over 12136.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2937, pruned_loss=0.06084, over 3026046.98 frames. ], batch size: 250, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:11:46,306 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:11:55,638 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:12:47,741 INFO [train.py:904] (2/8) Epoch 7, batch 8650, loss[loss=0.2121, simple_loss=0.2882, pruned_loss=0.068, over 12098.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2917, pruned_loss=0.05923, over 3020320.21 frames. ], batch size: 248, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:13:17,550 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9483, 5.2780, 4.9530, 4.9928, 4.7219, 4.6196, 4.7204, 5.3249], device='cuda:2'), covar=tensor([0.0776, 0.0843, 0.1126, 0.0602, 0.0734, 0.0800, 0.0879, 0.0843], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0536, 0.0452, 0.0358, 0.0335, 0.0364, 0.0447, 0.0393], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:13:53,716 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:14:19,150 INFO [optim.py:368] (2/8) 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 18:14:31,858 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:14:32,611 INFO [train.py:904] (2/8) Epoch 7, batch 8700, loss[loss=0.1806, simple_loss=0.2705, pruned_loss=0.04536, over 16836.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2881, pruned_loss=0.05705, over 3039641.55 frames. ], batch size: 83, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:15:22,789 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:16:08,632 INFO [train.py:904] (2/8) Epoch 7, batch 8750, loss[loss=0.2118, simple_loss=0.3023, pruned_loss=0.06066, over 16801.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2883, pruned_loss=0.05687, over 3032570.19 frames. ], batch size: 124, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:17:04,324 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:17:07,496 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:17:23,898 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0368, 1.3225, 1.6596, 1.9908, 2.0600, 2.0873, 1.5989, 2.2153], device='cuda:2'), covar=tensor([0.0153, 0.0312, 0.0178, 0.0188, 0.0189, 0.0155, 0.0280, 0.0074], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0151, 0.0134, 0.0133, 0.0143, 0.0098, 0.0150, 0.0089], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 18:17:32,178 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4397, 1.9116, 2.0790, 4.0399, 1.8398, 2.4679, 2.1037, 2.0651], device='cuda:2'), covar=tensor([0.0709, 0.3019, 0.1772, 0.0299, 0.3530, 0.1782, 0.2506, 0.3018], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0343, 0.0290, 0.0307, 0.0383, 0.0367, 0.0308, 0.0400], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:17:45,986 INFO [optim.py:368] (2/8) 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 18:17:55,723 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:18:01,165 INFO [train.py:904] (2/8) Epoch 7, batch 8800, loss[loss=0.2102, simple_loss=0.2978, pruned_loss=0.06126, over 16703.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.287, pruned_loss=0.05589, over 3034447.70 frames. ], batch size: 134, lr: 9.49e-03, grad_scale: 8.0 2023-04-28 18:18:47,924 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:11,511 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:19:13,366 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 18:19:33,327 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-04-28 18:19:36,494 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:43,876 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:46,919 INFO [train.py:904] (2/8) Epoch 7, batch 8850, loss[loss=0.1845, simple_loss=0.2913, pruned_loss=0.03882, over 16734.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2894, pruned_loss=0.05484, over 3045454.82 frames. ], batch size: 134, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:21:21,322 INFO [optim.py:368] (2/8) 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 18:21:26,795 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3446, 4.3135, 4.1474, 3.9835, 3.8332, 4.2094, 4.1256, 3.9355], device='cuda:2'), covar=tensor([0.0466, 0.0357, 0.0240, 0.0226, 0.0782, 0.0338, 0.0397, 0.0556], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0227, 0.0228, 0.0203, 0.0250, 0.0229, 0.0160, 0.0265], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:21:34,922 INFO [train.py:904] (2/8) Epoch 7, batch 8900, loss[loss=0.1812, simple_loss=0.268, pruned_loss=0.04724, over 12790.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2892, pruned_loss=0.05409, over 3052733.59 frames. ], batch size: 248, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:22:06,687 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:22:21,476 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:23:38,920 INFO [train.py:904] (2/8) Epoch 7, batch 8950, loss[loss=0.1981, simple_loss=0.2806, pruned_loss=0.05775, over 12985.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2891, pruned_loss=0.05445, over 3058559.30 frames. ], batch size: 250, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:23:53,825 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 18:24:09,509 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:21,486 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:42,892 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:25:14,731 INFO [optim.py:368] (2/8) 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 18:25:27,266 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:25:27,970 INFO [train.py:904] (2/8) Epoch 7, batch 9000, loss[loss=0.2125, simple_loss=0.2921, pruned_loss=0.06645, over 12108.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2859, pruned_loss=0.05295, over 3062679.61 frames. ], batch size: 248, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:25:27,970 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 18:25:37,193 INFO [train.py:938] (2/8) Epoch 7, validation: loss=0.1647, simple_loss=0.2682, pruned_loss=0.03062, over 944034.00 frames. 2023-04-28 18:25:37,194 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-28 18:26:33,160 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:26:33,203 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:27:14,336 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:27:19,733 INFO [train.py:904] (2/8) Epoch 7, batch 9050, loss[loss=0.1929, simple_loss=0.2804, pruned_loss=0.05272, over 16368.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.287, pruned_loss=0.05385, over 3069305.69 frames. ], batch size: 146, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:28:08,387 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:28:09,987 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:28:50,067 INFO [optim.py:368] (2/8) 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 18:29:06,545 INFO [train.py:904] (2/8) Epoch 7, batch 9100, loss[loss=0.2149, simple_loss=0.3052, pruned_loss=0.06231, over 15378.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2869, pruned_loss=0.05447, over 3067917.21 frames. ], batch size: 191, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:29:58,209 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:29:58,248 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:30:13,171 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:31:01,026 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:31:03,203 INFO [train.py:904] (2/8) Epoch 7, batch 9150, loss[loss=0.2155, simple_loss=0.2916, pruned_loss=0.06971, over 12086.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.288, pruned_loss=0.05425, over 3080843.77 frames. ], batch size: 247, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:31:47,574 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:35,334 INFO [optim.py:368] (2/8) 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 18:32:39,003 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:45,141 INFO [train.py:904] (2/8) Epoch 7, batch 9200, loss[loss=0.1964, simple_loss=0.2867, pruned_loss=0.05306, over 16254.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2824, pruned_loss=0.05284, over 3074751.03 frames. ], batch size: 165, lr: 9.47e-03, grad_scale: 8.0 2023-04-28 18:33:01,823 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:33:46,486 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1820, 4.9098, 5.1332, 5.3690, 5.5399, 4.7723, 5.4898, 5.4527], device='cuda:2'), covar=tensor([0.1170, 0.0931, 0.1221, 0.0493, 0.0457, 0.0614, 0.0404, 0.0524], device='cuda:2'), in_proj_covar=tensor([0.0417, 0.0522, 0.0637, 0.0521, 0.0395, 0.0400, 0.0411, 0.0457], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:34:22,200 INFO [train.py:904] (2/8) Epoch 7, batch 9250, loss[loss=0.1804, simple_loss=0.2722, pruned_loss=0.04428, over 16866.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2818, pruned_loss=0.05266, over 3074747.84 frames. ], batch size: 96, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:35:01,680 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:36:01,186 INFO [optim.py:368] (2/8) 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 18:36:13,551 INFO [train.py:904] (2/8) Epoch 7, batch 9300, loss[loss=0.1823, simple_loss=0.2759, pruned_loss=0.04434, over 16724.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2798, pruned_loss=0.05185, over 3065678.57 frames. ], batch size: 134, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:36:22,839 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1963, 3.3622, 3.6382, 3.6110, 3.6030, 3.3760, 3.4360, 3.5039], device='cuda:2'), covar=tensor([0.0312, 0.0503, 0.0343, 0.0362, 0.0397, 0.0402, 0.0690, 0.0350], device='cuda:2'), in_proj_covar=tensor([0.0252, 0.0255, 0.0254, 0.0245, 0.0292, 0.0275, 0.0357, 0.0223], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 18:37:43,884 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8202, 4.1763, 3.4174, 2.4390, 2.9476, 2.5568, 4.3146, 3.8261], device='cuda:2'), covar=tensor([0.2342, 0.0558, 0.1225, 0.1831, 0.1905, 0.1552, 0.0355, 0.0736], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0241, 0.0266, 0.0253, 0.0247, 0.0205, 0.0244, 0.0257], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:37:59,027 INFO [train.py:904] (2/8) Epoch 7, batch 9350, loss[loss=0.2023, simple_loss=0.2923, pruned_loss=0.05609, over 16154.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2797, pruned_loss=0.05213, over 3073492.19 frames. ], batch size: 165, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:38:18,311 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:39:29,825 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3131, 3.0028, 2.6379, 2.2293, 2.1663, 2.1079, 2.8696, 2.8404], device='cuda:2'), covar=tensor([0.2260, 0.0666, 0.1210, 0.1750, 0.2001, 0.1740, 0.0375, 0.0881], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0240, 0.0265, 0.0252, 0.0245, 0.0205, 0.0242, 0.0256], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:39:31,294 INFO [optim.py:368] (2/8) 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 18:39:40,121 INFO [train.py:904] (2/8) Epoch 7, batch 9400, loss[loss=0.216, simple_loss=0.3106, pruned_loss=0.06073, over 16336.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2789, pruned_loss=0.05154, over 3054485.74 frames. ], batch size: 146, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:40:18,235 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:40:36,518 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:41:20,563 INFO [train.py:904] (2/8) Epoch 7, batch 9450, loss[loss=0.1761, simple_loss=0.2701, pruned_loss=0.04108, over 16843.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2814, pruned_loss=0.05207, over 3055212.28 frames. ], batch size: 90, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:42:14,169 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:42:39,374 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 18:42:41,154 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6927, 2.7774, 1.8258, 2.8695, 2.1290, 2.8116, 2.0206, 2.4524], device='cuda:2'), covar=tensor([0.0194, 0.0325, 0.1180, 0.0139, 0.0648, 0.0652, 0.1143, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0148, 0.0174, 0.0090, 0.0156, 0.0179, 0.0184, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 18:42:50,912 INFO [optim.py:368] (2/8) 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 18:43:02,095 INFO [train.py:904] (2/8) Epoch 7, batch 9500, loss[loss=0.2142, simple_loss=0.297, pruned_loss=0.06568, over 15171.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2809, pruned_loss=0.05167, over 3054809.22 frames. ], batch size: 191, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:43:21,886 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1271, 3.3966, 2.8121, 5.0599, 4.1261, 4.8027, 1.6251, 3.5927], device='cuda:2'), covar=tensor([0.1309, 0.0546, 0.1055, 0.0107, 0.0212, 0.0242, 0.1452, 0.0577], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0146, 0.0169, 0.0103, 0.0175, 0.0194, 0.0170, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 18:44:47,823 INFO [train.py:904] (2/8) Epoch 7, batch 9550, loss[loss=0.2127, simple_loss=0.3078, pruned_loss=0.05879, over 16182.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2803, pruned_loss=0.05149, over 3061579.82 frames. ], batch size: 165, lr: 9.44e-03, grad_scale: 4.0 2023-04-28 18:45:18,541 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:46:18,720 INFO [optim.py:368] (2/8) 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 18:46:26,839 INFO [train.py:904] (2/8) Epoch 7, batch 9600, loss[loss=0.2262, simple_loss=0.314, pruned_loss=0.06917, over 16994.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2815, pruned_loss=0.05233, over 3062279.52 frames. ], batch size: 109, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:46:47,925 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 18:48:15,878 INFO [train.py:904] (2/8) Epoch 7, batch 9650, loss[loss=0.2121, simple_loss=0.3013, pruned_loss=0.06142, over 15375.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2839, pruned_loss=0.05307, over 3044509.31 frames. ], batch size: 191, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:49:55,419 INFO [optim.py:368] (2/8) 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 18:50:03,052 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6367, 4.3807, 4.6578, 4.8235, 5.0075, 4.4513, 4.9943, 4.9485], device='cuda:2'), covar=tensor([0.1234, 0.1015, 0.1241, 0.0590, 0.0429, 0.0716, 0.0420, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0520, 0.0634, 0.0523, 0.0392, 0.0398, 0.0401, 0.0451], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:50:06,083 INFO [train.py:904] (2/8) Epoch 7, batch 9700, loss[loss=0.1874, simple_loss=0.2765, pruned_loss=0.04916, over 16789.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2827, pruned_loss=0.05254, over 3057346.79 frames. ], batch size: 124, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:50:33,465 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:51:44,049 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6934, 2.9264, 2.6636, 4.6074, 3.5153, 4.4195, 1.3968, 3.2354], device='cuda:2'), covar=tensor([0.1449, 0.0620, 0.1021, 0.0090, 0.0175, 0.0282, 0.1538, 0.0664], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0145, 0.0168, 0.0104, 0.0172, 0.0194, 0.0169, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 18:51:47,796 INFO [train.py:904] (2/8) Epoch 7, batch 9750, loss[loss=0.1878, simple_loss=0.2817, pruned_loss=0.04693, over 15356.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.282, pruned_loss=0.05257, over 3057085.60 frames. ], batch size: 191, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:53:18,482 INFO [optim.py:368] (2/8) 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 18:53:27,255 INFO [train.py:904] (2/8) Epoch 7, batch 9800, loss[loss=0.1876, simple_loss=0.2783, pruned_loss=0.04846, over 16757.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2825, pruned_loss=0.05192, over 3069755.54 frames. ], batch size: 39, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:53:40,261 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0254, 3.1131, 3.0736, 2.1195, 2.9366, 3.0004, 3.0714, 1.9317], device='cuda:2'), covar=tensor([0.0320, 0.0024, 0.0030, 0.0246, 0.0060, 0.0062, 0.0038, 0.0304], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0056, 0.0059, 0.0117, 0.0065, 0.0073, 0.0065, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 18:54:36,678 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-28 18:55:12,741 INFO [train.py:904] (2/8) Epoch 7, batch 9850, loss[loss=0.1654, simple_loss=0.2701, pruned_loss=0.0303, over 17184.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2831, pruned_loss=0.05098, over 3060973.71 frames. ], batch size: 46, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:55:43,255 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:56:54,661 INFO [optim.py:368] (2/8) 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 18:57:00,194 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3944, 1.8980, 1.6345, 1.6103, 2.1964, 1.8650, 2.1728, 2.3217], device='cuda:2'), covar=tensor([0.0055, 0.0243, 0.0302, 0.0284, 0.0136, 0.0227, 0.0104, 0.0136], device='cuda:2'), in_proj_covar=tensor([0.0097, 0.0175, 0.0172, 0.0170, 0.0168, 0.0173, 0.0156, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 18:57:04,338 INFO [train.py:904] (2/8) Epoch 7, batch 9900, loss[loss=0.2034, simple_loss=0.2922, pruned_loss=0.0573, over 16925.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2831, pruned_loss=0.0508, over 3051136.64 frames. ], batch size: 116, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:57:33,007 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:59:01,222 INFO [train.py:904] (2/8) Epoch 7, batch 9950, loss[loss=0.1726, simple_loss=0.2698, pruned_loss=0.03771, over 17158.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2841, pruned_loss=0.05078, over 3045077.00 frames. ], batch size: 48, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 19:00:47,805 INFO [optim.py:368] (2/8) 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 19:01:00,568 INFO [train.py:904] (2/8) Epoch 7, batch 10000, loss[loss=0.1825, simple_loss=0.2749, pruned_loss=0.04508, over 16520.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2827, pruned_loss=0.0501, over 3073820.87 frames. ], batch size: 62, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:01:29,754 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:02:18,941 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 19:02:40,827 INFO [train.py:904] (2/8) Epoch 7, batch 10050, loss[loss=0.2036, simple_loss=0.2934, pruned_loss=0.05689, over 15377.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2826, pruned_loss=0.05002, over 3071652.51 frames. ], batch size: 190, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:02:51,297 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5499, 2.4968, 2.0027, 3.8620, 2.4543, 3.7900, 1.3470, 2.4867], device='cuda:2'), covar=tensor([0.1470, 0.0728, 0.1439, 0.0124, 0.0178, 0.0408, 0.1616, 0.1006], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0145, 0.0169, 0.0102, 0.0168, 0.0193, 0.0169, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 19:03:04,143 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:03:17,119 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6826, 4.9941, 5.0368, 4.9966, 4.9595, 5.5258, 4.9923, 4.7285], device='cuda:2'), covar=tensor([0.0860, 0.1383, 0.1581, 0.1405, 0.2147, 0.0833, 0.1399, 0.2360], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0401, 0.0423, 0.0353, 0.0460, 0.0440, 0.0333, 0.0469], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:03:56,369 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:04:04,309 INFO [optim.py:368] (2/8) 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 19:04:12,142 INFO [train.py:904] (2/8) Epoch 7, batch 10100, loss[loss=0.1874, simple_loss=0.2716, pruned_loss=0.05163, over 15524.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2828, pruned_loss=0.05022, over 3065731.88 frames. ], batch size: 191, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:05:07,003 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 8, batch 0, loss[loss=0.2921, simple_loss=0.3446, pruned_loss=0.1198, over 16520.00 frames. ], tot_loss[loss=0.2921, simple_loss=0.3446, pruned_loss=0.1198, over 16520.00 frames. ], batch size: 75, lr: 8.86e-03, grad_scale: 8.0 2023-04-28 19:05:54,945 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 19:06:02,572 INFO [train.py:938] (2/8) Epoch 8, validation: loss=0.1627, simple_loss=0.2663, pruned_loss=0.02958, over 944034.00 frames. 2023-04-28 19:06:02,572 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-28 19:06:04,012 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:06:13,870 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9912, 5.7190, 5.8149, 5.6181, 5.6857, 6.2100, 5.7984, 5.6355], device='cuda:2'), covar=tensor([0.0637, 0.1507, 0.1707, 0.1695, 0.2149, 0.0798, 0.1135, 0.2153], device='cuda:2'), in_proj_covar=tensor([0.0284, 0.0403, 0.0419, 0.0355, 0.0464, 0.0444, 0.0334, 0.0471], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:06:55,619 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71091.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:07:07,690 INFO [optim.py:368] (2/8) 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 19:07:09,983 INFO [train.py:904] (2/8) Epoch 8, batch 50, loss[loss=0.1812, simple_loss=0.2774, pruned_loss=0.04255, over 17137.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.2995, pruned_loss=0.07676, over 760409.99 frames. ], batch size: 49, lr: 8.86e-03, grad_scale: 1.0 2023-04-28 19:08:10,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1300, 4.7529, 4.9525, 5.3225, 5.4438, 4.8767, 5.3469, 5.4135], device='cuda:2'), covar=tensor([0.1156, 0.1021, 0.1775, 0.0616, 0.0545, 0.0533, 0.0558, 0.0529], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0542, 0.0661, 0.0543, 0.0410, 0.0411, 0.0426, 0.0470], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:08:17,899 INFO [train.py:904] (2/8) Epoch 8, batch 100, loss[loss=0.1965, simple_loss=0.2838, pruned_loss=0.05462, over 17201.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2924, pruned_loss=0.07019, over 1324650.14 frames. ], batch size: 46, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:09:00,026 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4446, 4.5211, 4.2995, 4.2137, 3.7709, 4.4262, 4.3761, 4.0537], device='cuda:2'), covar=tensor([0.0641, 0.0395, 0.0343, 0.0288, 0.1119, 0.0421, 0.0470, 0.0628], device='cuda:2'), in_proj_covar=tensor([0.0203, 0.0238, 0.0238, 0.0211, 0.0265, 0.0240, 0.0162, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:09:04,201 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 19:09:07,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9088, 5.0189, 4.7524, 4.5144, 4.0517, 4.8056, 4.8652, 4.4063], device='cuda:2'), covar=tensor([0.0666, 0.0365, 0.0355, 0.0326, 0.1438, 0.0442, 0.0312, 0.0721], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0240, 0.0239, 0.0212, 0.0267, 0.0242, 0.0164, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:09:14,393 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0668, 4.9376, 4.7843, 4.5061, 4.3914, 4.7753, 4.8887, 4.4983], device='cuda:2'), covar=tensor([0.0506, 0.0375, 0.0294, 0.0283, 0.1075, 0.0386, 0.0259, 0.0695], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0240, 0.0239, 0.0212, 0.0266, 0.0242, 0.0163, 0.0278], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:09:23,827 INFO [optim.py:368] (2/8) 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 19:09:26,730 INFO [train.py:904] (2/8) Epoch 8, batch 150, loss[loss=0.2653, simple_loss=0.3226, pruned_loss=0.104, over 16713.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2918, pruned_loss=0.06933, over 1762338.12 frames. ], batch size: 134, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:09:58,428 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-04-28 19:10:09,361 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 19:10:29,421 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-04-28 19:10:33,447 INFO [train.py:904] (2/8) Epoch 8, batch 200, loss[loss=0.2298, simple_loss=0.2933, pruned_loss=0.08313, over 16445.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2913, pruned_loss=0.06951, over 2113083.85 frames. ], batch size: 75, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:10:47,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1096, 4.0602, 4.4854, 3.4626, 4.1185, 4.3906, 4.1293, 2.7366], device='cuda:2'), covar=tensor([0.0276, 0.0036, 0.0020, 0.0184, 0.0041, 0.0031, 0.0032, 0.0275], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0063, 0.0062, 0.0119, 0.0066, 0.0074, 0.0067, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 19:11:40,015 INFO [optim.py:368] (2/8) 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 19:11:42,946 INFO [train.py:904] (2/8) Epoch 8, batch 250, loss[loss=0.2071, simple_loss=0.2836, pruned_loss=0.0653, over 16779.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2896, pruned_loss=0.06953, over 2377185.10 frames. ], batch size: 39, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:11:56,858 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1038, 1.6651, 2.3106, 2.8952, 2.7424, 2.9485, 1.8426, 3.0385], device='cuda:2'), covar=tensor([0.0093, 0.0291, 0.0194, 0.0153, 0.0141, 0.0110, 0.0300, 0.0056], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0155, 0.0138, 0.0138, 0.0144, 0.0098, 0.0152, 0.0090], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 19:12:47,160 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:12:52,125 INFO [train.py:904] (2/8) Epoch 8, batch 300, loss[loss=0.2188, simple_loss=0.2832, pruned_loss=0.07721, over 16933.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2863, pruned_loss=0.06703, over 2596334.86 frames. ], batch size: 116, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:13:38,970 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:13:58,499 INFO [optim.py:368] (2/8) 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 19:14:01,156 INFO [train.py:904] (2/8) Epoch 8, batch 350, loss[loss=0.2057, simple_loss=0.2735, pruned_loss=0.06893, over 16504.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2833, pruned_loss=0.06503, over 2759298.35 frames. ], batch size: 146, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:15:10,453 INFO [train.py:904] (2/8) Epoch 8, batch 400, loss[loss=0.1906, simple_loss=0.2672, pruned_loss=0.05697, over 16473.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2813, pruned_loss=0.06393, over 2890362.27 frames. ], batch size: 68, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:15:19,752 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-28 19:16:17,821 INFO [optim.py:368] (2/8) 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 19:16:20,152 INFO [train.py:904] (2/8) Epoch 8, batch 450, loss[loss=0.1814, simple_loss=0.273, pruned_loss=0.04487, over 17265.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2792, pruned_loss=0.06276, over 2985967.74 frames. ], batch size: 52, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:25,957 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2743, 3.5435, 3.4879, 2.0483, 2.9934, 2.3557, 3.6878, 3.5350], device='cuda:2'), covar=tensor([0.0233, 0.0611, 0.0551, 0.1545, 0.0693, 0.0876, 0.0474, 0.0901], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0131, 0.0152, 0.0138, 0.0131, 0.0122, 0.0133, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 19:17:28,361 INFO [train.py:904] (2/8) Epoch 8, batch 500, loss[loss=0.2005, simple_loss=0.2789, pruned_loss=0.06098, over 16598.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2783, pruned_loss=0.06154, over 3060685.10 frames. ], batch size: 68, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:51,968 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-28 19:18:33,696 INFO [optim.py:368] (2/8) 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 19:18:37,347 INFO [train.py:904] (2/8) Epoch 8, batch 550, loss[loss=0.1892, simple_loss=0.2583, pruned_loss=0.06003, over 16272.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2776, pruned_loss=0.06076, over 3110358.56 frames. ], batch size: 165, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:18:56,794 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:33,217 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 19:19:41,722 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:45,774 INFO [train.py:904] (2/8) Epoch 8, batch 600, loss[loss=0.1762, simple_loss=0.2696, pruned_loss=0.04141, over 17227.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2761, pruned_loss=0.06137, over 3151783.98 frames. ], batch size: 45, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:20:19,069 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:23,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3402, 1.5267, 1.9005, 2.2941, 2.3873, 2.2858, 1.6427, 2.4154], device='cuda:2'), covar=tensor([0.0131, 0.0288, 0.0169, 0.0162, 0.0139, 0.0155, 0.0270, 0.0067], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0157, 0.0141, 0.0141, 0.0148, 0.0103, 0.0155, 0.0093], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 19:20:34,024 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:39,587 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-04-28 19:20:45,949 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:50,937 INFO [optim.py:368] (2/8) 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 19:20:53,309 INFO [train.py:904] (2/8) Epoch 8, batch 650, loss[loss=0.2073, simple_loss=0.2894, pruned_loss=0.06256, over 17022.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2746, pruned_loss=0.06057, over 3186880.79 frames. ], batch size: 50, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:21:37,502 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:22:01,896 INFO [train.py:904] (2/8) Epoch 8, batch 700, loss[loss=0.1962, simple_loss=0.2652, pruned_loss=0.06359, over 16835.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2731, pruned_loss=0.05943, over 3213245.32 frames. ], batch size: 96, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:23:05,954 INFO [optim.py:368] (2/8) 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 19:23:08,640 INFO [train.py:904] (2/8) Epoch 8, batch 750, loss[loss=0.1809, simple_loss=0.2605, pruned_loss=0.05069, over 15920.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2731, pruned_loss=0.05913, over 3239832.71 frames. ], batch size: 35, lr: 8.81e-03, grad_scale: 2.0 2023-04-28 19:24:17,945 INFO [train.py:904] (2/8) Epoch 8, batch 800, loss[loss=0.1695, simple_loss=0.2606, pruned_loss=0.03915, over 17187.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2733, pruned_loss=0.05871, over 3251771.14 frames. ], batch size: 44, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:23,720 INFO [optim.py:368] (2/8) 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 19:25:25,971 INFO [train.py:904] (2/8) Epoch 8, batch 850, loss[loss=0.1865, simple_loss=0.2528, pruned_loss=0.06005, over 16332.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2731, pruned_loss=0.0586, over 3273835.40 frames. ], batch size: 165, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:49,833 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:26:32,698 INFO [train.py:904] (2/8) Epoch 8, batch 900, loss[loss=0.1918, simple_loss=0.2807, pruned_loss=0.05141, over 16688.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2725, pruned_loss=0.05784, over 3281337.70 frames. ], batch size: 57, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:27:00,548 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:13,325 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:37,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7297, 4.6550, 4.5933, 4.3401, 4.1771, 4.6409, 4.5301, 4.3298], device='cuda:2'), covar=tensor([0.0492, 0.0401, 0.0197, 0.0220, 0.0924, 0.0312, 0.0365, 0.0551], device='cuda:2'), in_proj_covar=tensor([0.0223, 0.0262, 0.0261, 0.0234, 0.0294, 0.0265, 0.0178, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 19:27:38,431 INFO [optim.py:368] (2/8) 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 19:27:44,210 INFO [train.py:904] (2/8) Epoch 8, batch 950, loss[loss=0.2053, simple_loss=0.2702, pruned_loss=0.07021, over 16898.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2729, pruned_loss=0.05802, over 3299713.46 frames. ], batch size: 109, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:28:52,162 INFO [train.py:904] (2/8) Epoch 8, batch 1000, loss[loss=0.192, simple_loss=0.2647, pruned_loss=0.05967, over 12080.00 frames. ], tot_loss[loss=0.193, simple_loss=0.271, pruned_loss=0.05752, over 3295877.11 frames. ], batch size: 246, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:29:14,713 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 19:29:58,311 INFO [optim.py:368] (2/8) 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 19:30:01,627 INFO [train.py:904] (2/8) Epoch 8, batch 1050, loss[loss=0.1642, simple_loss=0.2534, pruned_loss=0.03756, over 17174.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.27, pruned_loss=0.05693, over 3296219.78 frames. ], batch size: 46, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:30:40,170 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:31:10,578 INFO [train.py:904] (2/8) Epoch 8, batch 1100, loss[loss=0.1773, simple_loss=0.2431, pruned_loss=0.05572, over 16890.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2691, pruned_loss=0.05625, over 3310532.03 frames. ], batch size: 96, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:31:50,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8587, 4.8863, 5.4631, 5.3972, 5.3731, 5.0221, 4.9415, 4.7377], device='cuda:2'), covar=tensor([0.0253, 0.0420, 0.0313, 0.0400, 0.0381, 0.0273, 0.0784, 0.0390], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0310, 0.0303, 0.0292, 0.0349, 0.0325, 0.0428, 0.0264], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 19:32:03,069 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:32:15,692 INFO [optim.py:368] (2/8) 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 19:32:18,230 INFO [train.py:904] (2/8) Epoch 8, batch 1150, loss[loss=0.1836, simple_loss=0.2539, pruned_loss=0.05662, over 16902.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2687, pruned_loss=0.0555, over 3317019.92 frames. ], batch size: 109, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:36,583 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2710, 5.2955, 5.1108, 4.8056, 4.5790, 5.0860, 5.1393, 4.7274], device='cuda:2'), covar=tensor([0.0548, 0.0317, 0.0229, 0.0253, 0.1073, 0.0354, 0.0261, 0.0591], device='cuda:2'), in_proj_covar=tensor([0.0229, 0.0270, 0.0269, 0.0242, 0.0302, 0.0273, 0.0184, 0.0310], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 19:33:03,893 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2447, 3.4301, 3.2481, 2.0158, 2.7980, 2.3061, 3.7269, 3.5861], device='cuda:2'), covar=tensor([0.0209, 0.0655, 0.0599, 0.1602, 0.0768, 0.0963, 0.0462, 0.0793], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0135, 0.0154, 0.0138, 0.0133, 0.0123, 0.0135, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 19:33:26,638 INFO [train.py:904] (2/8) Epoch 8, batch 1200, loss[loss=0.1861, simple_loss=0.2838, pruned_loss=0.04421, over 17263.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2679, pruned_loss=0.05482, over 3315180.29 frames. ], batch size: 52, lr: 8.79e-03, grad_scale: 8.0 2023-04-28 19:33:45,241 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-28 19:33:53,224 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:33:56,992 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:34:30,638 INFO [optim.py:368] (2/8) 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 19:34:32,975 INFO [train.py:904] (2/8) Epoch 8, batch 1250, loss[loss=0.1912, simple_loss=0.2805, pruned_loss=0.05098, over 16681.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2685, pruned_loss=0.05563, over 3317583.72 frames. ], batch size: 57, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:34:58,450 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:06,090 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:27,424 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4804, 4.1282, 3.9253, 1.9618, 3.1649, 2.4403, 3.7645, 3.8945], device='cuda:2'), covar=tensor([0.0262, 0.0532, 0.0424, 0.1643, 0.0681, 0.0873, 0.0673, 0.0859], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0134, 0.0153, 0.0139, 0.0133, 0.0122, 0.0134, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 19:35:41,550 INFO [train.py:904] (2/8) Epoch 8, batch 1300, loss[loss=0.1678, simple_loss=0.2471, pruned_loss=0.04422, over 16819.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2679, pruned_loss=0.05537, over 3319669.03 frames. ], batch size: 39, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:35:50,799 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-28 19:36:30,367 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:36:49,363 INFO [optim.py:368] (2/8) 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 19:36:52,120 INFO [train.py:904] (2/8) Epoch 8, batch 1350, loss[loss=0.1565, simple_loss=0.2354, pruned_loss=0.03884, over 16894.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2677, pruned_loss=0.0544, over 3324459.17 frames. ], batch size: 41, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:37:29,932 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0578, 5.5030, 5.6094, 5.4421, 5.4044, 6.0523, 5.6502, 5.3493], device='cuda:2'), covar=tensor([0.0760, 0.1666, 0.1609, 0.1996, 0.2630, 0.0944, 0.1207, 0.2171], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0455, 0.0474, 0.0400, 0.0524, 0.0504, 0.0378, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 19:38:01,923 INFO [train.py:904] (2/8) Epoch 8, batch 1400, loss[loss=0.1785, simple_loss=0.2694, pruned_loss=0.04384, over 17041.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2689, pruned_loss=0.05518, over 3332759.00 frames. ], batch size: 55, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:38:47,782 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:39:07,466 INFO [optim.py:368] (2/8) 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 19:39:11,103 INFO [train.py:904] (2/8) Epoch 8, batch 1450, loss[loss=0.1977, simple_loss=0.2816, pruned_loss=0.05691, over 17090.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.268, pruned_loss=0.05515, over 3335978.60 frames. ], batch size: 48, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:39:17,410 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3506, 3.5167, 3.6186, 1.7511, 3.7792, 3.7408, 3.0379, 2.8282], device='cuda:2'), covar=tensor([0.0749, 0.0143, 0.0149, 0.1071, 0.0065, 0.0116, 0.0346, 0.0386], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0094, 0.0083, 0.0137, 0.0069, 0.0090, 0.0119, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 19:40:20,581 INFO [train.py:904] (2/8) Epoch 8, batch 1500, loss[loss=0.185, simple_loss=0.2697, pruned_loss=0.05014, over 16680.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2674, pruned_loss=0.05532, over 3330348.51 frames. ], batch size: 62, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:40:52,085 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:41:07,933 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1505, 5.1518, 4.9084, 4.2745, 4.8724, 1.6263, 4.7052, 4.9265], device='cuda:2'), covar=tensor([0.0063, 0.0050, 0.0132, 0.0364, 0.0077, 0.2269, 0.0113, 0.0145], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0104, 0.0155, 0.0148, 0.0119, 0.0167, 0.0141, 0.0146], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:41:25,446 INFO [optim.py:368] (2/8) 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 19:41:28,317 INFO [train.py:904] (2/8) Epoch 8, batch 1550, loss[loss=0.2133, simple_loss=0.2714, pruned_loss=0.07761, over 16904.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2698, pruned_loss=0.05711, over 3322694.19 frames. ], batch size: 109, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:41:52,372 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 19:41:58,081 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:42:24,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6815, 2.7703, 2.2740, 3.9420, 3.3441, 3.9623, 1.3920, 2.8614], device='cuda:2'), covar=tensor([0.1395, 0.0583, 0.1176, 0.0133, 0.0224, 0.0406, 0.1516, 0.0815], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0154, 0.0174, 0.0117, 0.0196, 0.0207, 0.0173, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 19:42:38,699 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3200, 3.5557, 3.9159, 2.7651, 3.6013, 3.8884, 3.7985, 2.1075], device='cuda:2'), covar=tensor([0.0379, 0.0146, 0.0047, 0.0252, 0.0076, 0.0081, 0.0063, 0.0389], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0065, 0.0062, 0.0117, 0.0067, 0.0076, 0.0069, 0.0114], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 19:42:39,324 INFO [train.py:904] (2/8) Epoch 8, batch 1600, loss[loss=0.233, simple_loss=0.2952, pruned_loss=0.08535, over 16795.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2715, pruned_loss=0.0578, over 3330388.49 frames. ], batch size: 83, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:43:20,418 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:43:21,776 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3763, 4.1592, 4.3495, 4.6085, 4.7126, 4.2588, 4.4547, 4.6456], device='cuda:2'), covar=tensor([0.1209, 0.0978, 0.1439, 0.0618, 0.0584, 0.0998, 0.1278, 0.0722], device='cuda:2'), in_proj_covar=tensor([0.0497, 0.0620, 0.0774, 0.0625, 0.0471, 0.0478, 0.0484, 0.0532], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:43:26,168 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 19:43:44,410 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 8, batch 1650, loss[loss=0.2062, simple_loss=0.2983, pruned_loss=0.05706, over 16665.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2729, pruned_loss=0.05882, over 3325097.39 frames. ], batch size: 57, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:44:06,006 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5230, 3.8658, 4.0023, 1.8891, 4.1494, 4.1297, 3.1820, 2.9438], device='cuda:2'), covar=tensor([0.0731, 0.0110, 0.0118, 0.1033, 0.0060, 0.0106, 0.0322, 0.0385], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0094, 0.0083, 0.0137, 0.0070, 0.0091, 0.0119, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 19:44:58,249 INFO [train.py:904] (2/8) Epoch 8, batch 1700, loss[loss=0.2185, simple_loss=0.3109, pruned_loss=0.06307, over 17027.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2743, pruned_loss=0.05957, over 3328504.64 frames. ], batch size: 50, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:45:06,945 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7870, 5.1580, 4.8846, 4.9469, 4.6252, 4.5483, 4.6240, 5.1816], device='cuda:2'), covar=tensor([0.0853, 0.0710, 0.1001, 0.0563, 0.0667, 0.0856, 0.0787, 0.0792], device='cuda:2'), in_proj_covar=tensor([0.0482, 0.0616, 0.0509, 0.0411, 0.0386, 0.0400, 0.0512, 0.0454], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:45:44,214 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 19:46:07,546 INFO [train.py:904] (2/8) Epoch 8, batch 1750, loss[loss=0.2165, simple_loss=0.2862, pruned_loss=0.07335, over 16886.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2756, pruned_loss=0.05917, over 3332440.11 frames. ], batch size: 116, lr: 8.75e-03, grad_scale: 8.0 2023-04-28 19:46:50,155 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:46:57,481 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7867, 3.8306, 4.2311, 4.1911, 4.1893, 3.8965, 3.9310, 3.8527], device='cuda:2'), covar=tensor([0.0376, 0.0471, 0.0326, 0.0420, 0.0406, 0.0353, 0.0737, 0.0493], device='cuda:2'), in_proj_covar=tensor([0.0303, 0.0311, 0.0310, 0.0295, 0.0351, 0.0329, 0.0432, 0.0266], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 19:47:02,643 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2298, 1.9112, 2.5707, 3.2080, 2.9015, 3.5568, 2.4834, 3.4670], device='cuda:2'), covar=tensor([0.0100, 0.0271, 0.0177, 0.0127, 0.0138, 0.0083, 0.0252, 0.0075], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0158, 0.0145, 0.0144, 0.0151, 0.0105, 0.0157, 0.0098], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 19:47:16,118 INFO [train.py:904] (2/8) Epoch 8, batch 1800, loss[loss=0.2233, simple_loss=0.2939, pruned_loss=0.07631, over 16790.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2763, pruned_loss=0.05874, over 3333653.29 frames. ], batch size: 124, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:48:15,291 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:48:21,293 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1640, 4.3470, 2.7085, 4.7950, 3.1711, 4.7254, 2.6725, 3.5008], device='cuda:2'), covar=tensor([0.0143, 0.0253, 0.1120, 0.0093, 0.0663, 0.0335, 0.1209, 0.0491], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0163, 0.0180, 0.0107, 0.0164, 0.0201, 0.0191, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 19:48:24,987 INFO [optim.py:368] (2/8) 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 19:48:26,868 INFO [train.py:904] (2/8) Epoch 8, batch 1850, loss[loss=0.2092, simple_loss=0.2855, pruned_loss=0.06647, over 16243.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2767, pruned_loss=0.05889, over 3339379.19 frames. ], batch size: 165, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:49:35,219 INFO [train.py:904] (2/8) Epoch 8, batch 1900, loss[loss=0.2477, simple_loss=0.3233, pruned_loss=0.08605, over 11989.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2771, pruned_loss=0.05865, over 3327389.87 frames. ], batch size: 246, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:49:39,054 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:49:51,989 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 19:50:11,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9318, 2.5026, 2.2912, 2.8481, 2.5167, 3.2912, 1.6818, 2.7452], device='cuda:2'), covar=tensor([0.1044, 0.0487, 0.0930, 0.0113, 0.0165, 0.0362, 0.1204, 0.0607], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0154, 0.0175, 0.0117, 0.0197, 0.0207, 0.0172, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 19:50:16,631 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:50:41,095 INFO [optim.py:368] (2/8) 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 19:50:43,027 INFO [train.py:904] (2/8) Epoch 8, batch 1950, loss[loss=0.1893, simple_loss=0.2865, pruned_loss=0.04599, over 17128.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2771, pruned_loss=0.05813, over 3324172.73 frames. ], batch size: 47, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:51:14,236 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 19:51:21,209 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:51:29,516 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-04-28 19:51:46,382 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9787, 4.8698, 4.7345, 4.1949, 4.8019, 1.7986, 4.5753, 4.7339], device='cuda:2'), covar=tensor([0.0065, 0.0059, 0.0155, 0.0346, 0.0073, 0.2198, 0.0114, 0.0154], device='cuda:2'), in_proj_covar=tensor([0.0116, 0.0104, 0.0157, 0.0151, 0.0121, 0.0168, 0.0142, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:51:50,195 INFO [train.py:904] (2/8) Epoch 8, batch 2000, loss[loss=0.1989, simple_loss=0.2823, pruned_loss=0.05773, over 16526.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2775, pruned_loss=0.05813, over 3323041.04 frames. ], batch size: 68, lr: 8.74e-03, grad_scale: 8.0 2023-04-28 19:52:58,781 INFO [optim.py:368] (2/8) 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 19:53:00,013 INFO [train.py:904] (2/8) Epoch 8, batch 2050, loss[loss=0.2012, simple_loss=0.2849, pruned_loss=0.05876, over 16574.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2772, pruned_loss=0.0583, over 3323058.79 frames. ], batch size: 68, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:53:19,030 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 19:53:36,702 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-28 19:54:09,330 INFO [train.py:904] (2/8) Epoch 8, batch 2100, loss[loss=0.2338, simple_loss=0.3075, pruned_loss=0.08005, over 11675.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2781, pruned_loss=0.05903, over 3322332.37 frames. ], batch size: 247, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:55:14,762 INFO [optim.py:368] (2/8) 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 19:55:16,495 INFO [train.py:904] (2/8) Epoch 8, batch 2150, loss[loss=0.2185, simple_loss=0.306, pruned_loss=0.06547, over 16705.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2792, pruned_loss=0.05951, over 3323824.17 frames. ], batch size: 57, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:56:09,859 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 19:56:21,782 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:56:23,750 INFO [train.py:904] (2/8) Epoch 8, batch 2200, loss[loss=0.1976, simple_loss=0.277, pruned_loss=0.0591, over 16505.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2803, pruned_loss=0.06009, over 3321697.90 frames. ], batch size: 68, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:56:24,315 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6574, 3.1636, 2.7597, 5.0968, 4.3893, 4.7898, 1.4920, 3.3570], device='cuda:2'), covar=tensor([0.1428, 0.0595, 0.1078, 0.0085, 0.0252, 0.0298, 0.1478, 0.0659], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0154, 0.0174, 0.0118, 0.0199, 0.0208, 0.0173, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 19:57:20,268 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:57:26,695 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5730, 4.3334, 4.5356, 4.7728, 4.8699, 4.3823, 4.6904, 4.7905], device='cuda:2'), covar=tensor([0.1235, 0.1030, 0.1243, 0.0655, 0.0584, 0.0976, 0.1142, 0.0888], device='cuda:2'), in_proj_covar=tensor([0.0493, 0.0617, 0.0767, 0.0618, 0.0468, 0.0472, 0.0483, 0.0529], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 19:57:29,785 INFO [optim.py:368] (2/8) 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 19:57:31,975 INFO [train.py:904] (2/8) Epoch 8, batch 2250, loss[loss=0.1979, simple_loss=0.2689, pruned_loss=0.06342, over 16752.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2809, pruned_loss=0.06038, over 3325741.56 frames. ], batch size: 124, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:57:56,979 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 19:58:40,764 INFO [train.py:904] (2/8) Epoch 8, batch 2300, loss[loss=0.2013, simple_loss=0.2898, pruned_loss=0.05643, over 16683.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2809, pruned_loss=0.06048, over 3329868.51 frames. ], batch size: 57, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:58:41,813 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 19:58:44,125 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:59:37,568 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9317, 3.3244, 2.8786, 5.1134, 4.3723, 4.7966, 1.6453, 3.3726], device='cuda:2'), covar=tensor([0.1285, 0.0579, 0.0959, 0.0124, 0.0239, 0.0294, 0.1337, 0.0662], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0153, 0.0172, 0.0117, 0.0198, 0.0207, 0.0170, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 19:59:48,860 INFO [optim.py:368] (2/8) 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 19:59:49,988 INFO [train.py:904] (2/8) Epoch 8, batch 2350, loss[loss=0.1666, simple_loss=0.2512, pruned_loss=0.04093, over 17190.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2814, pruned_loss=0.06098, over 3323782.95 frames. ], batch size: 43, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 20:00:50,339 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5940, 3.1762, 2.7215, 4.9038, 4.0146, 4.5794, 1.4468, 3.2176], device='cuda:2'), covar=tensor([0.1468, 0.0614, 0.1089, 0.0118, 0.0305, 0.0322, 0.1606, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0151, 0.0171, 0.0116, 0.0196, 0.0206, 0.0170, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 20:00:58,115 INFO [train.py:904] (2/8) Epoch 8, batch 2400, loss[loss=0.1789, simple_loss=0.2662, pruned_loss=0.04584, over 17206.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2829, pruned_loss=0.06099, over 3330355.33 frames. ], batch size: 44, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:02:04,601 INFO [optim.py:368] (2/8) 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 20:02:05,796 INFO [train.py:904] (2/8) Epoch 8, batch 2450, loss[loss=0.2278, simple_loss=0.3033, pruned_loss=0.0762, over 12620.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2829, pruned_loss=0.06012, over 3331446.50 frames. ], batch size: 246, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:02:13,813 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6242, 2.7053, 2.5973, 4.1143, 3.6205, 4.0420, 1.2136, 3.2041], device='cuda:2'), covar=tensor([0.1280, 0.0566, 0.0912, 0.0106, 0.0172, 0.0292, 0.1431, 0.0552], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0153, 0.0172, 0.0118, 0.0198, 0.0207, 0.0171, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 20:03:12,523 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:03:14,440 INFO [train.py:904] (2/8) Epoch 8, batch 2500, loss[loss=0.1863, simple_loss=0.2636, pruned_loss=0.05448, over 16680.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2825, pruned_loss=0.06044, over 3322622.08 frames. ], batch size: 89, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:03:37,630 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6206, 4.1864, 3.8850, 2.2793, 3.3664, 2.5761, 3.9313, 4.0346], device='cuda:2'), covar=tensor([0.0275, 0.0612, 0.0483, 0.1490, 0.0644, 0.0913, 0.0621, 0.0949], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0138, 0.0154, 0.0139, 0.0133, 0.0123, 0.0136, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 20:04:07,502 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:18,254 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:22,383 INFO [optim.py:368] (2/8) 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 20:04:23,439 INFO [train.py:904] (2/8) Epoch 8, batch 2550, loss[loss=0.2684, simple_loss=0.3301, pruned_loss=0.1034, over 15535.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.283, pruned_loss=0.06083, over 3319299.48 frames. ], batch size: 190, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:04:48,964 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:05:29,131 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,235 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,975 INFO [train.py:904] (2/8) Epoch 8, batch 2600, loss[loss=0.198, simple_loss=0.2783, pruned_loss=0.05881, over 16845.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2823, pruned_loss=0.06008, over 3309773.05 frames. ], batch size: 116, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:05:35,469 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9780, 2.1663, 2.2159, 4.7013, 1.9786, 2.8401, 2.3094, 2.4900], device='cuda:2'), covar=tensor([0.0687, 0.3036, 0.1883, 0.0307, 0.3624, 0.1870, 0.2592, 0.3067], device='cuda:2'), in_proj_covar=tensor([0.0350, 0.0365, 0.0307, 0.0325, 0.0392, 0.0407, 0.0328, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:05:55,318 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 20:06:42,999 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 8, batch 2650, loss[loss=0.2016, simple_loss=0.2944, pruned_loss=0.05435, over 16685.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2824, pruned_loss=0.05966, over 3314412.14 frames. ], batch size: 57, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:23,279 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:07:50,956 INFO [train.py:904] (2/8) Epoch 8, batch 2700, loss[loss=0.1865, simple_loss=0.2672, pruned_loss=0.05292, over 15461.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2824, pruned_loss=0.05895, over 3315528.10 frames. ], batch size: 190, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:57,604 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2390, 3.5569, 3.3105, 2.0129, 2.9579, 2.3232, 3.7140, 3.6176], device='cuda:2'), covar=tensor([0.0213, 0.0601, 0.0585, 0.1562, 0.0724, 0.0958, 0.0548, 0.0828], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0139, 0.0155, 0.0140, 0.0134, 0.0123, 0.0135, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 20:08:46,264 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:08:57,112 INFO [optim.py:368] (2/8) 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 20:08:57,128 INFO [train.py:904] (2/8) Epoch 8, batch 2750, loss[loss=0.1768, simple_loss=0.2694, pruned_loss=0.04205, over 16847.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2826, pruned_loss=0.05821, over 3315621.64 frames. ], batch size: 42, lr: 8.69e-03, grad_scale: 4.0 2023-04-28 20:09:09,936 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 20:09:30,111 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2317, 5.2222, 4.9945, 4.4964, 5.0550, 1.9788, 4.7688, 5.0771], device='cuda:2'), covar=tensor([0.0059, 0.0047, 0.0134, 0.0318, 0.0066, 0.1994, 0.0105, 0.0132], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0105, 0.0156, 0.0150, 0.0122, 0.0167, 0.0143, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:10:05,129 INFO [train.py:904] (2/8) Epoch 8, batch 2800, loss[loss=0.1938, simple_loss=0.283, pruned_loss=0.05229, over 16188.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2821, pruned_loss=0.05762, over 3316021.80 frames. ], batch size: 35, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:11:14,609 INFO [optim.py:368] (2/8) 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 20:11:14,626 INFO [train.py:904] (2/8) Epoch 8, batch 2850, loss[loss=0.2036, simple_loss=0.283, pruned_loss=0.06204, over 16524.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2822, pruned_loss=0.05835, over 3307448.18 frames. ], batch size: 68, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:11:51,018 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0620, 4.3700, 2.0949, 4.6416, 3.0533, 4.6058, 2.5292, 3.3148], device='cuda:2'), covar=tensor([0.0174, 0.0266, 0.1554, 0.0115, 0.0750, 0.0349, 0.1314, 0.0555], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0165, 0.0182, 0.0109, 0.0166, 0.0204, 0.0192, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 20:12:16,250 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:20,708 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:23,979 INFO [train.py:904] (2/8) Epoch 8, batch 2900, loss[loss=0.1878, simple_loss=0.2571, pruned_loss=0.05927, over 16706.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2795, pruned_loss=0.05833, over 3315741.96 frames. ], batch size: 89, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:12:59,051 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 20:13:28,886 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:13:38,231 INFO [optim.py:368] (2/8) 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 20:13:38,246 INFO [train.py:904] (2/8) Epoch 8, batch 2950, loss[loss=0.1547, simple_loss=0.2397, pruned_loss=0.03482, over 16826.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2788, pruned_loss=0.05912, over 3320646.32 frames. ], batch size: 42, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:13:38,529 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6282, 5.9958, 5.6746, 5.7545, 5.3388, 5.1708, 5.4427, 6.1242], device='cuda:2'), covar=tensor([0.1041, 0.0698, 0.1020, 0.0600, 0.0654, 0.0603, 0.0826, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0493, 0.0630, 0.0519, 0.0422, 0.0395, 0.0406, 0.0519, 0.0466], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:14:44,903 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 20:14:46,251 INFO [train.py:904] (2/8) Epoch 8, batch 3000, loss[loss=0.2173, simple_loss=0.2895, pruned_loss=0.07254, over 16896.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2784, pruned_loss=0.05922, over 3323269.97 frames. ], batch size: 96, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:14:46,252 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 20:14:55,848 INFO [train.py:938] (2/8) Epoch 8, validation: loss=0.1462, simple_loss=0.2525, pruned_loss=0.01995, over 944034.00 frames. 2023-04-28 20:14:55,849 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-28 20:15:13,433 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:15:44,979 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6840, 2.6887, 2.4200, 3.8078, 3.3257, 3.8790, 1.4420, 2.8000], device='cuda:2'), covar=tensor([0.1249, 0.0547, 0.0962, 0.0147, 0.0244, 0.0332, 0.1370, 0.0702], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0153, 0.0174, 0.0121, 0.0201, 0.0208, 0.0172, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 20:15:45,909 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:16:06,763 INFO [optim.py:368] (2/8) 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 20:16:06,778 INFO [train.py:904] (2/8) Epoch 8, batch 3050, loss[loss=0.1845, simple_loss=0.2775, pruned_loss=0.04573, over 17160.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2788, pruned_loss=0.05959, over 3321633.67 frames. ], batch size: 47, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:16:30,172 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2420, 4.0628, 4.2335, 4.4092, 4.5185, 4.1424, 4.2494, 4.4772], device='cuda:2'), covar=tensor([0.1262, 0.0905, 0.1328, 0.0581, 0.0549, 0.0998, 0.1795, 0.0565], device='cuda:2'), in_proj_covar=tensor([0.0503, 0.0619, 0.0772, 0.0628, 0.0474, 0.0475, 0.0490, 0.0538], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:16:38,641 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:16:41,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3231, 5.2129, 5.1058, 4.7428, 4.7028, 5.1285, 5.2022, 4.8270], device='cuda:2'), covar=tensor([0.0494, 0.0355, 0.0218, 0.0252, 0.0976, 0.0379, 0.0226, 0.0554], device='cuda:2'), in_proj_covar=tensor([0.0239, 0.0283, 0.0278, 0.0253, 0.0315, 0.0283, 0.0192, 0.0322], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 20:17:13,362 INFO [train.py:904] (2/8) Epoch 8, batch 3100, loss[loss=0.1573, simple_loss=0.2402, pruned_loss=0.03719, over 16747.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2778, pruned_loss=0.05903, over 3320887.34 frames. ], batch size: 39, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:17:29,376 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3818, 4.1979, 4.3757, 4.5768, 4.6470, 4.2396, 4.3082, 4.6017], device='cuda:2'), covar=tensor([0.1090, 0.0827, 0.1186, 0.0531, 0.0524, 0.1004, 0.1833, 0.0594], device='cuda:2'), in_proj_covar=tensor([0.0500, 0.0614, 0.0767, 0.0623, 0.0470, 0.0471, 0.0487, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:17:39,313 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 20:18:21,290 INFO [optim.py:368] (2/8) 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 20:18:21,306 INFO [train.py:904] (2/8) Epoch 8, batch 3150, loss[loss=0.173, simple_loss=0.2657, pruned_loss=0.04013, over 17126.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2774, pruned_loss=0.05966, over 3318430.88 frames. ], batch size: 49, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:18:55,787 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 20:19:14,378 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:23,308 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:32,368 INFO [train.py:904] (2/8) Epoch 8, batch 3200, loss[loss=0.2064, simple_loss=0.2781, pruned_loss=0.06728, over 16324.00 frames. ], tot_loss[loss=0.197, simple_loss=0.277, pruned_loss=0.05851, over 3325909.58 frames. ], batch size: 165, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:19:53,473 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8309, 5.0875, 5.3379, 5.1169, 5.0843, 5.7251, 5.3284, 4.9538], device='cuda:2'), covar=tensor([0.1067, 0.1742, 0.1791, 0.1797, 0.2727, 0.0986, 0.1277, 0.2441], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0473, 0.0486, 0.0412, 0.0544, 0.0515, 0.0390, 0.0555], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 20:20:30,673 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:40,565 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:41,302 INFO [train.py:904] (2/8) Epoch 8, batch 3250, loss[loss=0.1769, simple_loss=0.2598, pruned_loss=0.04698, over 16866.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2762, pruned_loss=0.0582, over 3320932.35 frames. ], batch size: 96, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:20:42,358 INFO [optim.py:368] (2/8) 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 20:21:52,435 INFO [train.py:904] (2/8) Epoch 8, batch 3300, loss[loss=0.1778, simple_loss=0.2653, pruned_loss=0.0451, over 17216.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2779, pruned_loss=0.05897, over 3330835.71 frames. ], batch size: 46, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:22:41,288 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:01,500 INFO [train.py:904] (2/8) Epoch 8, batch 3350, loss[loss=0.2297, simple_loss=0.3014, pruned_loss=0.07901, over 16421.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2779, pruned_loss=0.0585, over 3338796.21 frames. ], batch size: 146, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:23:02,765 INFO [optim.py:368] (2/8) 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 20:23:18,320 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2958, 4.6164, 4.4217, 4.4641, 4.1411, 4.0386, 4.1011, 4.6573], device='cuda:2'), covar=tensor([0.1001, 0.0723, 0.0895, 0.0547, 0.0645, 0.1267, 0.0797, 0.0753], device='cuda:2'), in_proj_covar=tensor([0.0491, 0.0630, 0.0518, 0.0422, 0.0393, 0.0409, 0.0521, 0.0466], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:23:27,943 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:33,874 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9512, 3.3588, 2.4323, 4.6047, 3.9631, 4.2569, 1.6741, 2.9591], device='cuda:2'), covar=tensor([0.1019, 0.0407, 0.0959, 0.0104, 0.0242, 0.0336, 0.1098, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0153, 0.0172, 0.0121, 0.0202, 0.0206, 0.0171, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 20:23:49,807 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:24:11,179 INFO [train.py:904] (2/8) Epoch 8, batch 3400, loss[loss=0.2286, simple_loss=0.3091, pruned_loss=0.07407, over 16230.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2776, pruned_loss=0.05799, over 3319884.96 frames. ], batch size: 165, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:24:49,962 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 20:25:08,049 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 20:25:21,723 INFO [train.py:904] (2/8) Epoch 8, batch 3450, loss[loss=0.1635, simple_loss=0.2506, pruned_loss=0.03818, over 16990.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2766, pruned_loss=0.05693, over 3331663.14 frames. ], batch size: 41, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:25:22,839 INFO [optim.py:368] (2/8) 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 20:25:27,070 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-28 20:26:30,766 INFO [train.py:904] (2/8) Epoch 8, batch 3500, loss[loss=0.205, simple_loss=0.292, pruned_loss=0.05901, over 17077.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2757, pruned_loss=0.05697, over 3330208.76 frames. ], batch size: 53, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:05,471 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7201, 3.1540, 2.9733, 1.8531, 2.6363, 2.1803, 3.2718, 3.2547], device='cuda:2'), covar=tensor([0.0246, 0.0672, 0.0519, 0.1582, 0.0719, 0.0860, 0.0550, 0.0721], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0142, 0.0154, 0.0139, 0.0134, 0.0123, 0.0135, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 20:27:12,593 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9786, 4.6627, 4.8829, 5.1565, 5.3609, 4.6627, 5.3087, 5.2895], device='cuda:2'), covar=tensor([0.1326, 0.1055, 0.1816, 0.0656, 0.0473, 0.0822, 0.0443, 0.0433], device='cuda:2'), in_proj_covar=tensor([0.0508, 0.0627, 0.0785, 0.0638, 0.0481, 0.0483, 0.0496, 0.0548], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:27:32,367 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:33,599 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:42,853 INFO [train.py:904] (2/8) Epoch 8, batch 3550, loss[loss=0.2158, simple_loss=0.2767, pruned_loss=0.07743, over 16903.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2746, pruned_loss=0.05618, over 3322261.56 frames. ], batch size: 116, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:43,958 INFO [optim.py:368] (2/8) 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 20:27:52,307 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9520, 4.2371, 2.3520, 4.6342, 3.0184, 4.6877, 2.5448, 3.3905], device='cuda:2'), covar=tensor([0.0183, 0.0241, 0.1379, 0.0135, 0.0746, 0.0360, 0.1256, 0.0549], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0164, 0.0179, 0.0111, 0.0165, 0.0206, 0.0191, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 20:28:51,888 INFO [train.py:904] (2/8) Epoch 8, batch 3600, loss[loss=0.2355, simple_loss=0.2955, pruned_loss=0.08774, over 11492.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2734, pruned_loss=0.05593, over 3318054.63 frames. ], batch size: 247, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:28:56,996 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74656.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:29:35,213 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8625, 4.8369, 5.3748, 5.3781, 5.3632, 5.0202, 4.9725, 4.8057], device='cuda:2'), covar=tensor([0.0280, 0.0437, 0.0352, 0.0341, 0.0369, 0.0268, 0.0776, 0.0337], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0314, 0.0313, 0.0303, 0.0358, 0.0332, 0.0440, 0.0269], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 20:30:00,903 INFO [train.py:904] (2/8) Epoch 8, batch 3650, loss[loss=0.1995, simple_loss=0.2603, pruned_loss=0.06935, over 16829.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2726, pruned_loss=0.05659, over 3304081.95 frames. ], batch size: 116, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:30:02,111 INFO [optim.py:368] (2/8) 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 20:30:27,858 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:31:13,893 INFO [train.py:904] (2/8) Epoch 8, batch 3700, loss[loss=0.168, simple_loss=0.245, pruned_loss=0.04548, over 16822.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2714, pruned_loss=0.05846, over 3274440.10 frames. ], batch size: 102, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:31:38,734 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:32:29,721 INFO [train.py:904] (2/8) Epoch 8, batch 3750, loss[loss=0.2307, simple_loss=0.3255, pruned_loss=0.06794, over 16988.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2726, pruned_loss=0.06027, over 3263718.23 frames. ], batch size: 41, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:32:30,687 INFO [optim.py:368] (2/8) 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 20:32:39,537 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5274, 3.7016, 3.7957, 1.8094, 3.9556, 3.9006, 3.1505, 2.8866], device='cuda:2'), covar=tensor([0.0712, 0.0128, 0.0130, 0.1158, 0.0055, 0.0119, 0.0314, 0.0409], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0096, 0.0085, 0.0137, 0.0070, 0.0093, 0.0119, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 20:32:52,316 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3020, 5.0690, 5.3043, 5.5243, 5.6696, 4.8804, 5.5600, 5.6493], device='cuda:2'), covar=tensor([0.1135, 0.0801, 0.1147, 0.0426, 0.0358, 0.0559, 0.0408, 0.0346], device='cuda:2'), in_proj_covar=tensor([0.0493, 0.0608, 0.0755, 0.0619, 0.0466, 0.0466, 0.0480, 0.0532], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:33:08,404 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9886, 1.6456, 2.2833, 2.9842, 2.7917, 3.1200, 1.8183, 3.0727], device='cuda:2'), covar=tensor([0.0115, 0.0347, 0.0220, 0.0154, 0.0172, 0.0113, 0.0326, 0.0062], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0160, 0.0146, 0.0149, 0.0155, 0.0111, 0.0160, 0.0101], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 20:33:41,192 INFO [train.py:904] (2/8) Epoch 8, batch 3800, loss[loss=0.2579, simple_loss=0.3251, pruned_loss=0.09534, over 12536.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2744, pruned_loss=0.06208, over 3255818.11 frames. ], batch size: 246, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:22,601 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 20:34:45,344 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:34:51,947 INFO [train.py:904] (2/8) Epoch 8, batch 3850, loss[loss=0.18, simple_loss=0.2553, pruned_loss=0.0524, over 16845.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2735, pruned_loss=0.06214, over 3263489.76 frames. ], batch size: 102, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:53,139 INFO [optim.py:368] (2/8) 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 20:35:03,411 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:35:52,590 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:01,444 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:02,285 INFO [train.py:904] (2/8) Epoch 8, batch 3900, loss[loss=0.189, simple_loss=0.2611, pruned_loss=0.05844, over 16417.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2726, pruned_loss=0.062, over 3270082.64 frames. ], batch size: 146, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:36:20,335 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:25,190 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6596, 3.8243, 4.0029, 2.0034, 4.1662, 4.1161, 3.2585, 3.0563], device='cuda:2'), covar=tensor([0.0668, 0.0143, 0.0128, 0.1036, 0.0053, 0.0117, 0.0281, 0.0376], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0098, 0.0085, 0.0138, 0.0070, 0.0093, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 20:36:27,577 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:27,594 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9609, 2.4323, 2.2365, 2.8982, 2.6200, 3.2897, 1.6154, 2.6576], device='cuda:2'), covar=tensor([0.1047, 0.0537, 0.0982, 0.0121, 0.0193, 0.0355, 0.1234, 0.0675], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0153, 0.0173, 0.0121, 0.0203, 0.0209, 0.0172, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 20:37:12,342 INFO [train.py:904] (2/8) Epoch 8, batch 3950, loss[loss=0.204, simple_loss=0.2698, pruned_loss=0.06915, over 16934.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2727, pruned_loss=0.06288, over 3268580.39 frames. ], batch size: 90, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:37:14,091 INFO [optim.py:368] (2/8) 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 20:37:18,580 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 20:37:37,118 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 20:37:46,130 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:38:23,629 INFO [train.py:904] (2/8) Epoch 8, batch 4000, loss[loss=0.1792, simple_loss=0.2597, pruned_loss=0.04938, over 16687.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2724, pruned_loss=0.06301, over 3268911.89 frames. ], batch size: 62, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:38:30,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3661, 2.0052, 2.2492, 3.9591, 2.0037, 2.5428, 2.1530, 2.1710], device='cuda:2'), covar=tensor([0.0805, 0.2992, 0.1742, 0.0412, 0.3157, 0.1979, 0.2707, 0.2707], device='cuda:2'), in_proj_covar=tensor([0.0355, 0.0375, 0.0313, 0.0331, 0.0401, 0.0424, 0.0338, 0.0445], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:39:36,988 INFO [train.py:904] (2/8) Epoch 8, batch 4050, loss[loss=0.1721, simple_loss=0.2539, pruned_loss=0.04509, over 17280.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2728, pruned_loss=0.0621, over 3261526.03 frames. ], batch size: 52, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:39:38,164 INFO [optim.py:368] (2/8) 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 20:40:23,302 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-28 20:40:49,054 INFO [train.py:904] (2/8) Epoch 8, batch 4100, loss[loss=0.2095, simple_loss=0.295, pruned_loss=0.06204, over 16760.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2742, pruned_loss=0.06138, over 3255500.74 frames. ], batch size: 124, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:41:07,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5882, 4.9357, 4.6745, 4.7208, 4.3944, 4.3173, 4.4372, 4.9739], device='cuda:2'), covar=tensor([0.1018, 0.0706, 0.1048, 0.0519, 0.0778, 0.0981, 0.0865, 0.0749], device='cuda:2'), in_proj_covar=tensor([0.0491, 0.0620, 0.0521, 0.0423, 0.0395, 0.0408, 0.0518, 0.0460], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:41:26,767 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3442, 1.9973, 2.0946, 3.9997, 1.9064, 2.5811, 2.1787, 2.2657], device='cuda:2'), covar=tensor([0.0786, 0.2778, 0.1744, 0.0329, 0.3261, 0.1761, 0.2443, 0.2697], device='cuda:2'), in_proj_covar=tensor([0.0355, 0.0375, 0.0312, 0.0329, 0.0399, 0.0424, 0.0337, 0.0444], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:41:44,491 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-28 20:42:02,416 INFO [train.py:904] (2/8) Epoch 8, batch 4150, loss[loss=0.1861, simple_loss=0.2782, pruned_loss=0.04698, over 16868.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.282, pruned_loss=0.06432, over 3226295.57 frames. ], batch size: 96, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:42:04,251 INFO [optim.py:368] (2/8) 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 20:42:43,290 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:07,275 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5863, 4.3592, 4.5624, 4.7381, 4.8531, 4.3672, 4.8458, 4.8402], device='cuda:2'), covar=tensor([0.1028, 0.0874, 0.1219, 0.0504, 0.0398, 0.0805, 0.0402, 0.0418], device='cuda:2'), in_proj_covar=tensor([0.0473, 0.0579, 0.0720, 0.0587, 0.0444, 0.0447, 0.0452, 0.0504], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:43:19,320 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:20,232 INFO [train.py:904] (2/8) Epoch 8, batch 4200, loss[loss=0.2257, simple_loss=0.3163, pruned_loss=0.06752, over 16745.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2891, pruned_loss=0.06595, over 3209600.79 frames. ], batch size: 134, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:43:40,419 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:47,330 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:54,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9833, 4.7621, 4.9387, 5.1612, 5.3160, 4.6477, 5.2523, 5.2913], device='cuda:2'), covar=tensor([0.1056, 0.0937, 0.1205, 0.0482, 0.0335, 0.0631, 0.0499, 0.0366], device='cuda:2'), in_proj_covar=tensor([0.0468, 0.0572, 0.0712, 0.0581, 0.0439, 0.0443, 0.0446, 0.0499], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:44:16,427 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:24,282 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 20:44:30,194 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:34,932 INFO [train.py:904] (2/8) Epoch 8, batch 4250, loss[loss=0.1748, simple_loss=0.2715, pruned_loss=0.03901, over 16916.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2912, pruned_loss=0.0652, over 3218361.25 frames. ], batch size: 96, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:44:36,196 INFO [optim.py:368] (2/8) 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 20:45:01,415 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:45:18,389 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 20:45:32,437 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4575, 2.8005, 2.5659, 4.2740, 3.3748, 4.0695, 1.5145, 3.0935], device='cuda:2'), covar=tensor([0.1509, 0.0723, 0.1146, 0.0138, 0.0363, 0.0381, 0.1559, 0.0786], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0151, 0.0170, 0.0118, 0.0201, 0.0203, 0.0170, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 20:45:48,478 INFO [train.py:904] (2/8) Epoch 8, batch 4300, loss[loss=0.2224, simple_loss=0.3087, pruned_loss=0.06811, over 16845.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2927, pruned_loss=0.06459, over 3219588.09 frames. ], batch size: 116, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:45:50,273 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 20:47:02,655 INFO [train.py:904] (2/8) Epoch 8, batch 4350, loss[loss=0.2245, simple_loss=0.3081, pruned_loss=0.07043, over 15444.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2966, pruned_loss=0.06607, over 3208031.26 frames. ], batch size: 191, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:47:03,853 INFO [optim.py:368] (2/8) 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 20:48:04,386 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4308, 3.5218, 1.7725, 3.9048, 2.5415, 3.8239, 2.0321, 2.6842], device='cuda:2'), covar=tensor([0.0169, 0.0261, 0.1666, 0.0056, 0.0766, 0.0353, 0.1435, 0.0715], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0161, 0.0181, 0.0103, 0.0164, 0.0200, 0.0189, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 20:48:17,344 INFO [train.py:904] (2/8) Epoch 8, batch 4400, loss[loss=0.2284, simple_loss=0.3118, pruned_loss=0.0725, over 16662.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2989, pruned_loss=0.06725, over 3222452.80 frames. ], batch size: 134, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:49:26,864 INFO [train.py:904] (2/8) Epoch 8, batch 4450, loss[loss=0.1978, simple_loss=0.2894, pruned_loss=0.0531, over 16671.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3022, pruned_loss=0.06776, over 3239649.32 frames. ], batch size: 89, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:49:28,916 INFO [optim.py:368] (2/8) 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 20:50:03,819 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2809, 5.2588, 5.0518, 4.8640, 4.6255, 5.1302, 5.0289, 4.7478], device='cuda:2'), covar=tensor([0.0413, 0.0180, 0.0194, 0.0169, 0.0793, 0.0197, 0.0192, 0.0439], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0251, 0.0253, 0.0225, 0.0280, 0.0254, 0.0172, 0.0286], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:50:14,727 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1715, 5.1446, 4.9757, 4.8205, 4.5891, 5.0483, 4.8952, 4.6817], device='cuda:2'), covar=tensor([0.0383, 0.0168, 0.0179, 0.0150, 0.0716, 0.0193, 0.0210, 0.0425], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0251, 0.0253, 0.0225, 0.0280, 0.0254, 0.0172, 0.0286], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:50:38,184 INFO [train.py:904] (2/8) Epoch 8, batch 4500, loss[loss=0.2314, simple_loss=0.314, pruned_loss=0.07438, over 16270.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3017, pruned_loss=0.06766, over 3241360.02 frames. ], batch size: 165, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:50:57,725 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:24,868 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:40,650 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4670, 3.5605, 1.8064, 3.9210, 2.5999, 3.8356, 2.0177, 2.7167], device='cuda:2'), covar=tensor([0.0173, 0.0266, 0.1624, 0.0072, 0.0752, 0.0356, 0.1427, 0.0672], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0160, 0.0181, 0.0102, 0.0163, 0.0199, 0.0189, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 20:51:51,099 INFO [train.py:904] (2/8) Epoch 8, batch 4550, loss[loss=0.2286, simple_loss=0.3192, pruned_loss=0.06899, over 16794.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3026, pruned_loss=0.06863, over 3245959.15 frames. ], batch size: 83, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:51:52,275 INFO [optim.py:368] (2/8) 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 20:52:06,459 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:07,791 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7986, 5.0529, 5.3194, 5.1432, 5.1225, 5.7339, 5.2014, 4.8955], device='cuda:2'), covar=tensor([0.0759, 0.1468, 0.1519, 0.1358, 0.2178, 0.0733, 0.1085, 0.1915], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0438, 0.0448, 0.0380, 0.0505, 0.0480, 0.0367, 0.0518], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 20:52:17,120 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:24,430 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 20:53:02,628 INFO [train.py:904] (2/8) Epoch 8, batch 4600, loss[loss=0.2178, simple_loss=0.3067, pruned_loss=0.06449, over 17220.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3032, pruned_loss=0.06832, over 3249528.19 frames. ], batch size: 45, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:53:25,888 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:54:12,076 INFO [train.py:904] (2/8) Epoch 8, batch 4650, loss[loss=0.2288, simple_loss=0.3036, pruned_loss=0.07699, over 16797.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3024, pruned_loss=0.06853, over 3250752.26 frames. ], batch size: 39, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:54:13,263 INFO [optim.py:368] (2/8) 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 20:55:23,490 INFO [train.py:904] (2/8) Epoch 8, batch 4700, loss[loss=0.2128, simple_loss=0.3022, pruned_loss=0.06168, over 15430.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2993, pruned_loss=0.06721, over 3242179.71 frames. ], batch size: 190, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:55:50,136 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-28 20:55:58,103 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9869, 3.0842, 3.1544, 1.5524, 3.3545, 3.3804, 2.6126, 2.5457], device='cuda:2'), covar=tensor([0.0806, 0.0145, 0.0142, 0.1110, 0.0058, 0.0078, 0.0369, 0.0380], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0096, 0.0083, 0.0138, 0.0068, 0.0090, 0.0118, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 20:56:31,925 INFO [train.py:904] (2/8) Epoch 8, batch 4750, loss[loss=0.2074, simple_loss=0.2894, pruned_loss=0.0627, over 16415.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2947, pruned_loss=0.06481, over 3236094.92 frames. ], batch size: 146, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:56:33,070 INFO [optim.py:368] (2/8) 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 20:57:25,852 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-28 20:57:44,145 INFO [train.py:904] (2/8) Epoch 8, batch 4800, loss[loss=0.2031, simple_loss=0.2942, pruned_loss=0.05597, over 15311.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2916, pruned_loss=0.06312, over 3223968.93 frames. ], batch size: 190, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:58:32,072 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:58:58,545 INFO [train.py:904] (2/8) Epoch 8, batch 4850, loss[loss=0.2376, simple_loss=0.3134, pruned_loss=0.0809, over 12244.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2922, pruned_loss=0.06285, over 3205286.50 frames. ], batch size: 248, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 20:59:01,503 INFO [optim.py:368] (2/8) 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 20:59:31,912 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4843, 3.5614, 2.8476, 2.1172, 2.4207, 2.2363, 3.6757, 3.4026], device='cuda:2'), covar=tensor([0.2401, 0.0636, 0.1387, 0.2071, 0.1854, 0.1606, 0.0407, 0.0780], device='cuda:2'), in_proj_covar=tensor([0.0288, 0.0249, 0.0273, 0.0261, 0.0278, 0.0209, 0.0254, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 20:59:36,817 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 20:59:46,583 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:59:51,048 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-28 21:00:17,705 INFO [train.py:904] (2/8) Epoch 8, batch 4900, loss[loss=0.1815, simple_loss=0.2716, pruned_loss=0.04566, over 16718.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2916, pruned_loss=0.06199, over 3183910.61 frames. ], batch size: 76, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:00:49,947 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:01:23,210 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4865, 5.3768, 5.2698, 5.0024, 4.7353, 5.2185, 5.1830, 4.9575], device='cuda:2'), covar=tensor([0.0506, 0.0371, 0.0220, 0.0194, 0.1068, 0.0395, 0.0183, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0213, 0.0250, 0.0250, 0.0223, 0.0279, 0.0252, 0.0169, 0.0283], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:01:34,530 INFO [train.py:904] (2/8) Epoch 8, batch 4950, loss[loss=0.2325, simple_loss=0.3111, pruned_loss=0.07701, over 12040.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2913, pruned_loss=0.06156, over 3169007.38 frames. ], batch size: 246, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:01:36,819 INFO [optim.py:368] (2/8) 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 21:02:02,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3935, 3.4477, 2.6387, 2.1231, 2.4565, 2.1618, 3.6586, 3.3364], device='cuda:2'), covar=tensor([0.2526, 0.0724, 0.1446, 0.1967, 0.1839, 0.1502, 0.0464, 0.0792], device='cuda:2'), in_proj_covar=tensor([0.0288, 0.0250, 0.0273, 0.0261, 0.0279, 0.0209, 0.0254, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:02:45,330 INFO [train.py:904] (2/8) Epoch 8, batch 5000, loss[loss=0.1752, simple_loss=0.276, pruned_loss=0.03716, over 16849.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2923, pruned_loss=0.06124, over 3193689.86 frames. ], batch size: 102, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:02:50,927 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9731, 3.0670, 1.6685, 3.2925, 2.3282, 3.3146, 1.8937, 2.4555], device='cuda:2'), covar=tensor([0.0206, 0.0359, 0.1533, 0.0078, 0.0718, 0.0380, 0.1379, 0.0666], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0159, 0.0182, 0.0101, 0.0164, 0.0196, 0.0188, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 21:03:55,754 INFO [train.py:904] (2/8) Epoch 8, batch 5050, loss[loss=0.2057, simple_loss=0.286, pruned_loss=0.06269, over 16476.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2923, pruned_loss=0.06084, over 3205356.51 frames. ], batch size: 75, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:57,927 INFO [optim.py:368] (2/8) 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 21:05:07,314 INFO [train.py:904] (2/8) Epoch 8, batch 5100, loss[loss=0.2047, simple_loss=0.3054, pruned_loss=0.05201, over 16428.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2912, pruned_loss=0.0605, over 3197611.75 frames. ], batch size: 146, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:05:08,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 21:06:20,870 INFO [train.py:904] (2/8) Epoch 8, batch 5150, loss[loss=0.2003, simple_loss=0.2969, pruned_loss=0.05189, over 16925.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2908, pruned_loss=0.0596, over 3189440.19 frames. ], batch size: 96, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:24,096 INFO [optim.py:368] (2/8) 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 21:07:33,713 INFO [train.py:904] (2/8) Epoch 8, batch 5200, loss[loss=0.1996, simple_loss=0.2801, pruned_loss=0.05957, over 16731.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2897, pruned_loss=0.05919, over 3180816.92 frames. ], batch size: 89, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:07:59,770 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:08:36,508 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:08:37,673 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 21:08:48,738 INFO [train.py:904] (2/8) Epoch 8, batch 5250, loss[loss=0.2283, simple_loss=0.3075, pruned_loss=0.07455, over 16740.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2874, pruned_loss=0.05861, over 3197022.76 frames. ], batch size: 124, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:08:51,148 INFO [optim.py:368] (2/8) 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 21:09:19,287 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:09:30,926 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:10:01,828 INFO [train.py:904] (2/8) Epoch 8, batch 5300, loss[loss=0.1758, simple_loss=0.2647, pruned_loss=0.04349, over 15358.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2841, pruned_loss=0.05721, over 3192118.52 frames. ], batch size: 190, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:10:06,458 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:10:47,985 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:10:49,875 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4515, 3.4520, 2.7979, 2.2015, 2.3564, 2.2553, 3.6160, 3.3213], device='cuda:2'), covar=tensor([0.2280, 0.0576, 0.1274, 0.1913, 0.1842, 0.1517, 0.0401, 0.0748], device='cuda:2'), in_proj_covar=tensor([0.0289, 0.0251, 0.0277, 0.0264, 0.0281, 0.0210, 0.0260, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:11:13,479 INFO [train.py:904] (2/8) Epoch 8, batch 5350, loss[loss=0.2168, simple_loss=0.2962, pruned_loss=0.06875, over 16630.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2827, pruned_loss=0.05682, over 3178739.10 frames. ], batch size: 62, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:11:15,924 INFO [optim.py:368] (2/8) 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 21:11:50,275 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 21:12:16,844 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 21:12:26,519 INFO [train.py:904] (2/8) Epoch 8, batch 5400, loss[loss=0.1998, simple_loss=0.2824, pruned_loss=0.05862, over 17101.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2855, pruned_loss=0.05757, over 3189584.83 frames. ], batch size: 47, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:31,995 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9999, 5.4038, 4.7791, 5.2656, 4.8841, 4.6202, 5.0504, 5.4325], device='cuda:2'), covar=tensor([0.1585, 0.1194, 0.2077, 0.0836, 0.1152, 0.1225, 0.1212, 0.1398], device='cuda:2'), in_proj_covar=tensor([0.0463, 0.0586, 0.0493, 0.0403, 0.0373, 0.0382, 0.0488, 0.0429], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:13:43,656 INFO [train.py:904] (2/8) Epoch 8, batch 5450, loss[loss=0.2767, simple_loss=0.35, pruned_loss=0.1017, over 15451.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2892, pruned_loss=0.0599, over 3185927.12 frames. ], batch size: 191, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:46,708 INFO [optim.py:368] (2/8) 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 21:14:33,573 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 21:14:35,616 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 21:14:54,743 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4617, 3.6541, 1.7422, 3.9727, 2.4726, 3.9922, 1.9845, 2.6910], device='cuda:2'), covar=tensor([0.0195, 0.0285, 0.1714, 0.0100, 0.0845, 0.0349, 0.1580, 0.0647], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0158, 0.0184, 0.0101, 0.0165, 0.0196, 0.0190, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 21:15:01,741 INFO [train.py:904] (2/8) Epoch 8, batch 5500, loss[loss=0.2571, simple_loss=0.3391, pruned_loss=0.08753, over 16612.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2971, pruned_loss=0.06527, over 3158104.91 frames. ], batch size: 57, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:16:22,240 INFO [train.py:904] (2/8) Epoch 8, batch 5550, loss[loss=0.3431, simple_loss=0.3865, pruned_loss=0.1499, over 10884.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3053, pruned_loss=0.07155, over 3127313.02 frames. ], batch size: 246, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:16:26,052 INFO [optim.py:368] (2/8) 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 21:17:00,669 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:17:11,592 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6269, 4.5360, 4.4235, 3.7719, 4.5130, 1.6045, 4.2730, 4.3152], device='cuda:2'), covar=tensor([0.0066, 0.0062, 0.0115, 0.0339, 0.0064, 0.2137, 0.0097, 0.0152], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0099, 0.0147, 0.0145, 0.0114, 0.0160, 0.0132, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:17:31,282 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7881, 3.6365, 3.8274, 3.6804, 3.7786, 4.1593, 3.8990, 3.6382], device='cuda:2'), covar=tensor([0.2000, 0.2174, 0.1855, 0.2404, 0.2698, 0.1686, 0.1584, 0.2658], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0436, 0.0453, 0.0383, 0.0509, 0.0482, 0.0369, 0.0516], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 21:17:40,834 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:17:43,810 INFO [train.py:904] (2/8) Epoch 8, batch 5600, loss[loss=0.2506, simple_loss=0.3252, pruned_loss=0.08805, over 16440.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3114, pruned_loss=0.07671, over 3099400.26 frames. ], batch size: 146, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:18:28,317 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 21:19:06,823 INFO [train.py:904] (2/8) Epoch 8, batch 5650, loss[loss=0.2491, simple_loss=0.325, pruned_loss=0.08665, over 16497.00 frames. ], tot_loss[loss=0.241, simple_loss=0.318, pruned_loss=0.08201, over 3075423.08 frames. ], batch size: 68, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:19:10,210 INFO [optim.py:368] (2/8) 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 21:19:11,157 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 21:19:35,751 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 21:20:27,956 INFO [train.py:904] (2/8) Epoch 8, batch 5700, loss[loss=0.2573, simple_loss=0.3473, pruned_loss=0.08371, over 16468.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3194, pruned_loss=0.0835, over 3081843.23 frames. ], batch size: 75, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:21:05,231 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0008, 3.3110, 3.4987, 3.4679, 3.4384, 3.2518, 3.2970, 3.3465], device='cuda:2'), covar=tensor([0.0394, 0.0585, 0.0441, 0.0462, 0.0489, 0.0450, 0.0829, 0.0515], device='cuda:2'), in_proj_covar=tensor([0.0287, 0.0292, 0.0294, 0.0281, 0.0332, 0.0312, 0.0416, 0.0256], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 21:21:15,118 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 21:21:36,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7553, 1.2428, 1.7156, 1.6252, 1.6146, 1.8378, 1.4362, 1.7105], device='cuda:2'), covar=tensor([0.0135, 0.0208, 0.0127, 0.0147, 0.0161, 0.0100, 0.0243, 0.0069], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0158, 0.0143, 0.0142, 0.0151, 0.0106, 0.0160, 0.0098], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 21:21:41,706 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 21:21:49,369 INFO [train.py:904] (2/8) Epoch 8, batch 5750, loss[loss=0.219, simple_loss=0.3026, pruned_loss=0.06772, over 16399.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3216, pruned_loss=0.08432, over 3070864.81 frames. ], batch size: 68, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:21:54,072 INFO [optim.py:368] (2/8) 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 21:22:52,018 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:23:12,544 INFO [train.py:904] (2/8) Epoch 8, batch 5800, loss[loss=0.1993, simple_loss=0.294, pruned_loss=0.05233, over 16853.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3208, pruned_loss=0.08241, over 3086277.74 frames. ], batch size: 102, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:23:28,948 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4389, 3.3262, 3.3972, 3.5279, 3.5491, 3.2569, 3.5014, 3.5790], device='cuda:2'), covar=tensor([0.0891, 0.0771, 0.1095, 0.0530, 0.0560, 0.2338, 0.0891, 0.0637], device='cuda:2'), in_proj_covar=tensor([0.0452, 0.0550, 0.0685, 0.0562, 0.0425, 0.0425, 0.0438, 0.0488], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:24:30,518 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:24:32,997 INFO [train.py:904] (2/8) Epoch 8, batch 5850, loss[loss=0.2501, simple_loss=0.312, pruned_loss=0.09412, over 11545.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3192, pruned_loss=0.08163, over 3055700.78 frames. ], batch size: 246, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:24:37,986 INFO [optim.py:368] (2/8) 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 21:25:09,371 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:25:52,773 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:25:56,469 INFO [train.py:904] (2/8) Epoch 8, batch 5900, loss[loss=0.2092, simple_loss=0.2914, pruned_loss=0.06353, over 15388.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3186, pruned_loss=0.08116, over 3066458.57 frames. ], batch size: 190, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:26:33,735 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:26:42,754 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:26:45,223 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7746, 4.1295, 3.2184, 2.4030, 3.1196, 2.5885, 4.2349, 4.0698], device='cuda:2'), covar=tensor([0.2289, 0.0610, 0.1335, 0.1925, 0.1824, 0.1505, 0.0476, 0.0684], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0251, 0.0273, 0.0262, 0.0277, 0.0210, 0.0259, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:26:55,092 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4011, 4.1874, 4.2528, 2.9340, 3.9176, 4.2571, 4.0005, 2.5886], device='cuda:2'), covar=tensor([0.0377, 0.0024, 0.0025, 0.0255, 0.0040, 0.0053, 0.0036, 0.0264], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0064, 0.0065, 0.0123, 0.0069, 0.0080, 0.0071, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 21:27:11,067 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:27:17,811 INFO [train.py:904] (2/8) Epoch 8, batch 5950, loss[loss=0.2236, simple_loss=0.3088, pruned_loss=0.06914, over 16618.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3195, pruned_loss=0.08031, over 3069754.51 frames. ], batch size: 62, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:27:21,560 INFO [optim.py:368] (2/8) 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 21:27:55,398 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:28:33,578 INFO [train.py:904] (2/8) Epoch 8, batch 6000, loss[loss=0.2182, simple_loss=0.3007, pruned_loss=0.06782, over 16646.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3184, pruned_loss=0.07977, over 3075444.23 frames. ], batch size: 62, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:28:33,578 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 21:28:44,108 INFO [train.py:938] (2/8) Epoch 8, validation: loss=0.1707, simple_loss=0.284, pruned_loss=0.02871, over 944034.00 frames. 2023-04-28 21:28:44,109 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-28 21:29:48,557 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:30:00,317 INFO [train.py:904] (2/8) Epoch 8, batch 6050, loss[loss=0.2277, simple_loss=0.3148, pruned_loss=0.07028, over 16694.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3164, pruned_loss=0.07849, over 3099235.27 frames. ], batch size: 134, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:30:04,223 INFO [optim.py:368] (2/8) 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 21:30:50,595 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8296, 4.8272, 4.6699, 4.3962, 4.2262, 4.7520, 4.6378, 4.4179], device='cuda:2'), covar=tensor([0.0594, 0.0441, 0.0260, 0.0235, 0.0952, 0.0436, 0.0295, 0.0583], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0253, 0.0249, 0.0219, 0.0278, 0.0256, 0.0171, 0.0287], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:31:00,638 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2720, 2.0578, 1.5071, 1.8071, 2.2672, 2.0803, 2.3639, 2.5055], device='cuda:2'), covar=tensor([0.0080, 0.0227, 0.0338, 0.0292, 0.0132, 0.0219, 0.0120, 0.0132], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0180, 0.0177, 0.0179, 0.0176, 0.0179, 0.0176, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:31:16,284 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6388, 2.1713, 2.4530, 4.3629, 2.0955, 2.6730, 2.2342, 2.3921], device='cuda:2'), covar=tensor([0.0724, 0.2747, 0.1643, 0.0278, 0.3311, 0.1820, 0.2634, 0.2423], device='cuda:2'), in_proj_covar=tensor([0.0343, 0.0365, 0.0306, 0.0319, 0.0398, 0.0406, 0.0327, 0.0432], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:31:19,254 INFO [train.py:904] (2/8) Epoch 8, batch 6100, loss[loss=0.2176, simple_loss=0.2973, pruned_loss=0.06897, over 17008.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3155, pruned_loss=0.07708, over 3110648.60 frames. ], batch size: 41, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:31:26,122 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:31:46,160 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4552, 2.9878, 2.5803, 2.2743, 2.4040, 2.1328, 2.9952, 2.9830], device='cuda:2'), covar=tensor([0.2137, 0.0666, 0.1201, 0.1812, 0.1948, 0.1669, 0.0505, 0.0876], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0253, 0.0275, 0.0264, 0.0280, 0.0211, 0.0260, 0.0280], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:31:48,523 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6428, 2.7908, 2.3537, 3.9168, 2.9374, 3.8864, 1.3613, 2.8739], device='cuda:2'), covar=tensor([0.1292, 0.0568, 0.1157, 0.0101, 0.0239, 0.0358, 0.1572, 0.0761], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0151, 0.0174, 0.0118, 0.0202, 0.0205, 0.0174, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 21:32:26,699 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:32:37,684 INFO [train.py:904] (2/8) Epoch 8, batch 6150, loss[loss=0.2321, simple_loss=0.3068, pruned_loss=0.07867, over 15469.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3148, pruned_loss=0.07764, over 3084380.93 frames. ], batch size: 191, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:32:42,674 INFO [optim.py:368] (2/8) 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 21:32:52,938 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:33:59,282 INFO [train.py:904] (2/8) Epoch 8, batch 6200, loss[loss=0.2074, simple_loss=0.2981, pruned_loss=0.05834, over 16471.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3129, pruned_loss=0.07752, over 3067173.00 frames. ], batch size: 75, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:34:31,237 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:33,162 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:39,087 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:35:16,095 INFO [train.py:904] (2/8) Epoch 8, batch 6250, loss[loss=0.2673, simple_loss=0.322, pruned_loss=0.1063, over 12185.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3116, pruned_loss=0.07642, over 3085598.67 frames. ], batch size: 247, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:35:22,794 INFO [optim.py:368] (2/8) 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 21:36:07,271 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:12,980 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:35,475 INFO [train.py:904] (2/8) Epoch 8, batch 6300, loss[loss=0.2396, simple_loss=0.3274, pruned_loss=0.07584, over 16599.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3117, pruned_loss=0.07571, over 3087980.98 frames. ], batch size: 62, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:37:01,384 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 21:37:36,884 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6770, 3.7562, 3.0629, 2.3325, 2.7041, 2.2828, 3.9293, 3.6412], device='cuda:2'), covar=tensor([0.2298, 0.0599, 0.1296, 0.1850, 0.1953, 0.1647, 0.0411, 0.0806], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0251, 0.0274, 0.0263, 0.0279, 0.0210, 0.0257, 0.0278], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:37:42,476 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7998, 4.0361, 3.3621, 2.3311, 3.0201, 2.4392, 4.3289, 3.9224], device='cuda:2'), covar=tensor([0.2253, 0.0589, 0.1156, 0.1865, 0.2050, 0.1530, 0.0359, 0.0685], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0250, 0.0273, 0.0262, 0.0278, 0.0209, 0.0257, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:37:54,085 INFO [train.py:904] (2/8) Epoch 8, batch 6350, loss[loss=0.2877, simple_loss=0.3394, pruned_loss=0.118, over 11527.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3116, pruned_loss=0.07652, over 3096042.93 frames. ], batch size: 247, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:38:00,461 INFO [optim.py:368] (2/8) 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 21:39:05,977 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 21:39:10,468 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 21:39:11,215 INFO [train.py:904] (2/8) Epoch 8, batch 6400, loss[loss=0.2283, simple_loss=0.3121, pruned_loss=0.07225, over 16910.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3127, pruned_loss=0.07787, over 3094586.06 frames. ], batch size: 109, lr: 8.49e-03, grad_scale: 8.0 2023-04-28 21:39:29,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3166, 4.2763, 4.7618, 4.6933, 4.7221, 4.3449, 4.3902, 4.1908], device='cuda:2'), covar=tensor([0.0249, 0.0423, 0.0279, 0.0361, 0.0361, 0.0311, 0.0793, 0.0426], device='cuda:2'), in_proj_covar=tensor([0.0289, 0.0294, 0.0296, 0.0282, 0.0331, 0.0312, 0.0417, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 21:40:15,893 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-28 21:40:17,149 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:40:26,236 INFO [train.py:904] (2/8) Epoch 8, batch 6450, loss[loss=0.2273, simple_loss=0.3087, pruned_loss=0.07296, over 16788.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3124, pruned_loss=0.07744, over 3079451.97 frames. ], batch size: 62, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:40:33,079 INFO [optim.py:368] (2/8) 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 21:40:58,003 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-28 21:41:25,901 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-28 21:41:31,679 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:41:45,633 INFO [train.py:904] (2/8) Epoch 8, batch 6500, loss[loss=0.2258, simple_loss=0.3091, pruned_loss=0.0712, over 16875.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.31, pruned_loss=0.07663, over 3068709.92 frames. ], batch size: 116, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:42:09,105 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:42:45,833 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5277, 2.6140, 1.7594, 2.7375, 2.1296, 2.7382, 2.0047, 2.3436], device='cuda:2'), covar=tensor([0.0227, 0.0312, 0.1145, 0.0135, 0.0565, 0.0426, 0.1032, 0.0501], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0158, 0.0182, 0.0102, 0.0163, 0.0197, 0.0191, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 21:43:05,178 INFO [train.py:904] (2/8) Epoch 8, batch 6550, loss[loss=0.22, simple_loss=0.3184, pruned_loss=0.06085, over 16717.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.313, pruned_loss=0.07725, over 3078750.65 frames. ], batch size: 76, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:43:11,237 INFO [optim.py:368] (2/8) 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 21:43:28,263 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:36,825 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8407, 2.5478, 2.5550, 1.9241, 2.3700, 2.5213, 2.4789, 1.8695], device='cuda:2'), covar=tensor([0.0303, 0.0053, 0.0048, 0.0247, 0.0070, 0.0066, 0.0051, 0.0291], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0062, 0.0063, 0.0119, 0.0067, 0.0078, 0.0069, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 21:43:47,921 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:53,394 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:44:22,369 INFO [train.py:904] (2/8) Epoch 8, batch 6600, loss[loss=0.2919, simple_loss=0.3488, pruned_loss=0.1175, over 11613.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3157, pruned_loss=0.07799, over 3085236.54 frames. ], batch size: 247, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:45:00,813 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:00,860 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:38,843 INFO [train.py:904] (2/8) Epoch 8, batch 6650, loss[loss=0.2638, simple_loss=0.3358, pruned_loss=0.09592, over 15431.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3171, pruned_loss=0.08025, over 3058179.87 frames. ], batch size: 191, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:45:40,032 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 21:45:45,538 INFO [optim.py:368] (2/8) 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 21:46:23,115 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 21:46:35,360 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:46:53,096 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:46:54,545 INFO [train.py:904] (2/8) Epoch 8, batch 6700, loss[loss=0.2397, simple_loss=0.3186, pruned_loss=0.08042, over 16651.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3157, pruned_loss=0.08015, over 3055276.46 frames. ], batch size: 124, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:47:49,763 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2179, 3.8291, 3.4019, 1.9049, 2.9247, 2.5396, 3.5096, 3.8121], device='cuda:2'), covar=tensor([0.0257, 0.0468, 0.0602, 0.1738, 0.0775, 0.0833, 0.0608, 0.0829], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0132, 0.0155, 0.0139, 0.0132, 0.0123, 0.0135, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 21:48:06,137 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:48:10,952 INFO [train.py:904] (2/8) Epoch 8, batch 6750, loss[loss=0.2308, simple_loss=0.2996, pruned_loss=0.08097, over 16626.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3135, pruned_loss=0.07964, over 3054885.29 frames. ], batch size: 62, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:18,410 INFO [optim.py:368] (2/8) 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 21:48:26,267 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 21:48:31,425 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 21:49:21,153 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 21:49:25,315 INFO [train.py:904] (2/8) Epoch 8, batch 6800, loss[loss=0.2187, simple_loss=0.3066, pruned_loss=0.06544, over 16905.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3126, pruned_loss=0.07838, over 3073771.81 frames. ], batch size: 96, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:49:49,081 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:49:50,728 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-28 21:50:44,158 INFO [train.py:904] (2/8) Epoch 8, batch 6850, loss[loss=0.2187, simple_loss=0.3153, pruned_loss=0.06106, over 16716.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3147, pruned_loss=0.07942, over 3062907.46 frames. ], batch size: 57, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:50:53,210 INFO [optim.py:368] (2/8) 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 21:51:03,698 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:24,281 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:30,559 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:52,160 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.31 vs. limit=5.0 2023-04-28 21:51:59,510 INFO [train.py:904] (2/8) Epoch 8, batch 6900, loss[loss=0.2295, simple_loss=0.3184, pruned_loss=0.07027, over 16564.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3168, pruned_loss=0.07914, over 3063277.60 frames. ], batch size: 68, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:52:13,289 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 21:52:31,117 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:39,049 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:45,251 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:53:20,813 INFO [train.py:904] (2/8) Epoch 8, batch 6950, loss[loss=0.2556, simple_loss=0.3326, pruned_loss=0.08931, over 16339.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3191, pruned_loss=0.08162, over 3045159.09 frames. ], batch size: 146, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:53:29,771 INFO [optim.py:368] (2/8) 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 21:54:10,657 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78033.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:54:38,871 INFO [train.py:904] (2/8) Epoch 8, batch 7000, loss[loss=0.2327, simple_loss=0.3154, pruned_loss=0.07501, over 16463.00 frames. ], tot_loss[loss=0.24, simple_loss=0.319, pruned_loss=0.08045, over 3056247.03 frames. ], batch size: 146, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:55:38,827 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2666, 5.2185, 4.9895, 4.2450, 5.0745, 1.7128, 4.8287, 4.8794], device='cuda:2'), covar=tensor([0.0056, 0.0043, 0.0112, 0.0345, 0.0057, 0.2066, 0.0085, 0.0148], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0097, 0.0144, 0.0140, 0.0114, 0.0161, 0.0129, 0.0135], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 21:55:53,774 INFO [train.py:904] (2/8) Epoch 8, batch 7050, loss[loss=0.2034, simple_loss=0.2927, pruned_loss=0.05707, over 16682.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3187, pruned_loss=0.07944, over 3058790.05 frames. ], batch size: 76, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:56:03,900 INFO [optim.py:368] (2/8) 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 21:57:09,151 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0056, 2.6501, 2.6098, 1.8182, 2.7733, 2.8161, 2.4470, 2.3186], device='cuda:2'), covar=tensor([0.0712, 0.0175, 0.0207, 0.0984, 0.0095, 0.0151, 0.0396, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0098, 0.0085, 0.0141, 0.0069, 0.0091, 0.0120, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 21:57:09,208 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0763, 2.4907, 2.2552, 2.8322, 2.2265, 3.2245, 1.7837, 2.7207], device='cuda:2'), covar=tensor([0.0979, 0.0451, 0.0864, 0.0133, 0.0147, 0.0336, 0.1218, 0.0596], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0151, 0.0171, 0.0117, 0.0199, 0.0202, 0.0172, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 21:57:11,209 INFO [train.py:904] (2/8) Epoch 8, batch 7100, loss[loss=0.2335, simple_loss=0.3109, pruned_loss=0.07802, over 15255.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3177, pruned_loss=0.07963, over 3053352.92 frames. ], batch size: 190, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:57:47,161 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0221, 3.0834, 1.7470, 3.2840, 2.2810, 3.2868, 1.9366, 2.4743], device='cuda:2'), covar=tensor([0.0217, 0.0386, 0.1596, 0.0118, 0.0818, 0.0582, 0.1479, 0.0735], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0160, 0.0183, 0.0102, 0.0166, 0.0200, 0.0194, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 21:58:08,851 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 21:58:14,434 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 21:58:26,606 INFO [train.py:904] (2/8) Epoch 8, batch 7150, loss[loss=0.2306, simple_loss=0.3094, pruned_loss=0.07588, over 15315.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3158, pruned_loss=0.07967, over 3057834.89 frames. ], batch size: 190, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:36,170 INFO [optim.py:368] (2/8) 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 21:59:09,645 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 8, batch 7200, loss[loss=0.2075, simple_loss=0.2971, pruned_loss=0.05896, over 16746.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3129, pruned_loss=0.07712, over 3062076.82 frames. ], batch size: 134, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:00:09,979 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:00:44,920 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:00:59,552 INFO [train.py:904] (2/8) Epoch 8, batch 7250, loss[loss=0.2368, simple_loss=0.3014, pruned_loss=0.08612, over 11221.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3106, pruned_loss=0.07557, over 3068318.77 frames. ], batch size: 246, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:01:10,047 INFO [optim.py:368] (2/8) 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 22:01:26,987 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:01:47,559 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:02:16,538 INFO [train.py:904] (2/8) Epoch 8, batch 7300, loss[loss=0.2221, simple_loss=0.3059, pruned_loss=0.06916, over 16708.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3092, pruned_loss=0.07466, over 3074528.86 frames. ], batch size: 76, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:02:33,360 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8452, 2.6654, 1.9263, 2.3295, 3.0721, 2.6935, 3.5982, 3.4108], device='cuda:2'), covar=tensor([0.0031, 0.0252, 0.0384, 0.0326, 0.0149, 0.0254, 0.0111, 0.0134], device='cuda:2'), in_proj_covar=tensor([0.0109, 0.0180, 0.0179, 0.0180, 0.0178, 0.0181, 0.0176, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:02:40,954 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8780, 4.1346, 3.9336, 3.9750, 3.6474, 3.7286, 3.8245, 4.1013], device='cuda:2'), covar=tensor([0.0884, 0.0779, 0.0887, 0.0572, 0.0716, 0.1434, 0.0767, 0.0848], device='cuda:2'), in_proj_covar=tensor([0.0460, 0.0579, 0.0492, 0.0394, 0.0362, 0.0387, 0.0480, 0.0426], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:03:02,794 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:03:06,378 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-04-28 22:03:34,470 INFO [train.py:904] (2/8) Epoch 8, batch 7350, loss[loss=0.2332, simple_loss=0.3071, pruned_loss=0.07965, over 16727.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3092, pruned_loss=0.07493, over 3066052.63 frames. ], batch size: 134, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:03:45,280 INFO [optim.py:368] (2/8) 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 22:04:55,409 INFO [train.py:904] (2/8) Epoch 8, batch 7400, loss[loss=0.2254, simple_loss=0.3037, pruned_loss=0.07352, over 16693.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3103, pruned_loss=0.07529, over 3084281.23 frames. ], batch size: 62, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:05:07,830 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:05:30,097 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 22:05:32,791 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0165, 4.0627, 4.4653, 4.4335, 4.3935, 4.0694, 4.0924, 3.9552], device='cuda:2'), covar=tensor([0.0296, 0.0417, 0.0343, 0.0376, 0.0486, 0.0321, 0.0861, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0286, 0.0288, 0.0294, 0.0280, 0.0331, 0.0309, 0.0407, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 22:06:13,356 INFO [train.py:904] (2/8) Epoch 8, batch 7450, loss[loss=0.2318, simple_loss=0.326, pruned_loss=0.06879, over 15382.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3124, pruned_loss=0.07693, over 3074382.45 frames. ], batch size: 191, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:06:26,556 INFO [optim.py:368] (2/8) 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 22:06:43,908 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0071, 2.3683, 2.2467, 2.9921, 2.1949, 3.3046, 1.5992, 2.7059], device='cuda:2'), covar=tensor([0.1149, 0.0508, 0.0990, 0.0157, 0.0153, 0.0440, 0.1360, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0154, 0.0174, 0.0119, 0.0202, 0.0205, 0.0175, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 22:06:47,401 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:07:06,593 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 22:07:34,693 INFO [train.py:904] (2/8) Epoch 8, batch 7500, loss[loss=0.2746, simple_loss=0.3341, pruned_loss=0.1076, over 11200.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3126, pruned_loss=0.07632, over 3084323.96 frames. ], batch size: 246, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:07:42,121 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:08:33,284 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:08:55,094 INFO [train.py:904] (2/8) Epoch 8, batch 7550, loss[loss=0.1912, simple_loss=0.2759, pruned_loss=0.05321, over 16506.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3118, pruned_loss=0.07652, over 3072151.78 frames. ], batch size: 75, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:08:55,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3916, 2.0627, 1.6507, 1.8938, 2.4493, 2.1800, 2.4843, 2.6388], device='cuda:2'), covar=tensor([0.0075, 0.0261, 0.0346, 0.0307, 0.0137, 0.0218, 0.0129, 0.0139], device='cuda:2'), in_proj_covar=tensor([0.0108, 0.0180, 0.0179, 0.0181, 0.0177, 0.0180, 0.0175, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:09:05,664 INFO [optim.py:368] (2/8) 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 22:09:19,442 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:09:38,229 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 22:09:50,952 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8448, 1.9836, 2.2996, 3.1642, 2.0843, 2.2348, 2.2048, 2.0585], device='cuda:2'), covar=tensor([0.0786, 0.2543, 0.1524, 0.0510, 0.3170, 0.1854, 0.2254, 0.2725], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0366, 0.0304, 0.0317, 0.0401, 0.0402, 0.0324, 0.0429], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:10:04,246 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3058, 3.2640, 3.3199, 3.4376, 3.4711, 3.1647, 3.4355, 3.5087], device='cuda:2'), covar=tensor([0.0986, 0.0798, 0.1013, 0.0573, 0.0637, 0.2052, 0.0893, 0.0616], device='cuda:2'), in_proj_covar=tensor([0.0464, 0.0569, 0.0699, 0.0576, 0.0444, 0.0428, 0.0463, 0.0503], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:10:12,495 INFO [train.py:904] (2/8) Epoch 8, batch 7600, loss[loss=0.2207, simple_loss=0.2983, pruned_loss=0.07155, over 16847.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3106, pruned_loss=0.07659, over 3065480.94 frames. ], batch size: 109, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:10:49,275 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3188, 5.2471, 5.1630, 4.9361, 4.6974, 5.1222, 5.1222, 4.8132], device='cuda:2'), covar=tensor([0.0592, 0.0383, 0.0252, 0.0245, 0.0935, 0.0452, 0.0316, 0.0614], device='cuda:2'), in_proj_covar=tensor([0.0211, 0.0251, 0.0247, 0.0219, 0.0273, 0.0253, 0.0172, 0.0286], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:10:58,279 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5952, 3.6903, 1.8583, 4.1183, 2.3627, 4.1065, 1.9138, 2.7982], device='cuda:2'), covar=tensor([0.0207, 0.0360, 0.1762, 0.0109, 0.0966, 0.0409, 0.1734, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0159, 0.0183, 0.0103, 0.0167, 0.0197, 0.0192, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 22:11:30,204 INFO [train.py:904] (2/8) Epoch 8, batch 7650, loss[loss=0.2327, simple_loss=0.3142, pruned_loss=0.0756, over 16776.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3104, pruned_loss=0.07634, over 3083408.03 frames. ], batch size: 83, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:11:40,445 INFO [optim.py:368] (2/8) 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 22:12:02,170 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0430, 3.0729, 1.7095, 3.2899, 2.2491, 3.2676, 1.9097, 2.5535], device='cuda:2'), covar=tensor([0.0205, 0.0334, 0.1393, 0.0094, 0.0743, 0.0443, 0.1366, 0.0584], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0159, 0.0183, 0.0103, 0.0167, 0.0198, 0.0193, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 22:12:45,889 INFO [train.py:904] (2/8) Epoch 8, batch 7700, loss[loss=0.2941, simple_loss=0.3432, pruned_loss=0.1225, over 11517.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3114, pruned_loss=0.07776, over 3065975.11 frames. ], batch size: 246, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:14:03,978 INFO [train.py:904] (2/8) Epoch 8, batch 7750, loss[loss=0.2014, simple_loss=0.2943, pruned_loss=0.05427, over 16876.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3116, pruned_loss=0.07779, over 3073380.21 frames. ], batch size: 42, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:14:09,717 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4593, 3.4151, 3.4112, 2.8128, 3.3411, 2.1083, 3.1120, 2.7731], device='cuda:2'), covar=tensor([0.0111, 0.0090, 0.0134, 0.0210, 0.0079, 0.1698, 0.0108, 0.0168], device='cuda:2'), in_proj_covar=tensor([0.0110, 0.0096, 0.0143, 0.0140, 0.0114, 0.0162, 0.0128, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:14:15,893 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8297, 3.8557, 4.2625, 4.2080, 4.1751, 3.9046, 3.9470, 3.9126], device='cuda:2'), covar=tensor([0.0335, 0.0525, 0.0378, 0.0430, 0.0452, 0.0382, 0.0868, 0.0473], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0300, 0.0302, 0.0291, 0.0343, 0.0319, 0.0422, 0.0262], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 22:14:17,807 INFO [optim.py:368] (2/8) 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 22:14:26,063 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:14:38,175 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1187, 4.1478, 3.9603, 3.8231, 3.6428, 4.0401, 3.7953, 3.7207], device='cuda:2'), covar=tensor([0.0590, 0.0465, 0.0282, 0.0248, 0.0777, 0.0460, 0.0764, 0.0651], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0255, 0.0250, 0.0222, 0.0276, 0.0257, 0.0174, 0.0292], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:14:39,537 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0862, 2.3820, 2.2364, 2.8774, 2.1965, 3.2113, 1.7191, 2.7822], device='cuda:2'), covar=tensor([0.0995, 0.0466, 0.0889, 0.0130, 0.0142, 0.0365, 0.1180, 0.0566], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0155, 0.0175, 0.0121, 0.0204, 0.0206, 0.0177, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 22:14:41,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8992, 1.6394, 2.3977, 2.7477, 2.7287, 3.1219, 1.7966, 3.0345], device='cuda:2'), covar=tensor([0.0105, 0.0337, 0.0196, 0.0184, 0.0157, 0.0094, 0.0336, 0.0087], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0157, 0.0141, 0.0140, 0.0147, 0.0106, 0.0156, 0.0098], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 22:15:19,809 INFO [train.py:904] (2/8) Epoch 8, batch 7800, loss[loss=0.2209, simple_loss=0.3058, pruned_loss=0.06805, over 16952.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3122, pruned_loss=0.07835, over 3073569.02 frames. ], batch size: 109, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:16,683 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:16:38,226 INFO [train.py:904] (2/8) Epoch 8, batch 7850, loss[loss=0.2325, simple_loss=0.3094, pruned_loss=0.07778, over 15320.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3137, pruned_loss=0.0783, over 3078397.93 frames. ], batch size: 190, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:50,959 INFO [optim.py:368] (2/8) 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 22:16:53,429 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:26,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8204, 3.7304, 3.8830, 4.0091, 4.0509, 3.6526, 3.9889, 4.0567], device='cuda:2'), covar=tensor([0.1096, 0.0784, 0.0967, 0.0498, 0.0528, 0.1578, 0.0660, 0.0517], device='cuda:2'), in_proj_covar=tensor([0.0463, 0.0567, 0.0697, 0.0579, 0.0443, 0.0427, 0.0462, 0.0504], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:17:29,115 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:54,229 INFO [train.py:904] (2/8) Epoch 8, batch 7900, loss[loss=0.2278, simple_loss=0.3094, pruned_loss=0.07308, over 17147.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3124, pruned_loss=0.07742, over 3081855.84 frames. ], batch size: 47, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:18:15,528 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:19:13,438 INFO [train.py:904] (2/8) Epoch 8, batch 7950, loss[loss=0.2132, simple_loss=0.2931, pruned_loss=0.06667, over 16890.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.313, pruned_loss=0.07803, over 3085910.71 frames. ], batch size: 116, lr: 8.40e-03, grad_scale: 2.0 2023-04-28 22:19:28,064 INFO [optim.py:368] (2/8) 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 22:19:52,441 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:20:21,140 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 22:20:32,280 INFO [train.py:904] (2/8) Epoch 8, batch 8000, loss[loss=0.2416, simple_loss=0.3251, pruned_loss=0.07909, over 16809.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3143, pruned_loss=0.07918, over 3070107.27 frames. ], batch size: 83, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:21:48,606 INFO [train.py:904] (2/8) Epoch 8, batch 8050, loss[loss=0.2284, simple_loss=0.3116, pruned_loss=0.07257, over 16686.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3138, pruned_loss=0.07872, over 3064718.91 frames. ], batch size: 57, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:21:55,238 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4189, 1.9715, 2.0612, 4.0636, 2.0050, 2.4760, 2.0907, 2.1730], device='cuda:2'), covar=tensor([0.0816, 0.3210, 0.2055, 0.0358, 0.3507, 0.2066, 0.2833, 0.3017], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0370, 0.0309, 0.0320, 0.0407, 0.0409, 0.0329, 0.0436], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:22:02,033 INFO [optim.py:368] (2/8) 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 22:22:11,216 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:22:31,083 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2528, 4.2524, 4.6809, 4.6226, 4.6355, 4.2665, 4.3299, 4.1359], device='cuda:2'), covar=tensor([0.0256, 0.0421, 0.0288, 0.0368, 0.0427, 0.0319, 0.0817, 0.0453], device='cuda:2'), in_proj_covar=tensor([0.0286, 0.0291, 0.0293, 0.0283, 0.0332, 0.0309, 0.0408, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 22:23:05,315 INFO [train.py:904] (2/8) Epoch 8, batch 8100, loss[loss=0.2292, simple_loss=0.3155, pruned_loss=0.0714, over 16201.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3126, pruned_loss=0.07754, over 3083322.92 frames. ], batch size: 35, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:23:15,716 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 22:23:23,752 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:23:28,370 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4410, 3.3859, 3.4078, 2.7447, 3.3610, 1.9223, 3.1535, 2.6701], device='cuda:2'), covar=tensor([0.0145, 0.0101, 0.0162, 0.0341, 0.0099, 0.2237, 0.0127, 0.0218], device='cuda:2'), in_proj_covar=tensor([0.0111, 0.0097, 0.0145, 0.0143, 0.0116, 0.0164, 0.0130, 0.0135], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:24:22,973 INFO [train.py:904] (2/8) Epoch 8, batch 8150, loss[loss=0.2403, simple_loss=0.2968, pruned_loss=0.09184, over 11449.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3102, pruned_loss=0.07644, over 3084672.27 frames. ], batch size: 247, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:24:36,887 INFO [optim.py:368] (2/8) 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 22:24:38,554 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:25:42,513 INFO [train.py:904] (2/8) Epoch 8, batch 8200, loss[loss=0.2635, simple_loss=0.3227, pruned_loss=0.1022, over 11683.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3072, pruned_loss=0.07514, over 3091758.90 frames. ], batch size: 247, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:25:55,508 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:26:17,899 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9690, 3.5318, 3.0046, 4.9757, 3.9061, 4.5690, 1.4908, 3.4005], device='cuda:2'), covar=tensor([0.1236, 0.0467, 0.0858, 0.0094, 0.0261, 0.0328, 0.1534, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0152, 0.0171, 0.0118, 0.0200, 0.0199, 0.0172, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 22:26:17,963 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8846, 2.2586, 2.3522, 4.6719, 2.1694, 2.8371, 2.3136, 2.5369], device='cuda:2'), covar=tensor([0.0694, 0.3222, 0.1869, 0.0247, 0.3600, 0.1992, 0.2734, 0.2813], device='cuda:2'), in_proj_covar=tensor([0.0339, 0.0366, 0.0305, 0.0315, 0.0400, 0.0404, 0.0324, 0.0430], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:26:36,282 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:27:04,309 INFO [train.py:904] (2/8) Epoch 8, batch 8250, loss[loss=0.23, simple_loss=0.2979, pruned_loss=0.08099, over 11986.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.306, pruned_loss=0.07353, over 3045806.93 frames. ], batch size: 246, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:27:19,442 INFO [optim.py:368] (2/8) 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 22:27:37,067 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:27:56,941 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:17,659 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:28:26,463 INFO [train.py:904] (2/8) Epoch 8, batch 8300, loss[loss=0.1976, simple_loss=0.2954, pruned_loss=0.04986, over 16854.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3036, pruned_loss=0.07046, over 3043048.10 frames. ], batch size: 102, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:28:38,044 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4710, 3.4523, 3.4888, 2.8775, 3.4031, 2.0011, 3.1963, 2.9114], device='cuda:2'), covar=tensor([0.0100, 0.0080, 0.0112, 0.0203, 0.0079, 0.1877, 0.0107, 0.0156], device='cuda:2'), in_proj_covar=tensor([0.0108, 0.0093, 0.0140, 0.0138, 0.0112, 0.0160, 0.0126, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:28:46,872 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:52,041 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 22:29:05,018 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:29:36,113 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:29:48,237 INFO [train.py:904] (2/8) Epoch 8, batch 8350, loss[loss=0.2, simple_loss=0.2946, pruned_loss=0.0527, over 15401.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3025, pruned_loss=0.0679, over 3048475.61 frames. ], batch size: 191, lr: 8.38e-03, grad_scale: 4.0 2023-04-28 22:30:02,871 INFO [optim.py:368] (2/8) 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 22:30:26,178 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:30:43,697 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:31:09,026 INFO [train.py:904] (2/8) Epoch 8, batch 8400, loss[loss=0.197, simple_loss=0.2903, pruned_loss=0.05182, over 16186.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2997, pruned_loss=0.06529, over 3044831.30 frames. ], batch size: 165, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:31:18,905 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-28 22:32:27,061 INFO [train.py:904] (2/8) Epoch 8, batch 8450, loss[loss=0.1736, simple_loss=0.2707, pruned_loss=0.03826, over 16355.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2972, pruned_loss=0.0633, over 3044230.97 frames. ], batch size: 165, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:42,127 INFO [optim.py:368] (2/8) 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 22:33:47,272 INFO [train.py:904] (2/8) Epoch 8, batch 8500, loss[loss=0.1828, simple_loss=0.26, pruned_loss=0.05284, over 11929.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2926, pruned_loss=0.06015, over 3057425.92 frames. ], batch size: 248, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:34:40,448 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:35:09,511 INFO [train.py:904] (2/8) Epoch 8, batch 8550, loss[loss=0.2237, simple_loss=0.3245, pruned_loss=0.06141, over 16175.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2903, pruned_loss=0.05898, over 3035525.35 frames. ], batch size: 165, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:35:26,481 INFO [optim.py:368] (2/8) 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 22:35:47,390 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:28,572 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:36:36,796 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:48,474 INFO [train.py:904] (2/8) Epoch 8, batch 8600, loss[loss=0.2021, simple_loss=0.2951, pruned_loss=0.05461, over 16734.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2904, pruned_loss=0.05801, over 3030467.81 frames. ], batch size: 134, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:36:59,392 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:37:21,576 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2547, 5.5487, 5.2649, 5.2896, 4.9019, 4.9202, 5.0235, 5.6044], device='cuda:2'), covar=tensor([0.0828, 0.0855, 0.1163, 0.0588, 0.0729, 0.0716, 0.0866, 0.0681], device='cuda:2'), in_proj_covar=tensor([0.0455, 0.0566, 0.0485, 0.0395, 0.0355, 0.0380, 0.0477, 0.0423], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:37:24,926 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:38:02,630 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:38:26,200 INFO [train.py:904] (2/8) Epoch 8, batch 8650, loss[loss=0.1891, simple_loss=0.2879, pruned_loss=0.04519, over 16291.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2876, pruned_loss=0.05552, over 3049896.92 frames. ], batch size: 146, lr: 8.37e-03, grad_scale: 4.0 2023-04-28 22:38:50,263 INFO [optim.py:368] (2/8) 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 22:39:04,693 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:32,714 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:40:12,066 INFO [train.py:904] (2/8) Epoch 8, batch 8700, loss[loss=0.1784, simple_loss=0.2692, pruned_loss=0.04383, over 16832.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.285, pruned_loss=0.0546, over 3046638.36 frames. ], batch size: 76, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:40:33,023 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:41:09,768 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 22:41:50,474 INFO [train.py:904] (2/8) Epoch 8, batch 8750, loss[loss=0.2035, simple_loss=0.2822, pruned_loss=0.06241, over 12639.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2842, pruned_loss=0.05415, over 3029738.88 frames. ], batch size: 250, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:42:11,412 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3550, 3.5292, 1.7663, 3.8153, 2.4090, 3.7800, 2.0293, 2.6780], device='cuda:2'), covar=tensor([0.0240, 0.0318, 0.1800, 0.0120, 0.0954, 0.0453, 0.1684, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0151, 0.0176, 0.0099, 0.0160, 0.0187, 0.0187, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 22:42:15,366 INFO [optim.py:368] (2/8) 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 22:42:43,973 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:42:49,619 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2834, 3.8710, 3.8069, 2.0908, 3.0482, 2.4752, 3.5896, 3.8354], device='cuda:2'), covar=tensor([0.0230, 0.0490, 0.0420, 0.1693, 0.0670, 0.0904, 0.0716, 0.0713], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0128, 0.0150, 0.0138, 0.0130, 0.0123, 0.0133, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-28 22:43:24,438 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 22:43:27,982 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1510, 3.1417, 3.1285, 1.6766, 3.3590, 3.4078, 2.7707, 2.5182], device='cuda:2'), covar=tensor([0.0697, 0.0166, 0.0182, 0.1128, 0.0058, 0.0084, 0.0338, 0.0450], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0091, 0.0078, 0.0133, 0.0062, 0.0084, 0.0112, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 22:43:36,339 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1213, 1.3698, 1.8144, 2.1001, 2.1910, 2.0986, 1.7223, 2.1798], device='cuda:2'), covar=tensor([0.0156, 0.0287, 0.0189, 0.0151, 0.0174, 0.0148, 0.0288, 0.0063], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0157, 0.0141, 0.0137, 0.0145, 0.0103, 0.0154, 0.0094], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 22:43:44,532 INFO [train.py:904] (2/8) Epoch 8, batch 8800, loss[loss=0.1962, simple_loss=0.2855, pruned_loss=0.05341, over 16264.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2833, pruned_loss=0.05313, over 3060594.12 frames. ], batch size: 166, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:45:22,361 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1713, 3.8363, 3.7564, 2.0920, 2.9059, 2.4565, 3.5146, 3.8717], device='cuda:2'), covar=tensor([0.0326, 0.0619, 0.0462, 0.1671, 0.0793, 0.0963, 0.0731, 0.0750], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0128, 0.0151, 0.0139, 0.0130, 0.0123, 0.0133, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 22:45:31,357 INFO [train.py:904] (2/8) Epoch 8, batch 8850, loss[loss=0.1852, simple_loss=0.2805, pruned_loss=0.04489, over 12282.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2852, pruned_loss=0.05275, over 3043232.44 frames. ], batch size: 246, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:45:51,485 INFO [optim.py:368] (2/8) 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 22:46:35,124 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9645, 2.2020, 1.9030, 1.9606, 2.6161, 2.2469, 2.8980, 2.8440], device='cuda:2'), covar=tensor([0.0058, 0.0280, 0.0299, 0.0306, 0.0142, 0.0259, 0.0111, 0.0143], device='cuda:2'), in_proj_covar=tensor([0.0106, 0.0180, 0.0177, 0.0178, 0.0175, 0.0179, 0.0168, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:46:57,814 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:46:57,839 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:47:20,944 INFO [train.py:904] (2/8) Epoch 8, batch 8900, loss[loss=0.1958, simple_loss=0.2739, pruned_loss=0.05884, over 12335.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.285, pruned_loss=0.05184, over 3042911.29 frames. ], batch size: 248, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:47:26,828 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5467, 3.5492, 2.6961, 2.1797, 2.2990, 2.1909, 3.6737, 3.2721], device='cuda:2'), covar=tensor([0.2404, 0.0657, 0.1450, 0.2042, 0.2129, 0.1729, 0.0411, 0.0789], device='cuda:2'), in_proj_covar=tensor([0.0283, 0.0245, 0.0268, 0.0254, 0.0254, 0.0205, 0.0249, 0.0261], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:47:46,481 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1656, 3.4287, 3.4296, 2.2970, 3.2104, 3.4544, 3.2930, 1.9229], device='cuda:2'), covar=tensor([0.0364, 0.0027, 0.0033, 0.0290, 0.0051, 0.0056, 0.0046, 0.0363], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0060, 0.0062, 0.0118, 0.0068, 0.0077, 0.0068, 0.0114], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 22:48:54,504 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:48:54,593 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:49:29,483 INFO [train.py:904] (2/8) Epoch 8, batch 8950, loss[loss=0.1551, simple_loss=0.2511, pruned_loss=0.02957, over 16265.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2848, pruned_loss=0.05213, over 3056634.88 frames. ], batch size: 165, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:49:50,485 INFO [optim.py:368] (2/8) 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 22:49:54,189 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:08,434 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:32,351 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:50:46,184 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:51:17,201 INFO [train.py:904] (2/8) Epoch 8, batch 9000, loss[loss=0.193, simple_loss=0.2713, pruned_loss=0.05735, over 12337.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2816, pruned_loss=0.05076, over 3043493.12 frames. ], batch size: 247, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:51:17,202 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 22:51:27,528 INFO [train.py:938] (2/8) Epoch 8, validation: loss=0.1608, simple_loss=0.265, pruned_loss=0.02828, over 944034.00 frames. 2023-04-28 22:51:27,528 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-28 22:52:04,336 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:52:04,744 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-28 22:52:26,510 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:53:14,309 INFO [train.py:904] (2/8) Epoch 8, batch 9050, loss[loss=0.1897, simple_loss=0.2755, pruned_loss=0.05191, over 16402.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2827, pruned_loss=0.05131, over 3039274.78 frames. ], batch size: 146, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:53:35,357 INFO [optim.py:368] (2/8) 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 22:53:48,932 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:54:59,313 INFO [train.py:904] (2/8) Epoch 8, batch 9100, loss[loss=0.1923, simple_loss=0.2882, pruned_loss=0.04818, over 15352.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2825, pruned_loss=0.05192, over 3029070.11 frames. ], batch size: 191, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:56:11,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5551, 5.8416, 5.5815, 5.6129, 5.2053, 5.1455, 5.2920, 5.9089], device='cuda:2'), covar=tensor([0.0807, 0.0743, 0.0940, 0.0528, 0.0676, 0.0547, 0.0770, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0451, 0.0571, 0.0483, 0.0392, 0.0355, 0.0380, 0.0477, 0.0425], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:56:59,464 INFO [train.py:904] (2/8) Epoch 8, batch 9150, loss[loss=0.1763, simple_loss=0.2682, pruned_loss=0.04219, over 16769.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2824, pruned_loss=0.05121, over 3033366.43 frames. ], batch size: 76, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:57:17,567 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-04-28 22:57:20,184 INFO [optim.py:368] (2/8) 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 22:58:07,874 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-04-28 22:58:26,798 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:58:28,290 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6975, 2.4729, 2.3037, 3.4577, 2.5601, 3.6576, 1.4320, 2.9359], device='cuda:2'), covar=tensor([0.1324, 0.0635, 0.1079, 0.0092, 0.0127, 0.0408, 0.1443, 0.0645], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0149, 0.0171, 0.0114, 0.0182, 0.0199, 0.0170, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-28 22:58:44,954 INFO [train.py:904] (2/8) Epoch 8, batch 9200, loss[loss=0.1529, simple_loss=0.2381, pruned_loss=0.03381, over 12045.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2771, pruned_loss=0.04928, over 3051775.59 frames. ], batch size: 248, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:58:51,014 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2774, 5.2717, 4.9891, 4.5510, 5.1298, 1.7678, 4.8540, 4.9995], device='cuda:2'), covar=tensor([0.0043, 0.0043, 0.0112, 0.0217, 0.0053, 0.2101, 0.0091, 0.0102], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0094, 0.0139, 0.0130, 0.0111, 0.0163, 0.0126, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 22:59:42,581 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:59:55,914 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:00:22,118 INFO [train.py:904] (2/8) Epoch 8, batch 9250, loss[loss=0.179, simple_loss=0.2518, pruned_loss=0.05311, over 12611.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2777, pruned_loss=0.04957, over 3051530.86 frames. ], batch size: 248, lr: 8.34e-03, grad_scale: 4.0 2023-04-28 23:00:30,619 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5051, 4.3113, 4.5719, 4.7203, 4.8280, 4.3161, 4.8135, 4.8050], device='cuda:2'), covar=tensor([0.1300, 0.0954, 0.1258, 0.0523, 0.0465, 0.0845, 0.0445, 0.0470], device='cuda:2'), in_proj_covar=tensor([0.0437, 0.0537, 0.0661, 0.0550, 0.0419, 0.0413, 0.0435, 0.0477], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:00:38,843 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 23:00:42,872 INFO [optim.py:368] (2/8) 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 23:00:44,518 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:01:56,375 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:02:12,075 INFO [train.py:904] (2/8) Epoch 8, batch 9300, loss[loss=0.1593, simple_loss=0.2503, pruned_loss=0.0342, over 16562.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2759, pruned_loss=0.0486, over 3067154.30 frames. ], batch size: 62, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:02:23,198 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5456, 3.5418, 3.5281, 3.0919, 3.4956, 1.9311, 3.2714, 3.0666], device='cuda:2'), covar=tensor([0.0118, 0.0110, 0.0133, 0.0253, 0.0091, 0.1956, 0.0123, 0.0188], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0095, 0.0138, 0.0130, 0.0110, 0.0162, 0.0126, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:02:31,762 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:02:57,508 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.24 vs. limit=5.0 2023-04-28 23:03:42,288 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:03:57,479 INFO [train.py:904] (2/8) Epoch 8, batch 9350, loss[loss=0.203, simple_loss=0.2921, pruned_loss=0.05697, over 16232.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2758, pruned_loss=0.04862, over 3066998.62 frames. ], batch size: 165, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:04:12,347 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-28 23:04:22,276 INFO [optim.py:368] (2/8) 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 23:04:34,195 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:05:40,910 INFO [train.py:904] (2/8) Epoch 8, batch 9400, loss[loss=0.15, simple_loss=0.2417, pruned_loss=0.02917, over 12504.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2762, pruned_loss=0.04853, over 3064647.19 frames. ], batch size: 247, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:05:46,253 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:05,125 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:09,457 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:32,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 23:07:07,254 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:07:19,862 INFO [train.py:904] (2/8) Epoch 8, batch 9450, loss[loss=0.204, simple_loss=0.2879, pruned_loss=0.06005, over 16898.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2765, pruned_loss=0.04853, over 3030283.23 frames. ], batch size: 116, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:07:38,810 INFO [optim.py:368] (2/8) 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 23:08:04,368 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2982, 3.4130, 1.8058, 3.6493, 2.4299, 3.5848, 1.9962, 2.7099], device='cuda:2'), covar=tensor([0.0194, 0.0310, 0.1636, 0.0114, 0.0890, 0.0471, 0.1578, 0.0674], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0151, 0.0179, 0.0100, 0.0161, 0.0185, 0.0189, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-28 23:08:04,387 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:08:58,865 INFO [train.py:904] (2/8) Epoch 8, batch 9500, loss[loss=0.1866, simple_loss=0.2759, pruned_loss=0.04862, over 16368.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2755, pruned_loss=0.04782, over 3054408.70 frames. ], batch size: 146, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:09:08,332 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:09:48,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9671, 5.3043, 5.0933, 5.0204, 4.7045, 4.5086, 4.7013, 5.3547], device='cuda:2'), covar=tensor([0.0811, 0.0820, 0.0863, 0.0596, 0.0719, 0.0862, 0.0884, 0.0735], device='cuda:2'), in_proj_covar=tensor([0.0447, 0.0567, 0.0471, 0.0387, 0.0352, 0.0377, 0.0472, 0.0422], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:10:46,570 INFO [train.py:904] (2/8) Epoch 8, batch 9550, loss[loss=0.2058, simple_loss=0.3023, pruned_loss=0.05464, over 15384.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2762, pruned_loss=0.04855, over 3061913.78 frames. ], batch size: 190, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:11:10,121 INFO [optim.py:368] (2/8) 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 23:12:03,960 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:12:27,095 INFO [train.py:904] (2/8) Epoch 8, batch 9600, loss[loss=0.2, simple_loss=0.2993, pruned_loss=0.05031, over 15325.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.278, pruned_loss=0.04945, over 3044286.65 frames. ], batch size: 191, lr: 8.32e-03, grad_scale: 8.0 2023-04-28 23:12:58,338 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1883, 4.1658, 4.3628, 4.2822, 4.2531, 4.7736, 4.3358, 4.0018], device='cuda:2'), covar=tensor([0.1424, 0.1790, 0.1529, 0.1882, 0.2384, 0.0986, 0.1364, 0.2484], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0418, 0.0446, 0.0370, 0.0484, 0.0464, 0.0361, 0.0486], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:14:15,050 INFO [train.py:904] (2/8) Epoch 8, batch 9650, loss[loss=0.1954, simple_loss=0.2942, pruned_loss=0.04823, over 15414.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2802, pruned_loss=0.04959, over 3048449.07 frames. ], batch size: 191, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:14:42,862 INFO [optim.py:368] (2/8) 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 23:15:05,713 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2107, 3.2688, 3.6184, 3.5930, 3.6098, 3.3405, 3.4086, 3.4531], device='cuda:2'), covar=tensor([0.0405, 0.0758, 0.0482, 0.0480, 0.0574, 0.0531, 0.0841, 0.0414], device='cuda:2'), in_proj_covar=tensor([0.0273, 0.0273, 0.0277, 0.0272, 0.0313, 0.0295, 0.0380, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 23:15:19,433 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8477, 4.7984, 4.6479, 4.1477, 4.6561, 1.7603, 4.4634, 4.6241], device='cuda:2'), covar=tensor([0.0068, 0.0065, 0.0116, 0.0250, 0.0073, 0.2002, 0.0105, 0.0148], device='cuda:2'), in_proj_covar=tensor([0.0104, 0.0093, 0.0135, 0.0126, 0.0108, 0.0160, 0.0122, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-28 23:15:30,024 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:15:59,671 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:16:03,264 INFO [train.py:904] (2/8) Epoch 8, batch 9700, loss[loss=0.192, simple_loss=0.2827, pruned_loss=0.05061, over 16963.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.279, pruned_loss=0.04914, over 3056978.73 frames. ], batch size: 109, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:16:46,522 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:38,614 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:46,348 INFO [train.py:904] (2/8) Epoch 8, batch 9750, loss[loss=0.2076, simple_loss=0.2773, pruned_loss=0.06892, over 12330.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2782, pruned_loss=0.04966, over 3051277.41 frames. ], batch size: 247, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:18:08,279 INFO [optim.py:368] (2/8) 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 23:18:20,510 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:18:47,421 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9902, 4.0641, 3.8270, 3.7082, 3.5968, 3.9557, 3.6841, 3.7476], device='cuda:2'), covar=tensor([0.0531, 0.0398, 0.0245, 0.0214, 0.0688, 0.0344, 0.0884, 0.0472], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0236, 0.0238, 0.0210, 0.0254, 0.0242, 0.0164, 0.0272], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:18:55,689 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:24,255 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:26,323 INFO [train.py:904] (2/8) Epoch 8, batch 9800, loss[loss=0.2032, simple_loss=0.3053, pruned_loss=0.05059, over 16473.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2783, pruned_loss=0.04848, over 3076864.52 frames. ], batch size: 147, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:20:07,433 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-28 23:21:11,973 INFO [train.py:904] (2/8) Epoch 8, batch 9850, loss[loss=0.1801, simple_loss=0.2726, pruned_loss=0.04375, over 16738.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2791, pruned_loss=0.04804, over 3084905.58 frames. ], batch size: 134, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:21:33,315 INFO [optim.py:368] (2/8) 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 23:22:37,375 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:22:53,315 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3367, 3.4693, 1.7389, 3.7965, 2.4717, 3.7304, 1.9359, 2.7012], device='cuda:2'), covar=tensor([0.0230, 0.0388, 0.1869, 0.0111, 0.0892, 0.0535, 0.1710, 0.0742], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0153, 0.0184, 0.0101, 0.0163, 0.0187, 0.0192, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 23:23:04,040 INFO [train.py:904] (2/8) Epoch 8, batch 9900, loss[loss=0.1988, simple_loss=0.2883, pruned_loss=0.05462, over 15316.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.28, pruned_loss=0.04841, over 3083852.15 frames. ], batch size: 191, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:23:39,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1309, 3.0506, 3.1316, 1.7653, 3.3453, 3.4072, 2.7062, 2.6456], device='cuda:2'), covar=tensor([0.0712, 0.0175, 0.0150, 0.1112, 0.0063, 0.0085, 0.0405, 0.0377], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0092, 0.0077, 0.0135, 0.0063, 0.0083, 0.0112, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 23:24:29,640 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:25:03,327 INFO [train.py:904] (2/8) Epoch 8, batch 9950, loss[loss=0.183, simple_loss=0.2811, pruned_loss=0.04248, over 16210.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2817, pruned_loss=0.04867, over 3068548.22 frames. ], batch size: 165, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:25:29,444 INFO [optim.py:368] (2/8) 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 23:26:16,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4838, 4.3944, 4.8994, 4.8093, 4.8372, 4.5070, 4.5271, 4.3550], device='cuda:2'), covar=tensor([0.0230, 0.0613, 0.0335, 0.0412, 0.0370, 0.0322, 0.0683, 0.0358], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0269, 0.0273, 0.0269, 0.0309, 0.0290, 0.0373, 0.0235], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 23:27:01,979 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:27:07,185 INFO [train.py:904] (2/8) Epoch 8, batch 10000, loss[loss=0.1681, simple_loss=0.2671, pruned_loss=0.0345, over 16763.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2803, pruned_loss=0.04836, over 3064629.67 frames. ], batch size: 76, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:27:53,013 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3745, 5.7809, 5.5344, 5.5817, 5.0123, 5.0452, 5.1777, 5.8390], device='cuda:2'), covar=tensor([0.0944, 0.0828, 0.0816, 0.0525, 0.0868, 0.0683, 0.0817, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0454, 0.0581, 0.0476, 0.0394, 0.0357, 0.0383, 0.0476, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:28:14,357 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7010, 1.5823, 2.0782, 2.6815, 2.4504, 2.7863, 2.0598, 2.7600], device='cuda:2'), covar=tensor([0.0111, 0.0324, 0.0206, 0.0157, 0.0177, 0.0123, 0.0269, 0.0091], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0156, 0.0142, 0.0138, 0.0146, 0.0103, 0.0155, 0.0093], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-28 23:28:30,671 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:28:47,984 INFO [train.py:904] (2/8) Epoch 8, batch 10050, loss[loss=0.1892, simple_loss=0.2788, pruned_loss=0.04976, over 12017.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2802, pruned_loss=0.04807, over 3064398.26 frames. ], batch size: 248, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:29:08,274 INFO [optim.py:368] (2/8) 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 23:29:15,162 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:21,232 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:39,801 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:30:02,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3637, 4.3791, 4.8435, 4.7223, 4.7707, 4.3852, 4.3793, 4.2059], device='cuda:2'), covar=tensor([0.0264, 0.0529, 0.0287, 0.0454, 0.0382, 0.0363, 0.0775, 0.0458], device='cuda:2'), in_proj_covar=tensor([0.0272, 0.0273, 0.0275, 0.0273, 0.0311, 0.0294, 0.0377, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-28 23:30:18,846 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:30:21,530 INFO [train.py:904] (2/8) Epoch 8, batch 10100, loss[loss=0.1889, simple_loss=0.263, pruned_loss=0.05739, over 12325.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2803, pruned_loss=0.0481, over 3073260.47 frames. ], batch size: 248, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:30:44,839 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 23:30:51,853 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:14,346 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:27,362 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1204, 4.0954, 3.9595, 3.3619, 3.9879, 1.6521, 3.7813, 3.7148], device='cuda:2'), covar=tensor([0.0082, 0.0079, 0.0142, 0.0300, 0.0090, 0.2256, 0.0125, 0.0202], device='cuda:2'), in_proj_covar=tensor([0.0107, 0.0095, 0.0138, 0.0128, 0.0111, 0.0165, 0.0124, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-28 23:31:37,453 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:39,853 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 23:32:08,878 INFO [train.py:904] (2/8) Epoch 9, batch 0, loss[loss=0.3069, simple_loss=0.3408, pruned_loss=0.1365, over 16487.00 frames. ], tot_loss[loss=0.3069, simple_loss=0.3408, pruned_loss=0.1365, over 16487.00 frames. ], batch size: 75, lr: 7.85e-03, grad_scale: 8.0 2023-04-28 23:32:08,879 INFO [train.py:929] (2/8) Computing validation loss 2023-04-28 23:32:16,257 INFO [train.py:938] (2/8) Epoch 9, validation: loss=0.1602, simple_loss=0.2637, pruned_loss=0.02837, over 944034.00 frames. 2023-04-28 23:32:16,258 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-28 23:32:33,669 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4915, 4.3840, 4.2935, 4.1411, 3.8025, 4.4726, 4.5086, 4.0461], device='cuda:2'), covar=tensor([0.0774, 0.0714, 0.0481, 0.0394, 0.1233, 0.0557, 0.0393, 0.0726], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0234, 0.0234, 0.0207, 0.0252, 0.0237, 0.0161, 0.0265], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-28 23:32:36,784 INFO [optim.py:368] (2/8) 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 23:33:25,084 INFO [train.py:904] (2/8) Epoch 9, batch 50, loss[loss=0.2029, simple_loss=0.2737, pruned_loss=0.06607, over 16706.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3023, pruned_loss=0.07703, over 745890.45 frames. ], batch size: 124, lr: 7.85e-03, grad_scale: 1.0 2023-04-28 23:34:31,196 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:34,160 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:35,495 INFO [train.py:904] (2/8) Epoch 9, batch 100, loss[loss=0.1891, simple_loss=0.2678, pruned_loss=0.05521, over 17209.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2928, pruned_loss=0.06877, over 1326752.24 frames. ], batch size: 44, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:34:54,536 INFO [optim.py:368] (2/8) 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 23:35:13,929 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4995, 5.9312, 5.6244, 5.6791, 5.2118, 5.1538, 5.3416, 6.0303], device='cuda:2'), covar=tensor([0.1089, 0.0925, 0.0992, 0.0611, 0.0805, 0.0723, 0.0886, 0.0841], device='cuda:2'), in_proj_covar=tensor([0.0480, 0.0611, 0.0507, 0.0416, 0.0376, 0.0401, 0.0506, 0.0451], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:35:16,980 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8461, 3.9484, 3.6915, 3.5946, 3.3015, 3.8432, 3.5579, 3.5319], device='cuda:2'), covar=tensor([0.0680, 0.0504, 0.0325, 0.0277, 0.0812, 0.0404, 0.0927, 0.0525], device='cuda:2'), in_proj_covar=tensor([0.0212, 0.0248, 0.0247, 0.0219, 0.0267, 0.0251, 0.0169, 0.0283], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:35:42,949 INFO [train.py:904] (2/8) Epoch 9, batch 150, loss[loss=0.1924, simple_loss=0.2697, pruned_loss=0.05753, over 16722.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2887, pruned_loss=0.06595, over 1764619.22 frames. ], batch size: 89, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:35:55,905 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:35:58,270 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:18,724 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 23:36:36,245 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.69 vs. limit=5.0 2023-04-28 23:36:40,002 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:54,435 INFO [train.py:904] (2/8) Epoch 9, batch 200, loss[loss=0.2445, simple_loss=0.2935, pruned_loss=0.09773, over 16859.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2877, pruned_loss=0.0653, over 2107380.12 frames. ], batch size: 96, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:37:13,043 INFO [optim.py:368] (2/8) 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 23:37:28,368 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3293, 5.3396, 5.1233, 4.7519, 5.1833, 2.0992, 4.9258, 5.0931], device='cuda:2'), covar=tensor([0.0056, 0.0051, 0.0115, 0.0246, 0.0060, 0.1959, 0.0089, 0.0145], device='cuda:2'), in_proj_covar=tensor([0.0113, 0.0100, 0.0147, 0.0137, 0.0117, 0.0171, 0.0132, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:37:31,276 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:37:45,245 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:01,970 INFO [train.py:904] (2/8) Epoch 9, batch 250, loss[loss=0.1722, simple_loss=0.257, pruned_loss=0.04373, over 17222.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2842, pruned_loss=0.06266, over 2376796.18 frames. ], batch size: 43, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:38:14,391 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6027, 3.7621, 3.9545, 2.1970, 4.1128, 4.1780, 3.1352, 2.9651], device='cuda:2'), covar=tensor([0.0709, 0.0142, 0.0147, 0.0966, 0.0058, 0.0108, 0.0343, 0.0396], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0094, 0.0081, 0.0140, 0.0067, 0.0090, 0.0117, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 23:38:29,683 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:36,874 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:46,987 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4507, 3.4345, 3.5446, 2.0276, 3.7600, 3.7971, 3.0301, 2.8230], device='cuda:2'), covar=tensor([0.0698, 0.0164, 0.0181, 0.1075, 0.0063, 0.0127, 0.0345, 0.0415], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0095, 0.0081, 0.0140, 0.0067, 0.0090, 0.0117, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-28 23:38:48,172 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2238, 4.5914, 4.7231, 3.5741, 4.1583, 4.5890, 4.1286, 3.0066], device='cuda:2'), covar=tensor([0.0292, 0.0054, 0.0023, 0.0205, 0.0042, 0.0049, 0.0040, 0.0265], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0065, 0.0065, 0.0121, 0.0069, 0.0079, 0.0070, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 23:39:04,716 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2280, 3.3894, 3.6126, 3.5865, 3.6200, 3.4014, 3.4314, 3.4603], device='cuda:2'), covar=tensor([0.0357, 0.0517, 0.0422, 0.0452, 0.0409, 0.0404, 0.0644, 0.0416], device='cuda:2'), in_proj_covar=tensor([0.0295, 0.0297, 0.0299, 0.0293, 0.0334, 0.0317, 0.0407, 0.0258], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 23:39:10,399 INFO [train.py:904] (2/8) Epoch 9, batch 300, loss[loss=0.1694, simple_loss=0.2566, pruned_loss=0.04114, over 17201.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2814, pruned_loss=0.06157, over 2590915.27 frames. ], batch size: 44, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:39:21,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5504, 2.4693, 1.9720, 2.1988, 2.8900, 2.6408, 3.4247, 3.1775], device='cuda:2'), covar=tensor([0.0051, 0.0278, 0.0317, 0.0286, 0.0163, 0.0246, 0.0138, 0.0139], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0189, 0.0184, 0.0185, 0.0183, 0.0187, 0.0183, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:39:29,676 INFO [optim.py:368] (2/8) 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 23:40:16,173 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6012, 2.6787, 2.1659, 2.4587, 3.0134, 2.8843, 3.4929, 3.2363], device='cuda:2'), covar=tensor([0.0059, 0.0229, 0.0308, 0.0282, 0.0156, 0.0223, 0.0160, 0.0153], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0189, 0.0184, 0.0185, 0.0183, 0.0188, 0.0183, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:40:17,833 INFO [train.py:904] (2/8) Epoch 9, batch 350, loss[loss=0.1872, simple_loss=0.2735, pruned_loss=0.05045, over 16703.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2787, pruned_loss=0.05914, over 2756876.45 frames. ], batch size: 57, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:41:23,824 INFO [train.py:904] (2/8) Epoch 9, batch 400, loss[loss=0.1835, simple_loss=0.2734, pruned_loss=0.04674, over 17126.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2767, pruned_loss=0.05881, over 2886760.04 frames. ], batch size: 48, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:41:43,586 INFO [optim.py:368] (2/8) 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 23:41:52,362 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7193, 6.1329, 5.8017, 5.8966, 5.5002, 5.3304, 5.5632, 6.2255], device='cuda:2'), covar=tensor([0.0996, 0.0777, 0.1120, 0.0659, 0.0784, 0.0633, 0.0813, 0.0791], device='cuda:2'), in_proj_covar=tensor([0.0492, 0.0631, 0.0523, 0.0430, 0.0390, 0.0410, 0.0522, 0.0466], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-28 23:42:33,320 INFO [train.py:904] (2/8) Epoch 9, batch 450, loss[loss=0.2189, simple_loss=0.3027, pruned_loss=0.0676, over 16765.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2745, pruned_loss=0.05748, over 2982861.91 frames. ], batch size: 62, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:42:37,319 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:40,763 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:58,118 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:43:21,580 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0171, 5.0719, 5.6648, 5.5386, 5.5682, 5.1970, 5.1103, 4.9004], device='cuda:2'), covar=tensor([0.0274, 0.0335, 0.0265, 0.0417, 0.0409, 0.0254, 0.0792, 0.0313], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0306, 0.0306, 0.0299, 0.0343, 0.0325, 0.0419, 0.0262], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-28 23:43:42,339 INFO [train.py:904] (2/8) Epoch 9, batch 500, loss[loss=0.1731, simple_loss=0.2553, pruned_loss=0.04546, over 16825.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2734, pruned_loss=0.05671, over 3050514.57 frames. ], batch size: 39, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:44:01,900 INFO [optim.py:368] (2/8) 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 23:44:22,649 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 23:44:42,296 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 23:44:45,509 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 23:44:50,825 INFO [train.py:904] (2/8) Epoch 9, batch 550, loss[loss=0.2117, simple_loss=0.2794, pruned_loss=0.07198, over 16406.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2727, pruned_loss=0.05611, over 3112293.37 frames. ], batch size: 146, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:45:19,862 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:45:59,697 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 23:46:02,609 INFO [train.py:904] (2/8) Epoch 9, batch 600, loss[loss=0.2119, simple_loss=0.2729, pruned_loss=0.07542, over 16898.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2722, pruned_loss=0.05688, over 3154354.27 frames. ], batch size: 109, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:46:21,575 INFO [optim.py:368] (2/8) 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 23:46:26,575 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:47:09,564 INFO [train.py:904] (2/8) Epoch 9, batch 650, loss[loss=0.1657, simple_loss=0.2425, pruned_loss=0.0445, over 17029.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2703, pruned_loss=0.05619, over 3196652.14 frames. ], batch size: 41, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:48:18,160 INFO [train.py:904] (2/8) Epoch 9, batch 700, loss[loss=0.1892, simple_loss=0.2791, pruned_loss=0.04968, over 16588.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2698, pruned_loss=0.05575, over 3222572.96 frames. ], batch size: 62, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:48:37,203 INFO [optim.py:368] (2/8) 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 23:49:25,153 INFO [train.py:904] (2/8) Epoch 9, batch 750, loss[loss=0.2038, simple_loss=0.2731, pruned_loss=0.06725, over 12541.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2711, pruned_loss=0.05654, over 3245505.10 frames. ], batch size: 246, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:49:29,042 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:49:31,581 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:14,420 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:37,832 INFO [train.py:904] (2/8) Epoch 9, batch 800, loss[loss=0.1806, simple_loss=0.2682, pruned_loss=0.0465, over 17126.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2699, pruned_loss=0.0563, over 3259097.98 frames. ], batch size: 49, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:50:39,915 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:56,930 INFO [optim.py:368] (2/8) 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 23:51:10,864 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 23:51:40,268 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:51:45,036 INFO [train.py:904] (2/8) Epoch 9, batch 850, loss[loss=0.1867, simple_loss=0.2615, pruned_loss=0.05591, over 16865.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2696, pruned_loss=0.05538, over 3277456.23 frames. ], batch size: 96, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:51:49,652 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:52:14,605 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9789, 3.9774, 3.7984, 3.5900, 3.4850, 3.9003, 3.6003, 3.6558], device='cuda:2'), covar=tensor([0.0640, 0.0500, 0.0291, 0.0299, 0.0884, 0.0451, 0.1008, 0.0680], device='cuda:2'), in_proj_covar=tensor([0.0236, 0.0278, 0.0274, 0.0246, 0.0301, 0.0282, 0.0189, 0.0318], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 23:52:54,644 INFO [train.py:904] (2/8) Epoch 9, batch 900, loss[loss=0.2115, simple_loss=0.2781, pruned_loss=0.07242, over 15498.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2684, pruned_loss=0.05441, over 3282450.51 frames. ], batch size: 190, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:53:07,678 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9393, 3.2867, 3.1418, 2.0416, 2.7796, 2.3987, 3.3763, 3.3930], device='cuda:2'), covar=tensor([0.0244, 0.0664, 0.0524, 0.1500, 0.0715, 0.0819, 0.0524, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0136, 0.0156, 0.0142, 0.0135, 0.0125, 0.0134, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-28 23:53:13,855 INFO [optim.py:368] (2/8) 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 23:53:14,352 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:53:17,458 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-04-28 23:54:03,810 INFO [train.py:904] (2/8) Epoch 9, batch 950, loss[loss=0.1848, simple_loss=0.2542, pruned_loss=0.05775, over 16959.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2689, pruned_loss=0.05478, over 3290619.05 frames. ], batch size: 41, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:11,106 INFO [train.py:904] (2/8) Epoch 9, batch 1000, loss[loss=0.1989, simple_loss=0.2871, pruned_loss=0.05535, over 16768.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2681, pruned_loss=0.05482, over 3305169.91 frames. ], batch size: 57, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:31,951 INFO [optim.py:368] (2/8) 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 23:56:20,445 INFO [train.py:904] (2/8) Epoch 9, batch 1050, loss[loss=0.1911, simple_loss=0.266, pruned_loss=0.05809, over 16336.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2681, pruned_loss=0.05462, over 3314509.37 frames. ], batch size: 165, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:09,441 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 23:57:21,035 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 23:57:28,537 INFO [train.py:904] (2/8) Epoch 9, batch 1100, loss[loss=0.1939, simple_loss=0.2882, pruned_loss=0.04984, over 17089.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.268, pruned_loss=0.05459, over 3319210.57 frames. ], batch size: 55, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:47,216 INFO [optim.py:368] (2/8) 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 23:57:49,650 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2860, 5.2167, 5.0934, 4.6836, 4.6354, 5.0955, 5.1561, 4.7510], device='cuda:2'), covar=tensor([0.0493, 0.0362, 0.0234, 0.0261, 0.1096, 0.0395, 0.0262, 0.0618], device='cuda:2'), in_proj_covar=tensor([0.0240, 0.0284, 0.0279, 0.0251, 0.0307, 0.0288, 0.0192, 0.0322], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-28 23:58:01,926 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:15,342 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 23:58:24,254 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:24,433 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8028, 4.3500, 4.4253, 3.0403, 3.8308, 4.2775, 3.8475, 2.5433], device='cuda:2'), covar=tensor([0.0320, 0.0026, 0.0025, 0.0251, 0.0056, 0.0076, 0.0056, 0.0300], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0066, 0.0065, 0.0120, 0.0070, 0.0080, 0.0071, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-28 23:58:28,053 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:35,308 INFO [train.py:904] (2/8) Epoch 9, batch 1150, loss[loss=0.1595, simple_loss=0.2492, pruned_loss=0.03484, over 17058.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2664, pruned_loss=0.05309, over 3327601.02 frames. ], batch size: 50, lr: 7.79e-03, grad_scale: 4.0 2023-04-28 23:59:04,896 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:44,528 INFO [train.py:904] (2/8) Epoch 9, batch 1200, loss[loss=0.1994, simple_loss=0.2736, pruned_loss=0.06258, over 15421.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2659, pruned_loss=0.05284, over 3318769.71 frames. ], batch size: 190, lr: 7.79e-03, grad_scale: 8.0 2023-04-28 23:59:45,383 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 23:59:50,717 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:55,207 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:00:02,692 INFO [optim.py:368] (2/8) 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 00:00:50,244 INFO [train.py:904] (2/8) Epoch 9, batch 1250, loss[loss=0.2144, simple_loss=0.303, pruned_loss=0.06294, over 17049.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.267, pruned_loss=0.05367, over 3318437.31 frames. ], batch size: 55, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:01:47,051 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:01:58,479 INFO [train.py:904] (2/8) Epoch 9, batch 1300, loss[loss=0.205, simple_loss=0.2749, pruned_loss=0.06754, over 16521.00 frames. ], tot_loss[loss=0.187, simple_loss=0.266, pruned_loss=0.05402, over 3310406.56 frames. ], batch size: 75, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:02:18,040 INFO [optim.py:368] (2/8) 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 00:02:46,490 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2699, 2.0776, 1.6264, 1.8842, 2.5002, 2.3005, 2.4850, 2.6012], device='cuda:2'), covar=tensor([0.0118, 0.0246, 0.0331, 0.0298, 0.0125, 0.0194, 0.0152, 0.0157], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0193, 0.0187, 0.0187, 0.0188, 0.0191, 0.0194, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:03:03,174 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1522, 4.1928, 4.6128, 4.5719, 4.6085, 4.2573, 4.3171, 4.1652], device='cuda:2'), covar=tensor([0.0276, 0.0477, 0.0346, 0.0417, 0.0353, 0.0306, 0.0674, 0.0465], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0319, 0.0322, 0.0308, 0.0361, 0.0338, 0.0438, 0.0272], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 00:03:05,221 INFO [train.py:904] (2/8) Epoch 9, batch 1350, loss[loss=0.2228, simple_loss=0.2856, pruned_loss=0.08006, over 16831.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.267, pruned_loss=0.05395, over 3309487.74 frames. ], batch size: 96, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:03:08,007 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:03:29,817 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:04:12,576 INFO [train.py:904] (2/8) Epoch 9, batch 1400, loss[loss=0.2007, simple_loss=0.2691, pruned_loss=0.06613, over 11963.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2664, pruned_loss=0.05331, over 3306260.68 frames. ], batch size: 246, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:04:31,669 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3624, 5.3056, 5.1195, 4.5929, 5.0821, 2.1736, 4.8845, 5.2283], device='cuda:2'), covar=tensor([0.0054, 0.0047, 0.0111, 0.0304, 0.0059, 0.1922, 0.0099, 0.0118], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0109, 0.0159, 0.0153, 0.0128, 0.0176, 0.0143, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:04:33,568 INFO [optim.py:368] (2/8) 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 00:04:37,620 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8018, 4.0271, 2.0382, 4.5357, 2.8043, 4.4717, 2.1490, 3.1323], device='cuda:2'), covar=tensor([0.0205, 0.0268, 0.1608, 0.0139, 0.0864, 0.0438, 0.1524, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0164, 0.0186, 0.0117, 0.0166, 0.0206, 0.0196, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 00:04:53,125 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:09,601 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:22,654 INFO [train.py:904] (2/8) Epoch 9, batch 1450, loss[loss=0.1927, simple_loss=0.2605, pruned_loss=0.06249, over 16756.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2657, pruned_loss=0.05357, over 3309362.61 frames. ], batch size: 83, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:05:47,876 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-29 00:06:00,999 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 00:06:04,733 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2418, 4.4942, 2.5057, 4.9271, 3.2894, 4.8561, 2.6595, 3.5284], device='cuda:2'), covar=tensor([0.0150, 0.0188, 0.1306, 0.0101, 0.0639, 0.0276, 0.1245, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0161, 0.0183, 0.0115, 0.0164, 0.0202, 0.0193, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 00:06:16,064 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:29,712 INFO [train.py:904] (2/8) Epoch 9, batch 1500, loss[loss=0.2034, simple_loss=0.2738, pruned_loss=0.06652, over 15495.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2654, pruned_loss=0.05375, over 3309416.52 frames. ], batch size: 190, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:06:29,979 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:42,751 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:51,613 INFO [optim.py:368] (2/8) 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 00:06:52,099 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1994, 4.0254, 4.2415, 4.3878, 4.4761, 4.0146, 4.2601, 4.4374], device='cuda:2'), covar=tensor([0.0996, 0.0832, 0.1056, 0.0494, 0.0488, 0.1171, 0.1410, 0.0543], device='cuda:2'), in_proj_covar=tensor([0.0527, 0.0641, 0.0807, 0.0659, 0.0495, 0.0488, 0.0514, 0.0571], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:07:39,172 INFO [train.py:904] (2/8) Epoch 9, batch 1550, loss[loss=0.1887, simple_loss=0.2745, pruned_loss=0.05143, over 17118.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2674, pruned_loss=0.05583, over 3314223.36 frames. ], batch size: 48, lr: 7.77e-03, grad_scale: 4.0 2023-04-29 00:07:49,785 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:07:58,630 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-29 00:08:48,744 INFO [train.py:904] (2/8) Epoch 9, batch 1600, loss[loss=0.1588, simple_loss=0.2495, pruned_loss=0.03405, over 16846.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2687, pruned_loss=0.05597, over 3306584.19 frames. ], batch size: 42, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:09:09,709 INFO [optim.py:368] (2/8) 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 00:09:52,365 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:09:56,031 INFO [train.py:904] (2/8) Epoch 9, batch 1650, loss[loss=0.1927, simple_loss=0.2662, pruned_loss=0.0596, over 16848.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2704, pruned_loss=0.05633, over 3316402.85 frames. ], batch size: 96, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:10:17,755 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-29 00:10:22,030 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1406, 5.0716, 5.6800, 5.6120, 5.6419, 5.2543, 5.2393, 4.9165], device='cuda:2'), covar=tensor([0.0259, 0.0363, 0.0288, 0.0409, 0.0389, 0.0272, 0.0847, 0.0344], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0317, 0.0322, 0.0304, 0.0364, 0.0337, 0.0437, 0.0271], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 00:11:05,716 INFO [train.py:904] (2/8) Epoch 9, batch 1700, loss[loss=0.1736, simple_loss=0.2642, pruned_loss=0.04148, over 17141.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2711, pruned_loss=0.05556, over 3328136.54 frames. ], batch size: 48, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:11:24,536 INFO [optim.py:368] (2/8) 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 00:11:37,132 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:11:38,475 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5957, 2.3750, 1.8030, 2.1648, 2.7929, 2.6539, 2.9655, 2.9633], device='cuda:2'), covar=tensor([0.0094, 0.0245, 0.0318, 0.0258, 0.0128, 0.0178, 0.0133, 0.0134], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0196, 0.0189, 0.0192, 0.0192, 0.0196, 0.0198, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:12:13,189 INFO [train.py:904] (2/8) Epoch 9, batch 1750, loss[loss=0.1771, simple_loss=0.26, pruned_loss=0.04709, over 17173.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2712, pruned_loss=0.05503, over 3336241.34 frames. ], batch size: 46, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:12:28,850 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:12:59,173 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 00:13:14,107 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 00:13:19,613 INFO [train.py:904] (2/8) Epoch 9, batch 1800, loss[loss=0.187, simple_loss=0.2714, pruned_loss=0.05132, over 16824.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2723, pruned_loss=0.05512, over 3331102.88 frames. ], batch size: 42, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:13:19,975 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:13:40,258 INFO [optim.py:368] (2/8) 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 00:13:50,209 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:19,813 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 00:14:26,662 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:28,692 INFO [train.py:904] (2/8) Epoch 9, batch 1850, loss[loss=0.1933, simple_loss=0.2884, pruned_loss=0.04912, over 17256.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2734, pruned_loss=0.05516, over 3334227.39 frames. ], batch size: 52, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:37,695 INFO [train.py:904] (2/8) Epoch 9, batch 1900, loss[loss=0.1507, simple_loss=0.2377, pruned_loss=0.03186, over 16732.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2727, pruned_loss=0.05407, over 3336776.93 frames. ], batch size: 39, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:59,152 INFO [optim.py:368] (2/8) 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 00:16:42,469 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:16:46,825 INFO [train.py:904] (2/8) Epoch 9, batch 1950, loss[loss=0.1713, simple_loss=0.2562, pruned_loss=0.04326, over 16779.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2729, pruned_loss=0.05404, over 3330709.98 frames. ], batch size: 39, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:16:58,068 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5604, 3.6927, 3.7342, 2.0621, 3.9481, 3.9275, 3.2508, 2.9134], device='cuda:2'), covar=tensor([0.0693, 0.0129, 0.0170, 0.1008, 0.0052, 0.0102, 0.0300, 0.0392], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0098, 0.0087, 0.0142, 0.0070, 0.0098, 0.0120, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 00:17:48,378 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:17:49,290 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:17:55,080 INFO [train.py:904] (2/8) Epoch 9, batch 2000, loss[loss=0.2352, simple_loss=0.2944, pruned_loss=0.08803, over 16449.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2734, pruned_loss=0.05446, over 3324873.33 frames. ], batch size: 75, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:18:17,511 INFO [optim.py:368] (2/8) 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 00:18:26,838 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:04,147 INFO [train.py:904] (2/8) Epoch 9, batch 2050, loss[loss=0.1827, simple_loss=0.2678, pruned_loss=0.04881, over 16833.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2731, pruned_loss=0.0548, over 3329015.24 frames. ], batch size: 42, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:19:04,882 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 00:19:12,731 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:33,894 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:48,900 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:20:13,197 INFO [train.py:904] (2/8) Epoch 9, batch 2100, loss[loss=0.1781, simple_loss=0.2603, pruned_loss=0.04789, over 16820.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2734, pruned_loss=0.05519, over 3332607.61 frames. ], batch size: 42, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:20:30,941 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1187, 5.5324, 5.6649, 5.5120, 5.5333, 6.0937, 5.7211, 5.4381], device='cuda:2'), covar=tensor([0.0766, 0.1689, 0.1807, 0.2024, 0.2741, 0.0942, 0.1254, 0.2511], device='cuda:2'), in_proj_covar=tensor([0.0342, 0.0475, 0.0506, 0.0420, 0.0557, 0.0535, 0.0406, 0.0561], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:20:35,022 INFO [optim.py:368] (2/8) 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 00:20:37,091 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:11,873 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:20,735 INFO [train.py:904] (2/8) Epoch 9, batch 2150, loss[loss=0.194, simple_loss=0.2726, pruned_loss=0.05776, over 16423.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2743, pruned_loss=0.0554, over 3338977.50 frames. ], batch size: 68, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:22:23,636 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7778, 2.6821, 2.3188, 2.4690, 3.0942, 2.8797, 3.6866, 3.3773], device='cuda:2'), covar=tensor([0.0054, 0.0254, 0.0298, 0.0290, 0.0165, 0.0219, 0.0143, 0.0138], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0192, 0.0186, 0.0190, 0.0189, 0.0193, 0.0195, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:22:31,366 INFO [train.py:904] (2/8) Epoch 9, batch 2200, loss[loss=0.1748, simple_loss=0.2725, pruned_loss=0.03858, over 17122.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2752, pruned_loss=0.05607, over 3342689.91 frames. ], batch size: 47, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:22:41,852 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1314, 4.1207, 4.3613, 2.1253, 4.6480, 4.6635, 3.2577, 3.6236], device='cuda:2'), covar=tensor([0.0562, 0.0133, 0.0190, 0.1010, 0.0043, 0.0100, 0.0352, 0.0301], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0098, 0.0087, 0.0142, 0.0070, 0.0098, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 00:22:54,054 INFO [optim.py:368] (2/8) 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 00:23:41,045 INFO [train.py:904] (2/8) Epoch 9, batch 2250, loss[loss=0.1959, simple_loss=0.2886, pruned_loss=0.05157, over 17123.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2767, pruned_loss=0.05702, over 3328597.63 frames. ], batch size: 47, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:23:51,948 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7350, 3.9108, 3.0693, 2.2995, 2.7462, 2.3638, 4.0857, 3.7109], device='cuda:2'), covar=tensor([0.2135, 0.0529, 0.1288, 0.1998, 0.2110, 0.1588, 0.0390, 0.0914], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0256, 0.0279, 0.0269, 0.0281, 0.0215, 0.0263, 0.0289], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:23:54,936 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1894, 5.6283, 5.7405, 5.6114, 5.6441, 6.1373, 5.7436, 5.4859], device='cuda:2'), covar=tensor([0.0736, 0.1746, 0.1788, 0.2125, 0.2775, 0.0946, 0.1425, 0.2481], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0476, 0.0504, 0.0423, 0.0558, 0.0532, 0.0405, 0.0562], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:23:55,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1153, 4.4194, 4.5523, 3.4693, 3.8520, 4.5163, 4.0111, 2.8409], device='cuda:2'), covar=tensor([0.0282, 0.0037, 0.0022, 0.0193, 0.0072, 0.0045, 0.0050, 0.0282], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0068, 0.0066, 0.0123, 0.0074, 0.0082, 0.0073, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:24:49,196 INFO [train.py:904] (2/8) Epoch 9, batch 2300, loss[loss=0.2046, simple_loss=0.2732, pruned_loss=0.06796, over 16884.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2765, pruned_loss=0.057, over 3337589.06 frames. ], batch size: 109, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:12,015 INFO [optim.py:368] (2/8) 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 00:25:55,681 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 00:25:59,039 INFO [train.py:904] (2/8) Epoch 9, batch 2350, loss[loss=0.2029, simple_loss=0.2917, pruned_loss=0.057, over 16661.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2765, pruned_loss=0.05698, over 3330408.15 frames. ], batch size: 57, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:59,382 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:26:06,626 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 00:27:02,835 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8270, 4.8267, 5.3438, 5.3221, 5.3673, 4.9447, 4.9343, 4.7065], device='cuda:2'), covar=tensor([0.0270, 0.0465, 0.0351, 0.0432, 0.0419, 0.0302, 0.0864, 0.0388], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0318, 0.0323, 0.0308, 0.0365, 0.0343, 0.0444, 0.0271], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 00:27:06,421 INFO [train.py:904] (2/8) Epoch 9, batch 2400, loss[loss=0.1895, simple_loss=0.2776, pruned_loss=0.05066, over 17216.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2782, pruned_loss=0.05771, over 3326293.49 frames. ], batch size: 46, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:27:29,672 INFO [optim.py:368] (2/8) 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 00:27:31,281 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:27:59,096 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:28:14,955 INFO [train.py:904] (2/8) Epoch 9, batch 2450, loss[loss=0.2135, simple_loss=0.2919, pruned_loss=0.06757, over 17237.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2788, pruned_loss=0.05749, over 3328840.24 frames. ], batch size: 45, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:28:34,505 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:29:22,511 INFO [train.py:904] (2/8) Epoch 9, batch 2500, loss[loss=0.1998, simple_loss=0.278, pruned_loss=0.06082, over 16257.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2783, pruned_loss=0.05716, over 3330338.74 frames. ], batch size: 36, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:29:44,588 INFO [optim.py:368] (2/8) 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 00:30:28,782 INFO [train.py:904] (2/8) Epoch 9, batch 2550, loss[loss=0.1893, simple_loss=0.2741, pruned_loss=0.05223, over 16468.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2785, pruned_loss=0.05775, over 3313684.18 frames. ], batch size: 75, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:30:41,825 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:31:38,615 INFO [train.py:904] (2/8) Epoch 9, batch 2600, loss[loss=0.226, simple_loss=0.3024, pruned_loss=0.0748, over 12520.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2784, pruned_loss=0.05767, over 3319819.47 frames. ], batch size: 246, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:31:46,913 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 00:31:48,090 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 00:31:48,997 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 00:31:59,378 INFO [optim.py:368] (2/8) 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 00:32:01,156 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7514, 4.2271, 3.2019, 2.1587, 2.7718, 2.3059, 4.4847, 3.7655], device='cuda:2'), covar=tensor([0.2342, 0.0563, 0.1296, 0.2191, 0.2461, 0.1780, 0.0321, 0.0904], device='cuda:2'), in_proj_covar=tensor([0.0291, 0.0255, 0.0278, 0.0269, 0.0280, 0.0215, 0.0262, 0.0288], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:32:04,216 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:09,596 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2750, 4.2198, 4.1933, 4.0198, 3.9239, 4.2648, 4.0543, 4.0337], device='cuda:2'), covar=tensor([0.0635, 0.0479, 0.0243, 0.0230, 0.0730, 0.0389, 0.0533, 0.0659], device='cuda:2'), in_proj_covar=tensor([0.0248, 0.0299, 0.0289, 0.0261, 0.0315, 0.0296, 0.0198, 0.0333], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:32:42,453 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0223, 1.8687, 2.4743, 2.9172, 2.7856, 3.3343, 2.1687, 3.4149], device='cuda:2'), covar=tensor([0.0117, 0.0288, 0.0184, 0.0168, 0.0161, 0.0103, 0.0281, 0.0076], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0166, 0.0150, 0.0154, 0.0157, 0.0115, 0.0164, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 00:32:45,549 INFO [train.py:904] (2/8) Epoch 9, batch 2650, loss[loss=0.1724, simple_loss=0.2554, pruned_loss=0.04468, over 16976.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2785, pruned_loss=0.05726, over 3314492.05 frames. ], batch size: 41, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:32:45,903 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:49,366 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8495, 5.0663, 5.2317, 5.0711, 4.9715, 5.6778, 5.2269, 4.9533], device='cuda:2'), covar=tensor([0.1060, 0.1683, 0.1527, 0.1770, 0.2621, 0.1015, 0.1242, 0.2101], device='cuda:2'), in_proj_covar=tensor([0.0336, 0.0472, 0.0502, 0.0414, 0.0550, 0.0527, 0.0399, 0.0557], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:33:51,885 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:33:53,984 INFO [train.py:904] (2/8) Epoch 9, batch 2700, loss[loss=0.1913, simple_loss=0.284, pruned_loss=0.04929, over 17086.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2784, pruned_loss=0.05691, over 3319323.75 frames. ], batch size: 53, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:34:17,451 INFO [optim.py:368] (2/8) 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 00:34:47,148 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:34:58,131 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:02,459 INFO [train.py:904] (2/8) Epoch 9, batch 2750, loss[loss=0.2223, simple_loss=0.2943, pruned_loss=0.07517, over 15421.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2796, pruned_loss=0.05705, over 3326664.26 frames. ], batch size: 190, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:35:46,143 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8563, 2.1265, 2.1736, 4.6791, 1.9570, 2.8145, 2.2997, 2.4665], device='cuda:2'), covar=tensor([0.0749, 0.3161, 0.2030, 0.0271, 0.3576, 0.1997, 0.2791, 0.2921], device='cuda:2'), in_proj_covar=tensor([0.0355, 0.0376, 0.0316, 0.0325, 0.0400, 0.0427, 0.0335, 0.0445], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:35:48,316 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:51,691 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:36:15,277 INFO [train.py:904] (2/8) Epoch 9, batch 2800, loss[loss=0.166, simple_loss=0.2605, pruned_loss=0.03581, over 17109.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2786, pruned_loss=0.05621, over 3327475.96 frames. ], batch size: 47, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:36:25,612 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:36:36,296 INFO [optim.py:368] (2/8) 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 00:37:16,037 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:37:23,281 INFO [train.py:904] (2/8) Epoch 9, batch 2850, loss[loss=0.1876, simple_loss=0.2587, pruned_loss=0.05829, over 16791.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2775, pruned_loss=0.05603, over 3318769.79 frames. ], batch size: 102, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:37:35,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3060, 3.4244, 1.8365, 3.5225, 2.5018, 3.5246, 1.9044, 2.6539], device='cuda:2'), covar=tensor([0.0233, 0.0332, 0.1411, 0.0218, 0.0717, 0.0575, 0.1319, 0.0627], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0166, 0.0186, 0.0121, 0.0165, 0.0209, 0.0192, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 00:38:25,931 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:32,599 INFO [train.py:904] (2/8) Epoch 9, batch 2900, loss[loss=0.1957, simple_loss=0.283, pruned_loss=0.05414, over 17166.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2775, pruned_loss=0.05689, over 3317282.19 frames. ], batch size: 48, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:38:52,444 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:54,443 INFO [optim.py:368] (2/8) 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 00:39:30,510 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-29 00:39:43,186 INFO [train.py:904] (2/8) Epoch 9, batch 2950, loss[loss=0.2019, simple_loss=0.2766, pruned_loss=0.06359, over 16872.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2766, pruned_loss=0.05671, over 3322832.30 frames. ], batch size: 90, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:39:51,130 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:47,953 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:52,809 INFO [train.py:904] (2/8) Epoch 9, batch 3000, loss[loss=0.1701, simple_loss=0.2542, pruned_loss=0.04297, over 16851.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2765, pruned_loss=0.05719, over 3323524.29 frames. ], batch size: 42, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:40:52,809 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 00:41:02,055 INFO [train.py:938] (2/8) Epoch 9, validation: loss=0.1444, simple_loss=0.2507, pruned_loss=0.019, over 944034.00 frames. 2023-04-29 00:41:02,056 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 00:41:23,147 INFO [optim.py:368] (2/8) 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 00:41:33,522 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7061, 4.6133, 4.5980, 4.4319, 4.2700, 4.6066, 4.4353, 4.3599], device='cuda:2'), covar=tensor([0.0603, 0.0601, 0.0239, 0.0206, 0.0827, 0.0451, 0.0412, 0.0633], device='cuda:2'), in_proj_covar=tensor([0.0246, 0.0297, 0.0285, 0.0260, 0.0312, 0.0296, 0.0196, 0.0331], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:41:47,896 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:09,946 INFO [train.py:904] (2/8) Epoch 9, batch 3050, loss[loss=0.1864, simple_loss=0.2625, pruned_loss=0.05521, over 16812.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2761, pruned_loss=0.05708, over 3328427.23 frames. ], batch size: 83, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:42:20,970 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:41,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6861, 4.7109, 5.1683, 5.1322, 5.1563, 4.7703, 4.7662, 4.5494], device='cuda:2'), covar=tensor([0.0267, 0.0427, 0.0342, 0.0405, 0.0400, 0.0295, 0.0781, 0.0387], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0326, 0.0329, 0.0314, 0.0372, 0.0347, 0.0458, 0.0279], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 00:42:45,156 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4132, 5.7605, 5.5356, 5.5901, 5.1711, 4.8943, 5.2595, 5.9436], device='cuda:2'), covar=tensor([0.1072, 0.0903, 0.0971, 0.0590, 0.0801, 0.0645, 0.0841, 0.0768], device='cuda:2'), in_proj_covar=tensor([0.0524, 0.0664, 0.0554, 0.0451, 0.0409, 0.0421, 0.0546, 0.0491], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:43:11,742 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:43:16,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0022, 5.1762, 4.9071, 4.7207, 3.9660, 5.1271, 5.1544, 4.5936], device='cuda:2'), covar=tensor([0.0799, 0.0401, 0.0462, 0.0330, 0.2006, 0.0414, 0.0285, 0.0648], device='cuda:2'), in_proj_covar=tensor([0.0248, 0.0299, 0.0288, 0.0262, 0.0315, 0.0297, 0.0198, 0.0334], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:43:17,163 INFO [train.py:904] (2/8) Epoch 9, batch 3100, loss[loss=0.2169, simple_loss=0.313, pruned_loss=0.06042, over 17104.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2761, pruned_loss=0.05658, over 3327000.06 frames. ], batch size: 47, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:43:22,113 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 00:43:39,988 INFO [optim.py:368] (2/8) 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 00:43:40,497 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0241, 1.7703, 2.4365, 2.8144, 2.7785, 3.2560, 2.0365, 3.1662], device='cuda:2'), covar=tensor([0.0125, 0.0324, 0.0215, 0.0187, 0.0174, 0.0097, 0.0312, 0.0085], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0165, 0.0152, 0.0155, 0.0157, 0.0115, 0.0164, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 00:44:06,108 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:12,240 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:19,933 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-29 00:44:28,370 INFO [train.py:904] (2/8) Epoch 9, batch 3150, loss[loss=0.2191, simple_loss=0.295, pruned_loss=0.07161, over 12023.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2749, pruned_loss=0.05626, over 3320483.17 frames. ], batch size: 247, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:44:30,012 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1041, 5.1994, 5.6241, 5.6565, 5.6438, 5.2615, 5.1987, 5.0132], device='cuda:2'), covar=tensor([0.0293, 0.0442, 0.0407, 0.0397, 0.0413, 0.0279, 0.0894, 0.0380], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0325, 0.0329, 0.0312, 0.0371, 0.0347, 0.0457, 0.0278], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 00:45:30,247 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:36,486 INFO [train.py:904] (2/8) Epoch 9, batch 3200, loss[loss=0.1723, simple_loss=0.2527, pruned_loss=0.04596, over 15968.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2742, pruned_loss=0.05598, over 3317356.49 frames. ], batch size: 35, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:45:41,762 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5761, 1.5753, 2.2141, 2.4588, 2.5328, 2.5024, 1.8064, 2.6016], device='cuda:2'), covar=tensor([0.0101, 0.0284, 0.0177, 0.0151, 0.0141, 0.0137, 0.0247, 0.0085], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0164, 0.0151, 0.0154, 0.0157, 0.0115, 0.0162, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 00:45:56,428 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:59,093 INFO [optim.py:368] (2/8) 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 00:46:28,740 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 00:46:38,663 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4127, 3.3362, 3.6113, 1.8635, 3.7361, 3.7604, 3.0079, 2.6699], device='cuda:2'), covar=tensor([0.0748, 0.0199, 0.0167, 0.1184, 0.0069, 0.0143, 0.0402, 0.0470], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0097, 0.0088, 0.0140, 0.0070, 0.0100, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 00:46:45,892 INFO [train.py:904] (2/8) Epoch 9, batch 3250, loss[loss=0.1576, simple_loss=0.2421, pruned_loss=0.03651, over 17221.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.275, pruned_loss=0.05637, over 3322076.26 frames. ], batch size: 45, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:46:47,423 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:46:52,261 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7752, 3.5350, 2.9142, 5.1990, 4.4128, 4.8299, 1.7057, 3.5332], device='cuda:2'), covar=tensor([0.1260, 0.0515, 0.0980, 0.0107, 0.0223, 0.0305, 0.1364, 0.0602], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0153, 0.0175, 0.0129, 0.0201, 0.0211, 0.0172, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 00:46:52,666 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 00:47:03,266 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:30,799 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2023-04-29 00:47:34,564 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9776, 4.4442, 4.5205, 3.2632, 3.8585, 4.5300, 4.1218, 2.8224], device='cuda:2'), covar=tensor([0.0319, 0.0023, 0.0023, 0.0224, 0.0066, 0.0048, 0.0046, 0.0285], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0068, 0.0067, 0.0122, 0.0074, 0.0083, 0.0074, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:47:51,060 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9274, 2.2411, 2.3375, 4.7994, 2.1021, 2.7984, 2.3708, 2.4817], device='cuda:2'), covar=tensor([0.0742, 0.3294, 0.2050, 0.0292, 0.3787, 0.2229, 0.2739, 0.3299], device='cuda:2'), in_proj_covar=tensor([0.0358, 0.0377, 0.0317, 0.0326, 0.0400, 0.0427, 0.0336, 0.0447], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:47:55,748 INFO [train.py:904] (2/8) Epoch 9, batch 3300, loss[loss=0.2116, simple_loss=0.2976, pruned_loss=0.06274, over 17087.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2754, pruned_loss=0.05623, over 3321947.90 frames. ], batch size: 53, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:48:18,092 INFO [optim.py:368] (2/8) 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 00:48:27,885 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:04,554 INFO [train.py:904] (2/8) Epoch 9, batch 3350, loss[loss=0.1895, simple_loss=0.2774, pruned_loss=0.05084, over 17092.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2765, pruned_loss=0.0566, over 3316979.61 frames. ], batch size: 47, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:49:10,236 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:39,432 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 00:49:53,323 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:01,570 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:16,096 INFO [train.py:904] (2/8) Epoch 9, batch 3400, loss[loss=0.2089, simple_loss=0.2778, pruned_loss=0.06999, over 16879.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2762, pruned_loss=0.05663, over 3327326.09 frames. ], batch size: 109, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:50:20,185 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:38,673 INFO [optim.py:368] (2/8) 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 00:50:59,424 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:10,788 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:24,588 INFO [train.py:904] (2/8) Epoch 9, batch 3450, loss[loss=0.1764, simple_loss=0.2509, pruned_loss=0.05098, over 15802.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2743, pruned_loss=0.05612, over 3323747.39 frames. ], batch size: 35, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:51:26,677 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:55,891 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8438, 2.3829, 1.9252, 2.2246, 2.8447, 2.6280, 3.0246, 2.9740], device='cuda:2'), covar=tensor([0.0099, 0.0267, 0.0313, 0.0303, 0.0121, 0.0222, 0.0146, 0.0146], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0194, 0.0190, 0.0190, 0.0191, 0.0196, 0.0201, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 00:52:16,067 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:20,243 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:24,205 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:35,103 INFO [train.py:904] (2/8) Epoch 9, batch 3500, loss[loss=0.2001, simple_loss=0.2771, pruned_loss=0.06156, over 11372.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.273, pruned_loss=0.05587, over 3327495.00 frames. ], batch size: 247, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:52:56,975 INFO [optim.py:368] (2/8) 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 00:53:44,895 INFO [train.py:904] (2/8) Epoch 9, batch 3550, loss[loss=0.1793, simple_loss=0.253, pruned_loss=0.0528, over 16830.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2719, pruned_loss=0.05549, over 3327100.02 frames. ], batch size: 42, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:53:46,580 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:53:50,240 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6215, 3.5194, 3.8768, 2.7482, 3.5490, 3.8765, 3.6687, 2.0658], device='cuda:2'), covar=tensor([0.0347, 0.0144, 0.0031, 0.0248, 0.0059, 0.0061, 0.0053, 0.0357], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0070, 0.0068, 0.0124, 0.0075, 0.0085, 0.0076, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 00:54:53,072 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:54,908 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:55,838 INFO [train.py:904] (2/8) Epoch 9, batch 3600, loss[loss=0.2086, simple_loss=0.2712, pruned_loss=0.07297, over 16911.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2711, pruned_loss=0.05449, over 3325837.78 frames. ], batch size: 116, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:55:17,924 INFO [optim.py:368] (2/8) 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 00:55:52,963 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 00:56:07,140 INFO [train.py:904] (2/8) Epoch 9, batch 3650, loss[loss=0.1951, simple_loss=0.2566, pruned_loss=0.06683, over 16913.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2702, pruned_loss=0.05471, over 3319381.49 frames. ], batch size: 116, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:56:12,759 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:21,101 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:46,851 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 00:56:50,576 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:07,339 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:21,437 INFO [train.py:904] (2/8) Epoch 9, batch 3700, loss[loss=0.1883, simple_loss=0.2635, pruned_loss=0.05654, over 16887.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2685, pruned_loss=0.05629, over 3309846.81 frames. ], batch size: 96, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:57:23,506 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:46,406 INFO [optim.py:368] (2/8) 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 00:58:19,043 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:34,932 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:35,655 INFO [train.py:904] (2/8) Epoch 9, batch 3750, loss[loss=0.1744, simple_loss=0.2466, pruned_loss=0.05115, over 16625.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2687, pruned_loss=0.05772, over 3292174.55 frames. ], batch size: 89, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 00:59:28,297 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:33,935 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:47,555 INFO [train.py:904] (2/8) Epoch 9, batch 3800, loss[loss=0.2104, simple_loss=0.287, pruned_loss=0.0669, over 16893.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2698, pruned_loss=0.0589, over 3289942.08 frames. ], batch size: 116, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 00:59:48,467 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 01:00:03,351 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:00:11,015 INFO [optim.py:368] (2/8) 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 01:00:43,934 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:01:00,353 INFO [train.py:904] (2/8) Epoch 9, batch 3850, loss[loss=0.1781, simple_loss=0.2584, pruned_loss=0.04892, over 15475.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2699, pruned_loss=0.05923, over 3295835.39 frames. ], batch size: 190, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:01:22,016 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4765, 3.4606, 2.4703, 2.1797, 2.4488, 1.9424, 3.3659, 3.1831], device='cuda:2'), covar=tensor([0.2332, 0.0644, 0.1578, 0.2071, 0.2224, 0.2140, 0.0633, 0.1098], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0257, 0.0279, 0.0271, 0.0285, 0.0217, 0.0263, 0.0293], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:01:43,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3010, 2.2210, 1.7437, 2.0408, 2.5406, 2.3912, 2.6333, 2.7504], device='cuda:2'), covar=tensor([0.0115, 0.0239, 0.0332, 0.0301, 0.0124, 0.0192, 0.0130, 0.0141], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0188, 0.0185, 0.0185, 0.0185, 0.0190, 0.0194, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:02:00,520 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0817, 3.9480, 4.1346, 4.2814, 4.3643, 3.9774, 4.1347, 4.3393], device='cuda:2'), covar=tensor([0.1300, 0.0857, 0.1124, 0.0563, 0.0541, 0.1254, 0.1235, 0.0553], device='cuda:2'), in_proj_covar=tensor([0.0524, 0.0639, 0.0800, 0.0652, 0.0496, 0.0494, 0.0509, 0.0570], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:02:13,000 INFO [train.py:904] (2/8) Epoch 9, batch 3900, loss[loss=0.1908, simple_loss=0.2641, pruned_loss=0.05876, over 15545.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2695, pruned_loss=0.05985, over 3301016.03 frames. ], batch size: 191, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:02:21,639 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:32,620 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:36,232 INFO [optim.py:368] (2/8) 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 01:03:10,363 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-04-29 01:03:24,358 INFO [train.py:904] (2/8) Epoch 9, batch 3950, loss[loss=0.2263, simple_loss=0.2973, pruned_loss=0.07761, over 15480.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2695, pruned_loss=0.06044, over 3294390.78 frames. ], batch size: 190, lr: 7.66e-03, grad_scale: 8.0 2023-04-29 01:03:30,540 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1174, 1.8999, 2.5367, 2.9596, 2.8941, 3.1130, 1.9458, 3.1260], device='cuda:2'), covar=tensor([0.0098, 0.0289, 0.0188, 0.0178, 0.0144, 0.0110, 0.0306, 0.0076], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0164, 0.0151, 0.0153, 0.0158, 0.0116, 0.0163, 0.0106], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 01:03:32,139 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:03:41,960 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2023-04-29 01:03:50,058 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:04:00,736 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:06,368 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 01:04:07,167 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:29,046 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2008, 2.1051, 1.6122, 1.8156, 2.3911, 2.2135, 2.3131, 2.5778], device='cuda:2'), covar=tensor([0.0116, 0.0232, 0.0348, 0.0323, 0.0128, 0.0198, 0.0135, 0.0136], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0189, 0.0187, 0.0186, 0.0186, 0.0189, 0.0195, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:04:37,098 INFO [train.py:904] (2/8) Epoch 9, batch 4000, loss[loss=0.1893, simple_loss=0.2707, pruned_loss=0.054, over 16321.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2691, pruned_loss=0.06069, over 3293698.75 frames. ], batch size: 165, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:04:49,126 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-29 01:05:00,769 INFO [optim.py:368] (2/8) 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 01:05:17,722 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:05:47,389 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5761, 5.9004, 5.5666, 5.7095, 5.1444, 4.9362, 5.3498, 5.9798], device='cuda:2'), covar=tensor([0.0961, 0.0767, 0.0991, 0.0577, 0.0820, 0.0696, 0.0813, 0.0778], device='cuda:2'), in_proj_covar=tensor([0.0510, 0.0650, 0.0536, 0.0442, 0.0406, 0.0413, 0.0540, 0.0483], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:05:51,192 INFO [train.py:904] (2/8) Epoch 9, batch 4050, loss[loss=0.1925, simple_loss=0.2714, pruned_loss=0.05681, over 15412.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2691, pruned_loss=0.05928, over 3285782.79 frames. ], batch size: 190, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:06:45,645 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:04,622 INFO [train.py:904] (2/8) Epoch 9, batch 4100, loss[loss=0.2482, simple_loss=0.3285, pruned_loss=0.08398, over 15325.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2704, pruned_loss=0.05829, over 3282316.19 frames. ], batch size: 190, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:07:12,414 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:21,502 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1690, 4.3668, 4.6466, 4.5689, 4.6237, 4.3360, 4.1284, 4.0902], device='cuda:2'), covar=tensor([0.0413, 0.0570, 0.0475, 0.0606, 0.0496, 0.0379, 0.1089, 0.0490], device='cuda:2'), in_proj_covar=tensor([0.0310, 0.0318, 0.0321, 0.0309, 0.0363, 0.0336, 0.0440, 0.0268], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 01:07:28,042 INFO [optim.py:368] (2/8) 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 01:07:57,462 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:08:20,247 INFO [train.py:904] (2/8) Epoch 9, batch 4150, loss[loss=0.237, simple_loss=0.3107, pruned_loss=0.08169, over 16635.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2781, pruned_loss=0.06138, over 3248470.31 frames. ], batch size: 57, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:09:02,294 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6499, 4.4788, 4.6890, 4.8275, 4.9701, 4.4933, 4.9481, 4.9712], device='cuda:2'), covar=tensor([0.1250, 0.0845, 0.1152, 0.0499, 0.0369, 0.0739, 0.0462, 0.0426], device='cuda:2'), in_proj_covar=tensor([0.0503, 0.0610, 0.0762, 0.0622, 0.0473, 0.0476, 0.0489, 0.0545], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:09:37,254 INFO [train.py:904] (2/8) Epoch 9, batch 4200, loss[loss=0.2419, simple_loss=0.3225, pruned_loss=0.08062, over 16871.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2859, pruned_loss=0.0638, over 3216833.49 frames. ], batch size: 116, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:09:44,848 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 01:10:02,491 INFO [optim.py:368] (2/8) 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 01:10:17,683 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:10:52,290 INFO [train.py:904] (2/8) Epoch 9, batch 4250, loss[loss=0.1875, simple_loss=0.2802, pruned_loss=0.04738, over 16686.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2889, pruned_loss=0.06419, over 3182560.06 frames. ], batch size: 89, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:59,202 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:08,338 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:11:14,299 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5528, 2.5211, 2.3629, 3.9458, 2.7891, 3.9093, 1.3512, 2.7998], device='cuda:2'), covar=tensor([0.1288, 0.0712, 0.1113, 0.0137, 0.0259, 0.0379, 0.1566, 0.0778], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0153, 0.0174, 0.0127, 0.0200, 0.0206, 0.0174, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 01:11:20,332 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:39,003 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:48,302 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 9, batch 4300, loss[loss=0.2262, simple_loss=0.3057, pruned_loss=0.0733, over 11416.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2902, pruned_loss=0.06315, over 3180697.48 frames. ], batch size: 247, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:12:08,302 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 01:12:11,409 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:21,378 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 01:12:30,788 INFO [optim.py:368] (2/8) 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 01:12:49,196 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:10,236 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:21,362 INFO [train.py:904] (2/8) Epoch 9, batch 4350, loss[loss=0.2065, simple_loss=0.29, pruned_loss=0.06148, over 16535.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2933, pruned_loss=0.06428, over 3176536.51 frames. ], batch size: 68, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:14:18,974 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:22,861 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 01:14:34,737 INFO [train.py:904] (2/8) Epoch 9, batch 4400, loss[loss=0.22, simple_loss=0.2974, pruned_loss=0.07131, over 16744.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2958, pruned_loss=0.06537, over 3161915.53 frames. ], batch size: 124, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:14:41,618 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 01:15:06,846 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:15:46,021 INFO [train.py:904] (2/8) Epoch 9, batch 4450, loss[loss=0.2198, simple_loss=0.3139, pruned_loss=0.06289, over 16820.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2992, pruned_loss=0.06629, over 3178283.39 frames. ], batch size: 83, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:15:50,416 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:16:33,316 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:16:56,347 INFO [train.py:904] (2/8) Epoch 9, batch 4500, loss[loss=0.2106, simple_loss=0.2931, pruned_loss=0.06407, over 16998.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2989, pruned_loss=0.0663, over 3180940.83 frames. ], batch size: 41, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:17:20,314 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 9, batch 4550, loss[loss=0.2351, simple_loss=0.3219, pruned_loss=0.07417, over 16878.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.3001, pruned_loss=0.06727, over 3190225.89 frames. ], batch size: 96, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:18:23,663 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:34,937 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:54,374 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:20,640 INFO [train.py:904] (2/8) Epoch 9, batch 4600, loss[loss=0.2075, simple_loss=0.2916, pruned_loss=0.06165, over 16651.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3008, pruned_loss=0.06759, over 3183695.26 frames. ], batch size: 134, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:19:32,652 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:40,510 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 01:19:42,422 INFO [optim.py:368] (2/8) 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 01:19:44,975 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:12,953 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:30,860 INFO [train.py:904] (2/8) Epoch 9, batch 4650, loss[loss=0.2084, simple_loss=0.2883, pruned_loss=0.06428, over 16812.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2993, pruned_loss=0.06715, over 3189653.82 frames. ], batch size: 83, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:21:04,821 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:20,371 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:42,720 INFO [train.py:904] (2/8) Epoch 9, batch 4700, loss[loss=0.2106, simple_loss=0.2884, pruned_loss=0.06645, over 16603.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2965, pruned_loss=0.06585, over 3174946.95 frames. ], batch size: 62, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:22:06,316 INFO [optim.py:368] (2/8) 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 01:22:20,135 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9253, 1.9949, 2.2833, 3.1432, 2.1058, 2.2993, 2.1843, 2.1035], device='cuda:2'), covar=tensor([0.0843, 0.2738, 0.1596, 0.0488, 0.3241, 0.1934, 0.2457, 0.2933], device='cuda:2'), in_proj_covar=tensor([0.0353, 0.0376, 0.0313, 0.0316, 0.0403, 0.0429, 0.0335, 0.0443], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:22:25,138 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:22:32,810 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:22:55,506 INFO [train.py:904] (2/8) Epoch 9, batch 4750, loss[loss=0.1805, simple_loss=0.2646, pruned_loss=0.04824, over 16619.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2923, pruned_loss=0.06359, over 3184350.88 frames. ], batch size: 57, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:23:04,425 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:23:39,895 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:23:44,214 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0084, 3.8934, 1.7582, 4.7434, 2.6359, 4.4462, 2.0072, 2.9741], device='cuda:2'), covar=tensor([0.0168, 0.0287, 0.2152, 0.0055, 0.0820, 0.0304, 0.1875, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0161, 0.0184, 0.0111, 0.0163, 0.0199, 0.0191, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 01:23:54,652 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:24:13,772 INFO [train.py:904] (2/8) Epoch 9, batch 4800, loss[loss=0.2174, simple_loss=0.2933, pruned_loss=0.07077, over 12316.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2889, pruned_loss=0.062, over 3174729.39 frames. ], batch size: 247, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:24:14,312 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:24:37,221 INFO [optim.py:368] (2/8) 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 01:24:38,383 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2591, 5.2060, 5.1954, 4.4408, 5.2450, 1.7388, 4.9339, 5.0694], device='cuda:2'), covar=tensor([0.0071, 0.0067, 0.0096, 0.0476, 0.0063, 0.2179, 0.0098, 0.0154], device='cuda:2'), in_proj_covar=tensor([0.0116, 0.0105, 0.0153, 0.0145, 0.0123, 0.0165, 0.0138, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:24:38,419 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:28,104 INFO [train.py:904] (2/8) Epoch 9, batch 4850, loss[loss=0.208, simple_loss=0.2944, pruned_loss=0.06084, over 16871.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2896, pruned_loss=0.06116, over 3177599.57 frames. ], batch size: 116, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:25:28,540 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:35,663 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 01:25:45,132 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:49,925 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5576, 3.7460, 3.9587, 1.8744, 4.2496, 4.2132, 3.0673, 2.9824], device='cuda:2'), covar=tensor([0.0708, 0.0155, 0.0116, 0.1139, 0.0036, 0.0068, 0.0292, 0.0413], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0096, 0.0085, 0.0139, 0.0068, 0.0096, 0.0119, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 01:25:54,692 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:00,609 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 01:26:16,450 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:24,518 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4049, 2.5898, 2.3329, 4.1614, 2.7469, 4.0368, 1.3312, 2.7272], device='cuda:2'), covar=tensor([0.1474, 0.0707, 0.1163, 0.0114, 0.0232, 0.0320, 0.1653, 0.0854], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0150, 0.0172, 0.0124, 0.0196, 0.0203, 0.0171, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 01:26:37,747 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8262, 2.2652, 2.2092, 2.8389, 2.1689, 3.2251, 1.5922, 2.7136], device='cuda:2'), covar=tensor([0.1271, 0.0597, 0.1000, 0.0155, 0.0123, 0.0300, 0.1495, 0.0622], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0150, 0.0172, 0.0124, 0.0195, 0.0202, 0.0170, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 01:26:40,904 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5033, 4.5669, 5.0349, 4.9602, 4.9557, 4.5793, 4.6371, 4.3839], device='cuda:2'), covar=tensor([0.0239, 0.0396, 0.0242, 0.0332, 0.0361, 0.0266, 0.0642, 0.0361], device='cuda:2'), in_proj_covar=tensor([0.0295, 0.0297, 0.0305, 0.0292, 0.0346, 0.0318, 0.0420, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-29 01:26:42,269 INFO [train.py:904] (2/8) Epoch 9, batch 4900, loss[loss=0.1832, simple_loss=0.2673, pruned_loss=0.04958, over 16753.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2885, pruned_loss=0.05942, over 3179673.05 frames. ], batch size: 76, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:26:51,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8108, 1.3363, 1.6077, 1.7126, 1.8774, 1.8538, 1.5217, 1.7973], device='cuda:2'), covar=tensor([0.0140, 0.0238, 0.0119, 0.0181, 0.0146, 0.0108, 0.0226, 0.0063], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0162, 0.0148, 0.0151, 0.0155, 0.0113, 0.0161, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 01:26:56,180 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 01:26:59,856 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86113.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:05,786 INFO [optim.py:368] (2/8) 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 01:27:23,804 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:25,906 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:37,326 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:38,859 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 01:27:55,672 INFO [train.py:904] (2/8) Epoch 9, batch 4950, loss[loss=0.2112, simple_loss=0.2934, pruned_loss=0.06454, over 12291.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2879, pruned_loss=0.05862, over 3179601.43 frames. ], batch size: 246, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:28:29,990 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9008, 4.1806, 3.9598, 4.0554, 3.6511, 3.7249, 3.8102, 4.1149], device='cuda:2'), covar=tensor([0.1054, 0.0875, 0.0971, 0.0593, 0.0785, 0.1614, 0.0875, 0.1017], device='cuda:2'), in_proj_covar=tensor([0.0486, 0.0611, 0.0511, 0.0415, 0.0382, 0.0395, 0.0508, 0.0461], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:28:44,735 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 01:28:45,917 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:28:46,997 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:08,681 INFO [train.py:904] (2/8) Epoch 9, batch 5000, loss[loss=0.2056, simple_loss=0.2922, pruned_loss=0.0595, over 17035.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2894, pruned_loss=0.05902, over 3178908.14 frames. ], batch size: 55, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:29:32,026 INFO [optim.py:368] (2/8) 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 01:29:49,841 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:54,564 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:11,780 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:19,716 INFO [train.py:904] (2/8) Epoch 9, batch 5050, loss[loss=0.2122, simple_loss=0.3, pruned_loss=0.0622, over 16239.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2899, pruned_loss=0.05873, over 3192119.71 frames. ], batch size: 165, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:31:01,187 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:07,940 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:31:12,964 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2661, 3.7592, 3.4503, 1.8761, 3.0292, 2.5856, 3.5995, 3.5299], device='cuda:2'), covar=tensor([0.0245, 0.0583, 0.0585, 0.1772, 0.0742, 0.0774, 0.0641, 0.0848], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0141, 0.0158, 0.0142, 0.0136, 0.0126, 0.0137, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 01:31:29,532 INFO [train.py:904] (2/8) Epoch 9, batch 5100, loss[loss=0.2473, simple_loss=0.3147, pruned_loss=0.08997, over 11819.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2879, pruned_loss=0.05768, over 3193977.31 frames. ], batch size: 247, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:31:37,460 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:42,194 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5416, 4.3536, 4.5437, 4.7750, 4.9047, 4.4146, 4.8654, 4.8904], device='cuda:2'), covar=tensor([0.1230, 0.0979, 0.1235, 0.0512, 0.0382, 0.0814, 0.0486, 0.0474], device='cuda:2'), in_proj_covar=tensor([0.0494, 0.0603, 0.0747, 0.0612, 0.0464, 0.0467, 0.0485, 0.0537], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:31:45,610 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:52,760 INFO [optim.py:368] (2/8) 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 01:32:08,055 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:32:33,271 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7785, 1.7473, 2.1024, 2.6684, 2.6226, 2.9026, 1.6931, 2.8184], device='cuda:2'), covar=tensor([0.0096, 0.0275, 0.0204, 0.0170, 0.0150, 0.0105, 0.0310, 0.0062], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0164, 0.0148, 0.0152, 0.0156, 0.0114, 0.0163, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 01:32:41,540 INFO [train.py:904] (2/8) Epoch 9, batch 5150, loss[loss=0.1768, simple_loss=0.268, pruned_loss=0.04281, over 17264.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2875, pruned_loss=0.05657, over 3203286.76 frames. ], batch size: 52, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:32:50,287 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:33:54,429 INFO [train.py:904] (2/8) Epoch 9, batch 5200, loss[loss=0.1817, simple_loss=0.2677, pruned_loss=0.04788, over 16905.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.286, pruned_loss=0.05624, over 3201976.31 frames. ], batch size: 109, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:34:02,829 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:17,298 INFO [optim.py:368] (2/8) 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 01:34:27,688 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:50,084 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:57,839 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:35:06,327 INFO [train.py:904] (2/8) Epoch 9, batch 5250, loss[loss=0.1969, simple_loss=0.2898, pruned_loss=0.05197, over 16749.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2841, pruned_loss=0.05598, over 3205147.75 frames. ], batch size: 83, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:35:07,236 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 01:36:17,626 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:36:19,004 INFO [train.py:904] (2/8) Epoch 9, batch 5300, loss[loss=0.2149, simple_loss=0.2814, pruned_loss=0.0742, over 12178.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2805, pruned_loss=0.05453, over 3209459.38 frames. ], batch size: 248, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:36:25,653 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:36:42,025 INFO [optim.py:368] (2/8) 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 01:36:58,146 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6684, 3.7065, 4.0825, 4.0311, 4.0253, 3.7783, 3.7450, 3.7578], device='cuda:2'), covar=tensor([0.0291, 0.0621, 0.0346, 0.0392, 0.0421, 0.0350, 0.0878, 0.0468], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0304, 0.0311, 0.0300, 0.0353, 0.0326, 0.0432, 0.0263], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 01:37:00,843 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 01:37:01,814 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:37:18,984 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8792, 2.2863, 2.2654, 2.9053, 2.0410, 3.1902, 1.6126, 2.6307], device='cuda:2'), covar=tensor([0.1111, 0.0631, 0.0937, 0.0140, 0.0128, 0.0346, 0.1263, 0.0682], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0153, 0.0175, 0.0126, 0.0200, 0.0205, 0.0173, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 01:37:32,903 INFO [train.py:904] (2/8) Epoch 9, batch 5350, loss[loss=0.229, simple_loss=0.3011, pruned_loss=0.07845, over 12144.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2791, pruned_loss=0.05412, over 3200134.93 frames. ], batch size: 248, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:37:33,668 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 01:38:11,778 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:13,783 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:23,807 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:38:45,868 INFO [train.py:904] (2/8) Epoch 9, batch 5400, loss[loss=0.2002, simple_loss=0.2947, pruned_loss=0.0529, over 16863.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2822, pruned_loss=0.05525, over 3191891.80 frames. ], batch size: 116, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:46,249 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:01,760 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:09,035 INFO [optim.py:368] (2/8) 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 01:39:31,534 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:39:41,849 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:42,930 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7866, 3.7907, 4.2367, 4.1977, 4.2150, 3.8653, 3.9056, 3.8220], device='cuda:2'), covar=tensor([0.0347, 0.0660, 0.0372, 0.0433, 0.0449, 0.0401, 0.0893, 0.0507], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0304, 0.0311, 0.0300, 0.0353, 0.0326, 0.0429, 0.0263], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 01:39:46,332 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8344, 2.7207, 2.6308, 1.9253, 2.5301, 2.6783, 2.5463, 1.8071], device='cuda:2'), covar=tensor([0.0326, 0.0041, 0.0037, 0.0259, 0.0072, 0.0074, 0.0059, 0.0308], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0065, 0.0065, 0.0121, 0.0072, 0.0082, 0.0071, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 01:40:00,476 INFO [train.py:904] (2/8) Epoch 9, batch 5450, loss[loss=0.2692, simple_loss=0.3454, pruned_loss=0.0965, over 16226.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.285, pruned_loss=0.05687, over 3185878.65 frames. ], batch size: 165, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:40:10,941 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:15,340 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:44,621 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 01:40:51,610 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2945, 4.5908, 4.4011, 4.3790, 4.0785, 4.0493, 4.1677, 4.6591], device='cuda:2'), covar=tensor([0.0978, 0.0855, 0.0948, 0.0651, 0.0780, 0.1386, 0.0856, 0.0805], device='cuda:2'), in_proj_covar=tensor([0.0487, 0.0615, 0.0509, 0.0416, 0.0381, 0.0394, 0.0506, 0.0460], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:41:18,953 INFO [train.py:904] (2/8) Epoch 9, batch 5500, loss[loss=0.2726, simple_loss=0.3411, pruned_loss=0.102, over 11599.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.293, pruned_loss=0.06235, over 3163021.21 frames. ], batch size: 248, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:41:24,318 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:27,639 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:43,579 INFO [optim.py:368] (2/8) 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 01:41:54,525 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:42:35,760 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0549, 1.7294, 2.5148, 2.9621, 2.8022, 3.3573, 1.9258, 3.2125], device='cuda:2'), covar=tensor([0.0108, 0.0364, 0.0202, 0.0169, 0.0177, 0.0093, 0.0320, 0.0088], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0164, 0.0148, 0.0151, 0.0158, 0.0113, 0.0164, 0.0104], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 01:42:36,877 INFO [train.py:904] (2/8) Epoch 9, batch 5550, loss[loss=0.2128, simple_loss=0.2936, pruned_loss=0.06604, over 16618.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3013, pruned_loss=0.06871, over 3150316.45 frames. ], batch size: 57, lr: 7.59e-03, grad_scale: 16.0 2023-04-29 01:42:43,065 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:09,698 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:44,626 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:54,006 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:43:54,666 INFO [train.py:904] (2/8) Epoch 9, batch 5600, loss[loss=0.3799, simple_loss=0.4015, pruned_loss=0.1792, over 11129.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3079, pruned_loss=0.07446, over 3098334.67 frames. ], batch size: 248, lr: 7.59e-03, grad_scale: 8.0 2023-04-29 01:44:14,070 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:44:15,971 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0208, 2.4625, 2.2256, 2.7013, 2.2613, 3.2708, 1.6670, 2.7089], device='cuda:2'), covar=tensor([0.1110, 0.0449, 0.1023, 0.0141, 0.0161, 0.0432, 0.1369, 0.0630], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0152, 0.0176, 0.0127, 0.0199, 0.0205, 0.0174, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 01:44:21,512 INFO [optim.py:368] (2/8) 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 01:44:24,172 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3909, 4.4545, 4.8843, 4.8504, 4.8876, 4.4586, 4.5192, 4.2987], device='cuda:2'), covar=tensor([0.0267, 0.0451, 0.0304, 0.0365, 0.0417, 0.0306, 0.0882, 0.0446], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0301, 0.0311, 0.0299, 0.0350, 0.0325, 0.0428, 0.0263], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 01:44:39,238 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 01:45:17,290 INFO [train.py:904] (2/8) Epoch 9, batch 5650, loss[loss=0.2147, simple_loss=0.3046, pruned_loss=0.06235, over 16535.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3132, pruned_loss=0.07881, over 3068053.81 frames. ], batch size: 75, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:45:53,431 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:45:58,561 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:17,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8221, 5.2099, 5.4162, 5.2476, 5.2148, 5.8055, 5.2554, 5.1151], device='cuda:2'), covar=tensor([0.0948, 0.1822, 0.1520, 0.1637, 0.2273, 0.0856, 0.1281, 0.2142], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0447, 0.0474, 0.0390, 0.0520, 0.0501, 0.0385, 0.0529], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 01:46:33,109 INFO [train.py:904] (2/8) Epoch 9, batch 5700, loss[loss=0.2791, simple_loss=0.3323, pruned_loss=0.113, over 11179.00 frames. ], tot_loss[loss=0.2385, simple_loss=0.3151, pruned_loss=0.08095, over 3046549.30 frames. ], batch size: 247, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:46:33,459 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:59,879 INFO [optim.py:368] (2/8) 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 01:47:07,858 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5939, 3.7526, 2.0240, 4.3402, 2.5821, 4.1959, 2.3326, 2.9254], device='cuda:2'), covar=tensor([0.0204, 0.0285, 0.1622, 0.0064, 0.0778, 0.0376, 0.1371, 0.0646], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0161, 0.0185, 0.0110, 0.0165, 0.0200, 0.0192, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 01:47:25,252 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:30,332 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:48,290 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:50,854 INFO [train.py:904] (2/8) Epoch 9, batch 5750, loss[loss=0.218, simple_loss=0.2998, pruned_loss=0.06815, over 16991.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3179, pruned_loss=0.08251, over 3035469.15 frames. ], batch size: 53, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:48:46,690 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 01:49:12,523 INFO [train.py:904] (2/8) Epoch 9, batch 5800, loss[loss=0.2143, simple_loss=0.3034, pruned_loss=0.06262, over 16235.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3173, pruned_loss=0.0812, over 3022202.48 frames. ], batch size: 165, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:49:13,747 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:49:40,720 INFO [optim.py:368] (2/8) 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 01:50:23,997 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4714, 3.4390, 3.3851, 2.9430, 3.3906, 2.0255, 3.1841, 2.7428], device='cuda:2'), covar=tensor([0.0110, 0.0097, 0.0138, 0.0217, 0.0079, 0.1883, 0.0103, 0.0172], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0105, 0.0153, 0.0147, 0.0122, 0.0168, 0.0137, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:50:30,440 INFO [train.py:904] (2/8) Epoch 9, batch 5850, loss[loss=0.2524, simple_loss=0.3209, pruned_loss=0.09197, over 15510.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3149, pruned_loss=0.07927, over 3035356.55 frames. ], batch size: 190, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:50:47,532 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:51:42,081 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:51:50,009 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:51:50,871 INFO [train.py:904] (2/8) Epoch 9, batch 5900, loss[loss=0.2622, simple_loss=0.321, pruned_loss=0.1017, over 11613.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3139, pruned_loss=0.07806, over 3072204.96 frames. ], batch size: 246, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:52:22,906 INFO [optim.py:368] (2/8) 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 01:53:01,037 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:09,063 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:13,003 INFO [train.py:904] (2/8) Epoch 9, batch 5950, loss[loss=0.2439, simple_loss=0.3233, pruned_loss=0.08225, over 16572.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.315, pruned_loss=0.07671, over 3083528.26 frames. ], batch size: 68, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:53:42,054 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:54:01,842 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9387, 3.3171, 3.3727, 2.1675, 2.9871, 3.3366, 3.2321, 1.7212], device='cuda:2'), covar=tensor([0.0409, 0.0037, 0.0039, 0.0307, 0.0078, 0.0072, 0.0051, 0.0377], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0065, 0.0066, 0.0124, 0.0073, 0.0083, 0.0072, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 01:54:16,162 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 01:54:33,095 INFO [train.py:904] (2/8) Epoch 9, batch 6000, loss[loss=0.2124, simple_loss=0.2937, pruned_loss=0.06558, over 16896.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3135, pruned_loss=0.07612, over 3097800.95 frames. ], batch size: 109, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:54:33,096 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 01:54:44,296 INFO [train.py:938] (2/8) Epoch 9, validation: loss=0.1674, simple_loss=0.2809, pruned_loss=0.02692, over 944034.00 frames. 2023-04-29 01:54:44,296 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 01:55:11,800 INFO [optim.py:368] (2/8) 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 01:55:28,578 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:34,789 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:37,518 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:56:03,544 INFO [train.py:904] (2/8) Epoch 9, batch 6050, loss[loss=0.2387, simple_loss=0.3174, pruned_loss=0.08003, over 16438.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.312, pruned_loss=0.07557, over 3095758.49 frames. ], batch size: 146, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:56:12,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7683, 1.8271, 2.2184, 3.0702, 1.9889, 2.1207, 2.1312, 1.9767], device='cuda:2'), covar=tensor([0.1052, 0.3604, 0.1743, 0.0604, 0.3967, 0.2277, 0.2837, 0.3465], device='cuda:2'), in_proj_covar=tensor([0.0351, 0.0374, 0.0312, 0.0317, 0.0404, 0.0424, 0.0335, 0.0441], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 01:56:52,085 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:07,008 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:23,335 INFO [train.py:904] (2/8) Epoch 9, batch 6100, loss[loss=0.1959, simple_loss=0.2821, pruned_loss=0.05488, over 16834.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3107, pruned_loss=0.0745, over 3088291.36 frames. ], batch size: 42, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:57:52,498 INFO [optim.py:368] (2/8) 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 01:58:02,684 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-04-29 01:58:42,518 INFO [train.py:904] (2/8) Epoch 9, batch 6150, loss[loss=0.1989, simple_loss=0.2872, pruned_loss=0.05529, over 16566.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.308, pruned_loss=0.07276, over 3114096.08 frames. ], batch size: 68, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:58:52,851 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:58:58,718 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:00:00,628 INFO [train.py:904] (2/8) Epoch 9, batch 6200, loss[loss=0.2104, simple_loss=0.2951, pruned_loss=0.0628, over 16672.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3063, pruned_loss=0.07235, over 3120548.61 frames. ], batch size: 89, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 02:00:28,300 INFO [optim.py:368] (2/8) 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 02:00:34,891 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:18,823 INFO [train.py:904] (2/8) Epoch 9, batch 6250, loss[loss=0.2705, simple_loss=0.3368, pruned_loss=0.1021, over 11885.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.307, pruned_loss=0.07284, over 3120490.41 frames. ], batch size: 248, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:01:20,641 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7091, 4.0199, 3.6626, 2.1807, 3.2668, 2.8506, 3.9340, 4.1784], device='cuda:2'), covar=tensor([0.0196, 0.0524, 0.0635, 0.1673, 0.0706, 0.0835, 0.0525, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0140, 0.0159, 0.0143, 0.0136, 0.0126, 0.0136, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 02:01:32,090 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:45,535 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:24,261 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0055, 3.6553, 3.6630, 2.4159, 3.3332, 3.6068, 3.4984, 1.8716], device='cuda:2'), covar=tensor([0.0397, 0.0027, 0.0030, 0.0277, 0.0063, 0.0078, 0.0046, 0.0369], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0064, 0.0065, 0.0122, 0.0071, 0.0083, 0.0071, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 02:02:33,519 INFO [train.py:904] (2/8) Epoch 9, batch 6300, loss[loss=0.2596, simple_loss=0.3344, pruned_loss=0.09238, over 15479.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3068, pruned_loss=0.07257, over 3090281.07 frames. ], batch size: 190, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:02:38,945 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) 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 02:03:06,736 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:24,687 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:51,224 INFO [train.py:904] (2/8) Epoch 9, batch 6350, loss[loss=0.2281, simple_loss=0.308, pruned_loss=0.07409, over 16670.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3079, pruned_loss=0.07361, over 3083041.72 frames. ], batch size: 134, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:04:13,014 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:04:37,473 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:40,645 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:44,420 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:49,455 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:01,442 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5367, 3.1449, 2.9676, 1.8389, 2.6255, 2.0982, 3.0907, 3.1817], device='cuda:2'), covar=tensor([0.0282, 0.0629, 0.0580, 0.1875, 0.0853, 0.1002, 0.0643, 0.0823], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0139, 0.0156, 0.0141, 0.0134, 0.0125, 0.0135, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 02:05:08,846 INFO [train.py:904] (2/8) Epoch 9, batch 6400, loss[loss=0.194, simple_loss=0.2869, pruned_loss=0.05054, over 16718.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3075, pruned_loss=0.07395, over 3099835.42 frames. ], batch size: 83, lr: 7.56e-03, grad_scale: 8.0 2023-04-29 02:05:13,469 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:37,568 INFO [optim.py:368] (2/8) 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 02:06:15,005 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:06:21,670 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:06:24,600 INFO [train.py:904] (2/8) Epoch 9, batch 6450, loss[loss=0.2159, simple_loss=0.3054, pruned_loss=0.06315, over 16754.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3066, pruned_loss=0.07302, over 3098041.92 frames. ], batch size: 83, lr: 7.55e-03, grad_scale: 4.0 2023-04-29 02:06:34,297 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:06:46,426 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:07:41,885 INFO [train.py:904] (2/8) Epoch 9, batch 6500, loss[loss=0.2444, simple_loss=0.3021, pruned_loss=0.09341, over 11736.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3052, pruned_loss=0.07266, over 3090320.00 frames. ], batch size: 247, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:07:48,426 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:08:06,443 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:08:13,250 INFO [optim.py:368] (2/8) 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 02:08:38,770 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9859, 4.0900, 2.1774, 4.5722, 2.7761, 4.5239, 2.0047, 3.0219], device='cuda:2'), covar=tensor([0.0178, 0.0279, 0.1548, 0.0113, 0.0779, 0.0402, 0.1764, 0.0703], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0161, 0.0186, 0.0113, 0.0164, 0.0201, 0.0193, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 02:09:00,845 INFO [train.py:904] (2/8) Epoch 9, batch 6550, loss[loss=0.2273, simple_loss=0.32, pruned_loss=0.06731, over 16309.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3078, pruned_loss=0.07323, over 3122704.04 frames. ], batch size: 146, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:09:33,658 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8316, 1.8310, 2.2295, 3.0648, 2.0038, 2.1134, 2.0722, 1.8720], device='cuda:2'), covar=tensor([0.0963, 0.3579, 0.1758, 0.0563, 0.4132, 0.2353, 0.3012, 0.3783], device='cuda:2'), in_proj_covar=tensor([0.0351, 0.0374, 0.0314, 0.0318, 0.0405, 0.0426, 0.0338, 0.0442], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:09:52,142 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-29 02:09:56,064 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6577, 3.7545, 2.0789, 4.1511, 2.6298, 4.1517, 2.0040, 2.9023], device='cuda:2'), covar=tensor([0.0180, 0.0315, 0.1556, 0.0114, 0.0773, 0.0350, 0.1548, 0.0662], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0160, 0.0185, 0.0112, 0.0163, 0.0199, 0.0192, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 02:09:59,035 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1274, 4.2495, 4.0314, 3.9427, 3.5252, 4.1633, 3.9209, 3.7706], device='cuda:2'), covar=tensor([0.0658, 0.0448, 0.0328, 0.0272, 0.1067, 0.0413, 0.0732, 0.0614], device='cuda:2'), in_proj_covar=tensor([0.0223, 0.0270, 0.0260, 0.0236, 0.0284, 0.0274, 0.0179, 0.0305], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:10:18,518 INFO [train.py:904] (2/8) Epoch 9, batch 6600, loss[loss=0.2355, simple_loss=0.3178, pruned_loss=0.07663, over 16200.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3104, pruned_loss=0.07395, over 3131328.38 frames. ], batch size: 165, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:10:42,559 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:10:51,644 INFO [optim.py:368] (2/8) 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 02:10:54,943 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8156, 5.3542, 5.5163, 5.2774, 5.3274, 5.8784, 5.3760, 5.1963], device='cuda:2'), covar=tensor([0.1034, 0.1628, 0.1863, 0.1679, 0.2578, 0.1026, 0.1436, 0.2196], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0452, 0.0483, 0.0400, 0.0526, 0.0509, 0.0394, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 02:11:36,975 INFO [train.py:904] (2/8) Epoch 9, batch 6650, loss[loss=0.2228, simple_loss=0.3079, pruned_loss=0.06886, over 16228.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.31, pruned_loss=0.0737, over 3148288.51 frames. ], batch size: 165, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:11:42,439 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7333, 2.9510, 2.5670, 4.9707, 3.8195, 4.4083, 1.7611, 3.1116], device='cuda:2'), covar=tensor([0.1351, 0.0673, 0.1264, 0.0104, 0.0389, 0.0350, 0.1422, 0.0857], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0154, 0.0179, 0.0129, 0.0203, 0.0206, 0.0175, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 02:11:50,956 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:12:28,651 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:12:45,086 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3392, 3.3230, 3.3320, 3.4745, 3.5077, 3.2794, 3.4621, 3.5316], device='cuda:2'), covar=tensor([0.0950, 0.0762, 0.0942, 0.0498, 0.0608, 0.1643, 0.0805, 0.0633], device='cuda:2'), in_proj_covar=tensor([0.0485, 0.0598, 0.0739, 0.0610, 0.0467, 0.0464, 0.0488, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:12:53,899 INFO [train.py:904] (2/8) Epoch 9, batch 6700, loss[loss=0.2254, simple_loss=0.3059, pruned_loss=0.0724, over 16975.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3089, pruned_loss=0.0741, over 3136689.98 frames. ], batch size: 109, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:13:03,961 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.8689, 6.1460, 5.8332, 5.9795, 5.5202, 5.2953, 5.6827, 6.2770], device='cuda:2'), covar=tensor([0.0836, 0.0740, 0.1077, 0.0599, 0.0736, 0.0631, 0.0871, 0.0748], device='cuda:2'), in_proj_covar=tensor([0.0498, 0.0626, 0.0523, 0.0428, 0.0390, 0.0408, 0.0522, 0.0469], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:13:26,700 INFO [optim.py:368] (2/8) 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 02:13:43,771 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:13:43,936 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3652, 4.0197, 4.0344, 2.5778, 3.6543, 3.9645, 3.7380, 2.2298], device='cuda:2'), covar=tensor([0.0381, 0.0024, 0.0026, 0.0292, 0.0058, 0.0064, 0.0043, 0.0315], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0063, 0.0063, 0.0120, 0.0070, 0.0082, 0.0071, 0.0114], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 02:13:49,021 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4924, 3.4282, 3.8918, 1.8307, 4.0853, 4.0767, 2.9753, 2.9555], device='cuda:2'), covar=tensor([0.0728, 0.0182, 0.0125, 0.1182, 0.0043, 0.0074, 0.0343, 0.0425], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0099, 0.0085, 0.0140, 0.0066, 0.0096, 0.0119, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 02:13:52,494 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:14:00,200 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:14:10,406 INFO [train.py:904] (2/8) Epoch 9, batch 6750, loss[loss=0.2389, simple_loss=0.3143, pruned_loss=0.08168, over 16892.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3074, pruned_loss=0.07398, over 3127877.63 frames. ], batch size: 116, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:14:24,858 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:14:51,983 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4625, 4.6573, 4.7795, 4.6390, 4.6465, 5.2034, 4.7340, 4.4548], device='cuda:2'), covar=tensor([0.1167, 0.1578, 0.1550, 0.1718, 0.2142, 0.0934, 0.1361, 0.2501], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0449, 0.0482, 0.0399, 0.0524, 0.0510, 0.0391, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 02:15:29,295 INFO [train.py:904] (2/8) Epoch 9, batch 6800, loss[loss=0.2157, simple_loss=0.301, pruned_loss=0.06518, over 16614.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3075, pruned_loss=0.07378, over 3133230.57 frames. ], batch size: 134, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:15:54,179 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:16:01,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8462, 1.6167, 1.4420, 1.4333, 1.7845, 1.5410, 1.7243, 1.9195], device='cuda:2'), covar=tensor([0.0092, 0.0207, 0.0304, 0.0261, 0.0142, 0.0187, 0.0130, 0.0128], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0187, 0.0186, 0.0183, 0.0184, 0.0188, 0.0187, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:16:02,237 INFO [optim.py:368] (2/8) 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 02:16:45,308 INFO [train.py:904] (2/8) Epoch 9, batch 6850, loss[loss=0.2112, simple_loss=0.31, pruned_loss=0.05615, over 17281.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.31, pruned_loss=0.07516, over 3110564.27 frames. ], batch size: 52, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:17:02,729 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-29 02:17:06,648 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:17:59,091 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:01,001 INFO [train.py:904] (2/8) Epoch 9, batch 6900, loss[loss=0.2236, simple_loss=0.3171, pruned_loss=0.06507, over 16669.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3127, pruned_loss=0.07566, over 3104593.84 frames. ], batch size: 89, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:18:24,662 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:33,100 INFO [optim.py:368] (2/8) 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 02:19:18,075 INFO [train.py:904] (2/8) Epoch 9, batch 6950, loss[loss=0.2337, simple_loss=0.314, pruned_loss=0.0767, over 16300.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.314, pruned_loss=0.07691, over 3122306.74 frames. ], batch size: 35, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:19:32,118 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:19:32,197 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:19:38,178 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:20:31,302 INFO [train.py:904] (2/8) Epoch 9, batch 7000, loss[loss=0.2192, simple_loss=0.311, pruned_loss=0.0637, over 16666.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3146, pruned_loss=0.07689, over 3097376.38 frames. ], batch size: 62, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:20:36,677 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3653, 2.0122, 2.2091, 4.0505, 1.9530, 2.5490, 2.0878, 2.1982], device='cuda:2'), covar=tensor([0.0955, 0.3233, 0.1922, 0.0335, 0.3584, 0.2052, 0.2928, 0.2775], device='cuda:2'), in_proj_covar=tensor([0.0351, 0.0374, 0.0311, 0.0318, 0.0404, 0.0423, 0.0334, 0.0437], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:20:43,256 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:03,407 INFO [optim.py:368] (2/8) 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 02:21:12,419 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2033, 2.0109, 2.2169, 3.7592, 1.9482, 2.4569, 2.0945, 2.1789], device='cuda:2'), covar=tensor([0.0926, 0.3096, 0.1903, 0.0397, 0.3513, 0.2020, 0.2810, 0.2730], device='cuda:2'), in_proj_covar=tensor([0.0352, 0.0375, 0.0312, 0.0319, 0.0405, 0.0424, 0.0336, 0.0437], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:21:12,536 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-29 02:21:25,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3974, 2.8494, 2.6493, 2.2346, 2.2789, 2.1109, 2.7871, 2.8346], device='cuda:2'), covar=tensor([0.2042, 0.0803, 0.1338, 0.1821, 0.1758, 0.1777, 0.0463, 0.0904], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0254, 0.0280, 0.0270, 0.0279, 0.0214, 0.0263, 0.0281], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:21:29,012 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:35,934 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:47,200 INFO [train.py:904] (2/8) Epoch 9, batch 7050, loss[loss=0.254, simple_loss=0.3148, pruned_loss=0.09664, over 11870.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3157, pruned_loss=0.07692, over 3094986.37 frames. ], batch size: 250, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:22:00,347 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:22:40,450 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:22:45,475 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:22:47,888 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:02,083 INFO [train.py:904] (2/8) Epoch 9, batch 7100, loss[loss=0.2303, simple_loss=0.3076, pruned_loss=0.07644, over 15307.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3146, pruned_loss=0.07716, over 3067024.51 frames. ], batch size: 190, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:23:12,558 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:33,099 INFO [optim.py:368] (2/8) 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 02:23:43,576 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9900, 1.6742, 2.4053, 3.0527, 2.8349, 3.4040, 2.0061, 3.2346], device='cuda:2'), covar=tensor([0.0125, 0.0336, 0.0196, 0.0151, 0.0162, 0.0079, 0.0322, 0.0065], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0159, 0.0143, 0.0143, 0.0152, 0.0111, 0.0160, 0.0103], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 02:24:15,985 INFO [train.py:904] (2/8) Epoch 9, batch 7150, loss[loss=0.2813, simple_loss=0.3273, pruned_loss=0.1176, over 11273.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3123, pruned_loss=0.0768, over 3065420.72 frames. ], batch size: 246, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:24:17,092 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:25:28,098 INFO [train.py:904] (2/8) Epoch 9, batch 7200, loss[loss=0.211, simple_loss=0.2922, pruned_loss=0.06491, over 16723.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3097, pruned_loss=0.07444, over 3073417.59 frames. ], batch size: 62, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:00,148 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 9, batch 7250, loss[loss=0.2039, simple_loss=0.2792, pruned_loss=0.06434, over 16926.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3071, pruned_loss=0.07292, over 3078732.92 frames. ], batch size: 109, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:53,243 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:28:00,985 INFO [train.py:904] (2/8) Epoch 9, batch 7300, loss[loss=0.2039, simple_loss=0.2913, pruned_loss=0.0582, over 16453.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3057, pruned_loss=0.07226, over 3072235.58 frames. ], batch size: 146, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:28:33,453 INFO [optim.py:368] (2/8) 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 02:29:03,826 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 02:29:14,111 INFO [train.py:904] (2/8) Epoch 9, batch 7350, loss[loss=0.1947, simple_loss=0.2776, pruned_loss=0.05588, over 16536.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3058, pruned_loss=0.07227, over 3086267.97 frames. ], batch size: 57, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:30:27,926 INFO [train.py:904] (2/8) Epoch 9, batch 7400, loss[loss=0.2273, simple_loss=0.3138, pruned_loss=0.07039, over 16715.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.307, pruned_loss=0.07275, over 3086846.97 frames. ], batch size: 134, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:31:01,751 INFO [optim.py:368] (2/8) 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 02:31:38,444 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:31:44,621 INFO [train.py:904] (2/8) Epoch 9, batch 7450, loss[loss=0.217, simple_loss=0.2971, pruned_loss=0.06848, over 16868.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.308, pruned_loss=0.0736, over 3090505.16 frames. ], batch size: 116, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:03,515 INFO [train.py:904] (2/8) Epoch 9, batch 7500, loss[loss=0.2133, simple_loss=0.2955, pruned_loss=0.06558, over 16554.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3087, pruned_loss=0.07337, over 3088542.69 frames. ], batch size: 68, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:04,100 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9084, 2.6541, 2.5754, 1.9036, 2.5312, 2.5929, 2.5866, 1.8503], device='cuda:2'), covar=tensor([0.0305, 0.0047, 0.0042, 0.0248, 0.0078, 0.0067, 0.0059, 0.0294], device='cuda:2'), in_proj_covar=tensor([0.0124, 0.0063, 0.0065, 0.0123, 0.0072, 0.0082, 0.0072, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 02:33:22,927 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7820, 3.6158, 3.8184, 3.6682, 3.7464, 4.1716, 3.9152, 3.5558], device='cuda:2'), covar=tensor([0.1734, 0.2275, 0.1972, 0.2331, 0.2626, 0.1628, 0.1467, 0.2886], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0447, 0.0480, 0.0396, 0.0524, 0.0511, 0.0388, 0.0536], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 02:33:36,929 INFO [optim.py:368] (2/8) 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 02:34:18,158 INFO [train.py:904] (2/8) Epoch 9, batch 7550, loss[loss=0.2093, simple_loss=0.2989, pruned_loss=0.05988, over 16727.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3079, pruned_loss=0.07333, over 3090863.15 frames. ], batch size: 124, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:34:23,592 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:35:27,101 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 02:35:33,134 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 02:35:33,598 INFO [train.py:904] (2/8) Epoch 9, batch 7600, loss[loss=0.1944, simple_loss=0.2733, pruned_loss=0.05772, over 16312.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.307, pruned_loss=0.07357, over 3091522.54 frames. ], batch size: 35, lr: 7.51e-03, grad_scale: 8.0 2023-04-29 02:35:37,853 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:36:05,554 INFO [optim.py:368] (2/8) 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 02:36:45,917 INFO [train.py:904] (2/8) Epoch 9, batch 7650, loss[loss=0.2913, simple_loss=0.3425, pruned_loss=0.1201, over 11174.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3081, pruned_loss=0.07464, over 3087297.63 frames. ], batch size: 247, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:38:01,878 INFO [train.py:904] (2/8) Epoch 9, batch 7700, loss[loss=0.1995, simple_loss=0.2881, pruned_loss=0.05542, over 17221.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3084, pruned_loss=0.07542, over 3077796.62 frames. ], batch size: 45, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:38:09,047 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7964, 3.9649, 4.2856, 2.1165, 4.6314, 4.5937, 3.0048, 3.3285], device='cuda:2'), covar=tensor([0.0760, 0.0169, 0.0173, 0.1118, 0.0038, 0.0076, 0.0419, 0.0400], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0099, 0.0087, 0.0140, 0.0068, 0.0097, 0.0120, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 02:38:34,565 INFO [optim.py:368] (2/8) 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 02:39:08,821 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:39:16,835 INFO [train.py:904] (2/8) Epoch 9, batch 7750, loss[loss=0.2117, simple_loss=0.2952, pruned_loss=0.0641, over 16923.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3088, pruned_loss=0.0753, over 3073757.84 frames. ], batch size: 109, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:39:28,516 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1216, 4.1082, 4.5817, 4.4998, 4.5220, 4.1975, 4.2050, 4.0470], device='cuda:2'), covar=tensor([0.0318, 0.0507, 0.0312, 0.0459, 0.0425, 0.0342, 0.0943, 0.0509], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0304, 0.0312, 0.0300, 0.0347, 0.0326, 0.0431, 0.0264], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 02:39:36,464 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6687, 3.9262, 2.6711, 2.3729, 2.9587, 2.3149, 3.9436, 3.6061], device='cuda:2'), covar=tensor([0.2690, 0.0738, 0.1861, 0.1960, 0.2192, 0.1776, 0.0627, 0.0943], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0257, 0.0281, 0.0270, 0.0278, 0.0215, 0.0264, 0.0283], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:40:20,774 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:40:24,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1700, 5.5064, 5.2743, 5.2817, 4.8968, 4.7739, 4.9826, 5.6129], device='cuda:2'), covar=tensor([0.0901, 0.0786, 0.0942, 0.0646, 0.0766, 0.0758, 0.0905, 0.0879], device='cuda:2'), in_proj_covar=tensor([0.0489, 0.0622, 0.0525, 0.0422, 0.0384, 0.0405, 0.0516, 0.0468], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:40:29,342 INFO [train.py:904] (2/8) Epoch 9, batch 7800, loss[loss=0.2722, simple_loss=0.3275, pruned_loss=0.1085, over 11280.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3093, pruned_loss=0.07567, over 3091285.85 frames. ], batch size: 248, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:41:02,672 INFO [optim.py:368] (2/8) 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 02:41:07,716 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:41:10,225 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4755, 4.4323, 4.3096, 3.2916, 4.3388, 1.3568, 4.0453, 4.0516], device='cuda:2'), covar=tensor([0.0108, 0.0093, 0.0158, 0.0531, 0.0108, 0.2877, 0.0150, 0.0274], device='cuda:2'), in_proj_covar=tensor([0.0116, 0.0102, 0.0150, 0.0145, 0.0118, 0.0166, 0.0133, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:41:44,457 INFO [train.py:904] (2/8) Epoch 9, batch 7850, loss[loss=0.2139, simple_loss=0.3023, pruned_loss=0.06279, over 16639.00 frames. ], tot_loss[loss=0.23, simple_loss=0.31, pruned_loss=0.07497, over 3103672.16 frames. ], batch size: 89, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:42:38,966 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:42:57,226 INFO [train.py:904] (2/8) Epoch 9, batch 7900, loss[loss=0.3032, simple_loss=0.3526, pruned_loss=0.127, over 11735.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3096, pruned_loss=0.07515, over 3084994.20 frames. ], batch size: 250, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:43:28,442 INFO [optim.py:368] (2/8) 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 02:44:12,924 INFO [train.py:904] (2/8) Epoch 9, batch 7950, loss[loss=0.2618, simple_loss=0.334, pruned_loss=0.09479, over 15400.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3106, pruned_loss=0.07633, over 3059692.24 frames. ], batch size: 191, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:44:15,116 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8427, 4.1373, 3.9111, 3.9363, 3.6788, 3.7687, 3.8087, 4.0956], device='cuda:2'), covar=tensor([0.1004, 0.0835, 0.0991, 0.0727, 0.0755, 0.1518, 0.0815, 0.0958], device='cuda:2'), in_proj_covar=tensor([0.0487, 0.0619, 0.0524, 0.0423, 0.0383, 0.0405, 0.0518, 0.0465], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:44:39,069 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:44:44,142 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8884, 2.0330, 2.3596, 3.1571, 2.1893, 2.2855, 2.2471, 2.1321], device='cuda:2'), covar=tensor([0.0793, 0.2743, 0.1570, 0.0456, 0.2988, 0.1856, 0.2225, 0.2632], device='cuda:2'), in_proj_covar=tensor([0.0349, 0.0377, 0.0314, 0.0317, 0.0407, 0.0427, 0.0337, 0.0439], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:44:54,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6954, 3.3244, 3.0387, 1.7921, 2.7187, 2.1674, 3.2615, 3.3663], device='cuda:2'), covar=tensor([0.0295, 0.0592, 0.0622, 0.1834, 0.0814, 0.0956, 0.0683, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0137, 0.0154, 0.0140, 0.0132, 0.0123, 0.0135, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 02:45:26,764 INFO [train.py:904] (2/8) Epoch 9, batch 8000, loss[loss=0.2367, simple_loss=0.3178, pruned_loss=0.07782, over 15329.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3117, pruned_loss=0.07804, over 3032107.74 frames. ], batch size: 190, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:45:59,520 INFO [optim.py:368] (2/8) 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 02:46:01,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4109, 2.8778, 2.6589, 2.3323, 2.3564, 2.1863, 2.8517, 2.9000], device='cuda:2'), covar=tensor([0.2035, 0.0680, 0.1352, 0.1640, 0.1800, 0.1754, 0.0455, 0.0901], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0257, 0.0282, 0.0271, 0.0279, 0.0216, 0.0263, 0.0283], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:46:09,133 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:46:40,073 INFO [train.py:904] (2/8) Epoch 9, batch 8050, loss[loss=0.2236, simple_loss=0.3013, pruned_loss=0.07297, over 15318.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3117, pruned_loss=0.07785, over 3028753.10 frames. ], batch size: 190, lr: 7.49e-03, grad_scale: 4.0 2023-04-29 02:47:55,803 INFO [train.py:904] (2/8) Epoch 9, batch 8100, loss[loss=0.1802, simple_loss=0.2667, pruned_loss=0.04691, over 17214.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3108, pruned_loss=0.07664, over 3047759.90 frames. ], batch size: 52, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:48:29,435 INFO [optim.py:368] (2/8) 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 02:49:10,971 INFO [train.py:904] (2/8) Epoch 9, batch 8150, loss[loss=0.2078, simple_loss=0.2903, pruned_loss=0.0626, over 16195.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3086, pruned_loss=0.07547, over 3074399.57 frames. ], batch size: 165, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:49:45,531 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:49:59,363 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:50:28,007 INFO [train.py:904] (2/8) Epoch 9, batch 8200, loss[loss=0.2165, simple_loss=0.3041, pruned_loss=0.06443, over 16376.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3062, pruned_loss=0.07509, over 3050552.70 frames. ], batch size: 146, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:51:05,354 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:06,153 INFO [optim.py:368] (2/8) 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 02:51:20,844 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:27,390 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7424, 3.6104, 3.8814, 3.6124, 3.8312, 4.2270, 3.9397, 3.6129], device='cuda:2'), covar=tensor([0.2150, 0.2261, 0.1706, 0.2420, 0.2582, 0.1423, 0.1415, 0.2569], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0455, 0.0483, 0.0400, 0.0522, 0.0511, 0.0392, 0.0536], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 02:51:27,494 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:51:47,433 INFO [train.py:904] (2/8) Epoch 9, batch 8250, loss[loss=0.2138, simple_loss=0.313, pruned_loss=0.05732, over 16754.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3056, pruned_loss=0.07282, over 3039236.89 frames. ], batch size: 102, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:51:53,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5764, 4.0407, 4.0642, 2.3297, 3.4611, 2.7470, 3.9412, 3.8699], device='cuda:2'), covar=tensor([0.0224, 0.0573, 0.0401, 0.1581, 0.0623, 0.0809, 0.0557, 0.0889], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0138, 0.0156, 0.0141, 0.0133, 0.0125, 0.0135, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 02:52:43,263 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:06,325 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:53:08,808 INFO [train.py:904] (2/8) Epoch 9, batch 8300, loss[loss=0.2001, simple_loss=0.2769, pruned_loss=0.06166, over 12222.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3017, pruned_loss=0.06905, over 3024698.37 frames. ], batch size: 248, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:53:19,073 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2368, 2.3249, 2.0176, 2.0516, 2.7216, 2.4915, 3.0822, 2.9421], device='cuda:2'), covar=tensor([0.0070, 0.0288, 0.0362, 0.0356, 0.0185, 0.0254, 0.0130, 0.0179], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0185, 0.0185, 0.0183, 0.0184, 0.0185, 0.0185, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:53:21,228 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:24,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8405, 4.7460, 4.6586, 4.3827, 4.2725, 4.7443, 4.5941, 4.4099], device='cuda:2'), covar=tensor([0.0478, 0.0525, 0.0262, 0.0240, 0.0922, 0.0394, 0.0306, 0.0622], device='cuda:2'), in_proj_covar=tensor([0.0223, 0.0271, 0.0260, 0.0240, 0.0279, 0.0270, 0.0179, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 02:53:47,008 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:48,528 INFO [optim.py:368] (2/8) 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 02:54:31,537 INFO [train.py:904] (2/8) Epoch 9, batch 8350, loss[loss=0.1974, simple_loss=0.2862, pruned_loss=0.0543, over 15459.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2996, pruned_loss=0.06593, over 3028907.97 frames. ], batch size: 191, lr: 7.47e-03, grad_scale: 2.0 2023-04-29 02:55:01,146 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:55:52,199 INFO [train.py:904] (2/8) Epoch 9, batch 8400, loss[loss=0.2022, simple_loss=0.2897, pruned_loss=0.05735, over 16753.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2963, pruned_loss=0.06288, over 3047024.01 frames. ], batch size: 124, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:56:31,315 INFO [optim.py:368] (2/8) 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 02:57:13,775 INFO [train.py:904] (2/8) Epoch 9, batch 8450, loss[loss=0.1747, simple_loss=0.2746, pruned_loss=0.03742, over 16839.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2945, pruned_loss=0.06138, over 3035294.70 frames. ], batch size: 96, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:57:59,610 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:04,776 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:34,963 INFO [train.py:904] (2/8) Epoch 9, batch 8500, loss[loss=0.1859, simple_loss=0.2756, pruned_loss=0.04809, over 16801.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2904, pruned_loss=0.05869, over 3030219.58 frames. ], batch size: 102, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:59:01,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1077, 2.6251, 2.7107, 1.9379, 2.8640, 2.9124, 2.5280, 2.5032], device='cuda:2'), covar=tensor([0.0596, 0.0162, 0.0178, 0.0881, 0.0073, 0.0138, 0.0338, 0.0325], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0094, 0.0082, 0.0135, 0.0065, 0.0090, 0.0114, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-29 02:59:14,097 INFO [optim.py:368] (2/8) 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 02:59:20,521 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:38,911 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:56,042 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 02:59:58,598 INFO [train.py:904] (2/8) Epoch 9, batch 8550, loss[loss=0.2056, simple_loss=0.2824, pruned_loss=0.06439, over 11819.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2878, pruned_loss=0.05739, over 3030647.65 frames. ], batch size: 246, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 03:00:01,761 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4047, 3.3464, 3.4342, 3.5247, 3.5598, 3.2739, 3.5378, 3.5779], device='cuda:2'), covar=tensor([0.1034, 0.0769, 0.0895, 0.0502, 0.0540, 0.2179, 0.0669, 0.0591], device='cuda:2'), in_proj_covar=tensor([0.0470, 0.0580, 0.0704, 0.0587, 0.0454, 0.0454, 0.0470, 0.0529], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:00:51,084 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7123, 2.1465, 2.2260, 4.4185, 2.0088, 2.6669, 2.1778, 2.3195], device='cuda:2'), covar=tensor([0.0665, 0.3145, 0.2106, 0.0300, 0.3889, 0.2037, 0.2935, 0.3202], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0368, 0.0309, 0.0309, 0.0398, 0.0412, 0.0328, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:00:54,503 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:01:24,773 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:01:37,952 INFO [train.py:904] (2/8) Epoch 9, batch 8600, loss[loss=0.2331, simple_loss=0.3176, pruned_loss=0.07429, over 15400.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.288, pruned_loss=0.05666, over 3019162.85 frames. ], batch size: 190, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:02:25,684 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:02:26,425 INFO [optim.py:368] (2/8) 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 03:03:04,618 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4286, 1.8247, 1.6046, 1.6255, 2.1837, 1.9330, 2.1418, 2.3026], device='cuda:2'), covar=tensor([0.0076, 0.0289, 0.0355, 0.0359, 0.0181, 0.0265, 0.0122, 0.0186], device='cuda:2'), in_proj_covar=tensor([0.0120, 0.0188, 0.0186, 0.0187, 0.0186, 0.0187, 0.0184, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:03:15,781 INFO [train.py:904] (2/8) Epoch 9, batch 8650, loss[loss=0.1835, simple_loss=0.2767, pruned_loss=0.04513, over 16360.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2861, pruned_loss=0.05504, over 3026083.84 frames. ], batch size: 146, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:03:47,304 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:47,462 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:55,348 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9415, 2.3264, 2.1984, 2.9048, 1.9869, 3.2879, 1.6026, 2.7803], device='cuda:2'), covar=tensor([0.1240, 0.0588, 0.1003, 0.0121, 0.0092, 0.0335, 0.1448, 0.0637], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0150, 0.0174, 0.0124, 0.0192, 0.0202, 0.0173, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 03:04:05,356 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:04:56,959 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5786, 3.5778, 3.5379, 2.9720, 3.4763, 1.9947, 3.2589, 3.0500], device='cuda:2'), covar=tensor([0.0112, 0.0090, 0.0108, 0.0177, 0.0072, 0.1947, 0.0100, 0.0175], device='cuda:2'), in_proj_covar=tensor([0.0114, 0.0100, 0.0146, 0.0138, 0.0117, 0.0164, 0.0131, 0.0136], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:05:02,348 INFO [train.py:904] (2/8) Epoch 9, batch 8700, loss[loss=0.2065, simple_loss=0.2839, pruned_loss=0.06453, over 12370.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2829, pruned_loss=0.05314, over 3047126.00 frames. ], batch size: 248, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:05:18,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0243, 1.6714, 1.5591, 1.4059, 1.8956, 1.5930, 1.7549, 1.9173], device='cuda:2'), covar=tensor([0.0105, 0.0205, 0.0274, 0.0275, 0.0154, 0.0210, 0.0109, 0.0166], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0187, 0.0185, 0.0185, 0.0185, 0.0186, 0.0181, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:05:45,060 INFO [optim.py:368] (2/8) 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 03:05:45,846 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:05:47,075 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3055, 3.4463, 3.6648, 3.6446, 3.6296, 3.4319, 3.5145, 3.5205], device='cuda:2'), covar=tensor([0.0320, 0.0506, 0.0373, 0.0454, 0.0490, 0.0370, 0.0655, 0.0363], device='cuda:2'), in_proj_covar=tensor([0.0289, 0.0292, 0.0297, 0.0287, 0.0335, 0.0310, 0.0407, 0.0251], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-29 03:06:36,264 INFO [train.py:904] (2/8) Epoch 9, batch 8750, loss[loss=0.1948, simple_loss=0.2869, pruned_loss=0.0513, over 16475.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2821, pruned_loss=0.05252, over 3040108.80 frames. ], batch size: 62, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:06:55,531 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8591, 5.1879, 5.3319, 5.2815, 5.3025, 5.7676, 5.3500, 5.0553], device='cuda:2'), covar=tensor([0.0764, 0.1537, 0.1543, 0.1524, 0.2058, 0.0811, 0.0953, 0.2056], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0440, 0.0470, 0.0383, 0.0502, 0.0493, 0.0383, 0.0514], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 03:07:19,867 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 03:07:34,190 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7844, 2.9087, 2.9483, 4.6579, 3.3324, 4.4449, 1.4308, 3.3527], device='cuda:2'), covar=tensor([0.1302, 0.0656, 0.0922, 0.0089, 0.0169, 0.0245, 0.1549, 0.0581], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0149, 0.0173, 0.0123, 0.0189, 0.0200, 0.0171, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 03:08:32,638 INFO [train.py:904] (2/8) Epoch 9, batch 8800, loss[loss=0.1865, simple_loss=0.2801, pruned_loss=0.04646, over 15425.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.281, pruned_loss=0.05121, over 3052611.12 frames. ], batch size: 191, lr: 7.46e-03, grad_scale: 8.0 2023-04-29 03:08:57,425 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 03:09:21,968 INFO [optim.py:368] (2/8) 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 03:09:31,803 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:44,355 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:10:17,857 INFO [train.py:904] (2/8) Epoch 9, batch 8850, loss[loss=0.194, simple_loss=0.289, pruned_loss=0.04952, over 16771.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2826, pruned_loss=0.05043, over 3048685.53 frames. ], batch size: 124, lr: 7.45e-03, grad_scale: 8.0 2023-04-29 03:11:11,513 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:18,284 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:49,570 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:11:59,876 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5442, 3.5403, 3.4951, 2.9486, 3.4984, 1.8783, 3.3086, 3.0074], device='cuda:2'), covar=tensor([0.0100, 0.0088, 0.0122, 0.0200, 0.0079, 0.2079, 0.0106, 0.0191], device='cuda:2'), in_proj_covar=tensor([0.0115, 0.0102, 0.0148, 0.0140, 0.0119, 0.0168, 0.0134, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:12:02,853 INFO [train.py:904] (2/8) Epoch 9, batch 8900, loss[loss=0.2008, simple_loss=0.2995, pruned_loss=0.05107, over 16482.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2827, pruned_loss=0.0499, over 3030372.32 frames. ], batch size: 62, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:12:35,684 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3119, 1.6238, 2.5663, 3.2732, 2.9394, 3.5392, 1.8078, 3.4869], device='cuda:2'), covar=tensor([0.0071, 0.0427, 0.0187, 0.0122, 0.0154, 0.0085, 0.0388, 0.0072], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0161, 0.0145, 0.0144, 0.0154, 0.0110, 0.0162, 0.0100], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 03:12:49,177 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-29 03:12:57,513 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:07,661 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0009, 3.5334, 2.7438, 5.1324, 4.1004, 4.6381, 1.5869, 3.4839], device='cuda:2'), covar=tensor([0.1291, 0.0528, 0.1117, 0.0124, 0.0291, 0.0280, 0.1453, 0.0616], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0151, 0.0175, 0.0124, 0.0189, 0.0201, 0.0173, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 03:13:10,495 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:48,654 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:13:53,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7329, 1.5424, 2.0920, 2.7197, 2.5509, 2.9066, 1.8989, 2.8429], device='cuda:2'), covar=tensor([0.0109, 0.0386, 0.0254, 0.0154, 0.0200, 0.0117, 0.0345, 0.0086], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0161, 0.0145, 0.0144, 0.0154, 0.0110, 0.0162, 0.0100], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 03:14:08,705 INFO [train.py:904] (2/8) Epoch 9, batch 8950, loss[loss=0.1899, simple_loss=0.2796, pruned_loss=0.05013, over 15538.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2815, pruned_loss=0.04936, over 3060559.59 frames. ], batch size: 193, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:14:15,264 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 03:14:38,299 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:15:35,055 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:15:57,360 INFO [train.py:904] (2/8) Epoch 9, batch 9000, loss[loss=0.181, simple_loss=0.2683, pruned_loss=0.04688, over 16239.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2785, pruned_loss=0.04816, over 3053431.84 frames. ], batch size: 165, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:15:57,360 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 03:16:07,526 INFO [train.py:938] (2/8) Epoch 9, validation: loss=0.1581, simple_loss=0.2623, pruned_loss=0.02697, over 944034.00 frames. 2023-04-29 03:16:07,527 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 03:16:31,270 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:16:31,859 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-29 03:16:42,292 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7499, 3.8745, 4.1831, 1.9413, 4.4104, 4.4399, 3.0464, 3.1752], device='cuda:2'), covar=tensor([0.0673, 0.0164, 0.0127, 0.1116, 0.0036, 0.0078, 0.0329, 0.0399], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0095, 0.0081, 0.0137, 0.0065, 0.0091, 0.0116, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-29 03:16:47,982 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:16:58,705 INFO [optim.py:368] (2/8) 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 03:17:05,333 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 03:17:08,273 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:17:51,878 INFO [train.py:904] (2/8) Epoch 9, batch 9050, loss[loss=0.1766, simple_loss=0.2629, pruned_loss=0.04514, over 15484.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2796, pruned_loss=0.04899, over 3054715.87 frames. ], batch size: 192, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:18:00,929 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5454, 4.5333, 4.3734, 4.0912, 4.0255, 4.4856, 4.3255, 4.1372], device='cuda:2'), covar=tensor([0.0449, 0.0399, 0.0264, 0.0229, 0.0734, 0.0350, 0.0326, 0.0505], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0261, 0.0252, 0.0233, 0.0268, 0.0262, 0.0172, 0.0291], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:19:19,200 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:19:36,841 INFO [train.py:904] (2/8) Epoch 9, batch 9100, loss[loss=0.169, simple_loss=0.2574, pruned_loss=0.04029, over 12457.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2793, pruned_loss=0.04924, over 3069665.59 frames. ], batch size: 248, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:20:34,109 INFO [optim.py:368] (2/8) 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 03:20:48,714 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 03:20:58,250 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:33,808 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:35,357 INFO [train.py:904] (2/8) Epoch 9, batch 9150, loss[loss=0.1797, simple_loss=0.2705, pruned_loss=0.04443, over 15465.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2797, pruned_loss=0.04871, over 3069066.78 frames. ], batch size: 192, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:22:44,951 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:22:55,889 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7833, 3.8120, 4.0049, 1.8951, 4.2217, 4.2838, 3.1328, 3.0351], device='cuda:2'), covar=tensor([0.0650, 0.0179, 0.0187, 0.1117, 0.0042, 0.0080, 0.0340, 0.0405], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0094, 0.0081, 0.0135, 0.0065, 0.0090, 0.0115, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:2') 2023-04-29 03:23:22,760 INFO [train.py:904] (2/8) Epoch 9, batch 9200, loss[loss=0.1568, simple_loss=0.2506, pruned_loss=0.0315, over 16617.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2748, pruned_loss=0.04725, over 3087007.43 frames. ], batch size: 75, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:23:28,068 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1951, 4.2075, 4.4038, 4.2259, 4.2789, 4.7977, 4.4014, 4.0693], device='cuda:2'), covar=tensor([0.1500, 0.1770, 0.1472, 0.1940, 0.2428, 0.0967, 0.1372, 0.2285], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0436, 0.0461, 0.0383, 0.0504, 0.0488, 0.0379, 0.0507], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:23:42,376 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90412.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:24:07,625 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 9, batch 9250, loss[loss=0.1708, simple_loss=0.2646, pruned_loss=0.03851, over 15303.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2748, pruned_loss=0.04745, over 3072901.62 frames. ], batch size: 191, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:25:14,868 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:25:52,840 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 03:26:12,490 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:38,482 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:49,730 INFO [train.py:904] (2/8) Epoch 9, batch 9300, loss[loss=0.1845, simple_loss=0.2734, pruned_loss=0.04781, over 16274.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2734, pruned_loss=0.047, over 3067738.50 frames. ], batch size: 165, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:27:35,093 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:27:36,986 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 03:27:46,915 INFO [optim.py:368] (2/8) 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 03:28:36,077 INFO [train.py:904] (2/8) Epoch 9, batch 9350, loss[loss=0.1793, simple_loss=0.2628, pruned_loss=0.04793, over 12212.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2732, pruned_loss=0.04676, over 3075662.69 frames. ], batch size: 248, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:28:48,619 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:13,851 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:50,762 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:30:05,917 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1086, 1.4269, 1.8249, 2.0679, 2.1778, 2.2043, 1.6297, 2.2317], device='cuda:2'), covar=tensor([0.0151, 0.0314, 0.0194, 0.0197, 0.0185, 0.0144, 0.0312, 0.0089], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0160, 0.0144, 0.0143, 0.0154, 0.0110, 0.0161, 0.0100], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 03:30:18,678 INFO [train.py:904] (2/8) Epoch 9, batch 9400, loss[loss=0.1807, simple_loss=0.2868, pruned_loss=0.0373, over 16677.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.273, pruned_loss=0.04625, over 3078292.03 frames. ], batch size: 83, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:31:09,564 INFO [optim.py:368] (2/8) 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 03:31:51,727 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9382, 1.7508, 1.5621, 1.4857, 1.9268, 1.5773, 1.7261, 1.9548], device='cuda:2'), covar=tensor([0.0074, 0.0200, 0.0286, 0.0276, 0.0148, 0.0217, 0.0110, 0.0143], device='cuda:2'), in_proj_covar=tensor([0.0118, 0.0188, 0.0185, 0.0184, 0.0182, 0.0186, 0.0177, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:32:00,903 INFO [train.py:904] (2/8) Epoch 9, batch 9450, loss[loss=0.1961, simple_loss=0.28, pruned_loss=0.05611, over 12709.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2749, pruned_loss=0.04682, over 3065074.98 frames. ], batch size: 248, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:32:09,242 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2860, 3.6499, 3.7628, 2.4225, 3.4117, 3.7195, 3.6040, 2.2750], device='cuda:2'), covar=tensor([0.0365, 0.0029, 0.0023, 0.0292, 0.0063, 0.0053, 0.0044, 0.0322], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0062, 0.0064, 0.0121, 0.0072, 0.0080, 0.0071, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 03:33:38,987 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 03:33:43,968 INFO [train.py:904] (2/8) Epoch 9, batch 9500, loss[loss=0.1709, simple_loss=0.2607, pruned_loss=0.04051, over 16631.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2742, pruned_loss=0.04652, over 3079867.14 frames. ], batch size: 57, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:33:58,229 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:34:35,314 INFO [optim.py:368] (2/8) 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 03:35:22,706 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8930, 1.7796, 1.5672, 1.4635, 1.8978, 1.5793, 1.7487, 1.9529], device='cuda:2'), covar=tensor([0.0064, 0.0194, 0.0265, 0.0252, 0.0128, 0.0204, 0.0109, 0.0128], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0188, 0.0184, 0.0182, 0.0182, 0.0185, 0.0178, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:35:30,038 INFO [train.py:904] (2/8) Epoch 9, batch 9550, loss[loss=0.2043, simple_loss=0.2964, pruned_loss=0.05609, over 16249.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2743, pruned_loss=0.04704, over 3065524.02 frames. ], batch size: 165, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:36:43,090 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:11,179 INFO [train.py:904] (2/8) Epoch 9, batch 9600, loss[loss=0.2034, simple_loss=0.2969, pruned_loss=0.05492, over 16168.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2752, pruned_loss=0.04795, over 3025222.27 frames. ], batch size: 165, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:37:37,097 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:37,439 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-29 03:37:59,117 INFO [optim.py:368] (2/8) 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 03:38:17,844 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:38:55,290 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:39:01,494 INFO [train.py:904] (2/8) Epoch 9, batch 9650, loss[loss=0.1796, simple_loss=0.2664, pruned_loss=0.04634, over 16628.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2773, pruned_loss=0.04818, over 3032802.99 frames. ], batch size: 57, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:39:03,429 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:40:18,918 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:40:48,119 INFO [train.py:904] (2/8) Epoch 9, batch 9700, loss[loss=0.1754, simple_loss=0.2607, pruned_loss=0.04509, over 12344.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2764, pruned_loss=0.04782, over 3035740.44 frames. ], batch size: 248, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:41:04,608 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:41:40,521 INFO [optim.py:368] (2/8) 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 03:41:59,658 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:42:08,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9713, 3.8271, 3.9884, 4.1739, 4.2443, 3.8377, 4.2517, 4.2352], device='cuda:2'), covar=tensor([0.1306, 0.1024, 0.1317, 0.0576, 0.0523, 0.1202, 0.0513, 0.0560], device='cuda:2'), in_proj_covar=tensor([0.0452, 0.0561, 0.0676, 0.0560, 0.0432, 0.0436, 0.0453, 0.0509], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:42:31,488 INFO [train.py:904] (2/8) Epoch 9, batch 9750, loss[loss=0.1741, simple_loss=0.2569, pruned_loss=0.04566, over 12374.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2753, pruned_loss=0.04769, over 3044813.70 frames. ], batch size: 246, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:43:06,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8331, 3.7375, 3.9049, 3.7128, 3.8830, 4.2352, 3.9383, 3.6301], device='cuda:2'), covar=tensor([0.1989, 0.2420, 0.2169, 0.2500, 0.2759, 0.1676, 0.1560, 0.2874], device='cuda:2'), in_proj_covar=tensor([0.0306, 0.0435, 0.0466, 0.0381, 0.0502, 0.0486, 0.0382, 0.0504], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:43:38,913 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7700, 5.0374, 4.8170, 4.8257, 4.4827, 4.4963, 4.4699, 5.0883], device='cuda:2'), covar=tensor([0.0895, 0.0770, 0.0895, 0.0662, 0.0794, 0.0861, 0.0901, 0.0801], device='cuda:2'), in_proj_covar=tensor([0.0477, 0.0606, 0.0494, 0.0414, 0.0376, 0.0394, 0.0502, 0.0447], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:44:10,866 INFO [train.py:904] (2/8) Epoch 9, batch 9800, loss[loss=0.1753, simple_loss=0.2604, pruned_loss=0.04511, over 12077.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2763, pruned_loss=0.04685, over 3056395.68 frames. ], batch size: 248, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:44:21,887 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:44:24,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5263, 3.5675, 3.2962, 3.1021, 3.1365, 3.4769, 3.2550, 3.2537], device='cuda:2'), covar=tensor([0.0503, 0.0444, 0.0239, 0.0237, 0.0532, 0.0331, 0.1045, 0.0446], device='cuda:2'), in_proj_covar=tensor([0.0211, 0.0255, 0.0246, 0.0228, 0.0264, 0.0256, 0.0169, 0.0285], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-29 03:44:57,914 INFO [optim.py:368] (2/8) 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 03:45:57,974 INFO [train.py:904] (2/8) Epoch 9, batch 9850, loss[loss=0.1719, simple_loss=0.2687, pruned_loss=0.03761, over 16698.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2773, pruned_loss=0.04664, over 3058284.89 frames. ], batch size: 83, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:46:04,899 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:47:24,593 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8616, 3.8866, 3.9976, 3.8933, 3.9586, 4.3224, 4.0041, 3.7136], device='cuda:2'), covar=tensor([0.1785, 0.1821, 0.1208, 0.1949, 0.2142, 0.1139, 0.1245, 0.2210], device='cuda:2'), in_proj_covar=tensor([0.0306, 0.0436, 0.0464, 0.0382, 0.0499, 0.0485, 0.0379, 0.0506], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:47:48,707 INFO [train.py:904] (2/8) Epoch 9, batch 9900, loss[loss=0.1711, simple_loss=0.2557, pruned_loss=0.04323, over 12532.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2779, pruned_loss=0.04665, over 3057413.18 frames. ], batch size: 250, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:47:52,779 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 03:48:14,971 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1866, 4.2363, 4.6075, 4.5729, 4.5613, 4.3410, 4.2912, 4.1022], device='cuda:2'), covar=tensor([0.0257, 0.0448, 0.0305, 0.0323, 0.0348, 0.0283, 0.0649, 0.0377], device='cuda:2'), in_proj_covar=tensor([0.0289, 0.0289, 0.0294, 0.0281, 0.0328, 0.0309, 0.0397, 0.0250], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-29 03:48:20,526 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:48:47,932 INFO [optim.py:368] (2/8) 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 03:48:59,546 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2978, 4.2981, 4.1507, 3.7048, 4.1669, 1.7170, 3.9533, 3.9848], device='cuda:2'), covar=tensor([0.0072, 0.0067, 0.0122, 0.0225, 0.0076, 0.2179, 0.0117, 0.0171], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0099, 0.0144, 0.0131, 0.0115, 0.0165, 0.0130, 0.0134], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-29 03:49:08,449 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2671, 2.4206, 2.0803, 2.0089, 2.8427, 2.4850, 3.0951, 2.9735], device='cuda:2'), covar=tensor([0.0054, 0.0310, 0.0335, 0.0356, 0.0150, 0.0254, 0.0139, 0.0141], device='cuda:2'), in_proj_covar=tensor([0.0117, 0.0190, 0.0184, 0.0184, 0.0184, 0.0187, 0.0178, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:49:47,991 INFO [train.py:904] (2/8) Epoch 9, batch 9950, loss[loss=0.1652, simple_loss=0.2615, pruned_loss=0.03446, over 16763.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2795, pruned_loss=0.0471, over 3052038.04 frames. ], batch size: 83, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:49:49,272 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:14,537 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:29,702 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4189, 5.4712, 5.1537, 4.8494, 5.1898, 1.7456, 4.9079, 5.1382], device='cuda:2'), covar=tensor([0.0050, 0.0041, 0.0112, 0.0192, 0.0053, 0.2088, 0.0085, 0.0127], device='cuda:2'), in_proj_covar=tensor([0.0112, 0.0099, 0.0144, 0.0131, 0.0115, 0.0166, 0.0131, 0.0134], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-29 03:51:44,349 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:51:47,236 INFO [train.py:904] (2/8) Epoch 9, batch 10000, loss[loss=0.2056, simple_loss=0.309, pruned_loss=0.05111, over 16729.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2771, pruned_loss=0.04616, over 3082247.98 frames. ], batch size: 134, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:51:53,876 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 9, batch 10050, loss[loss=0.1782, simple_loss=0.2673, pruned_loss=0.04455, over 12050.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2769, pruned_loss=0.04608, over 3080371.19 frames. ], batch size: 250, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:54:07,076 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7892, 1.1853, 1.5613, 1.6153, 1.7258, 1.8488, 1.5531, 1.8854], device='cuda:2'), covar=tensor([0.0175, 0.0291, 0.0153, 0.0183, 0.0180, 0.0152, 0.0280, 0.0091], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0164, 0.0149, 0.0146, 0.0158, 0.0112, 0.0166, 0.0101], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 03:54:30,375 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91285.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:55:01,275 INFO [train.py:904] (2/8) Epoch 9, batch 10100, loss[loss=0.181, simple_loss=0.273, pruned_loss=0.04448, over 16418.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2772, pruned_loss=0.0464, over 3076375.04 frames. ], batch size: 35, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:55:52,059 INFO [optim.py:368] (2/8) 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 03:56:14,330 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:56:45,259 INFO [train.py:904] (2/8) Epoch 10, batch 0, loss[loss=0.275, simple_loss=0.3186, pruned_loss=0.1157, over 16828.00 frames. ], tot_loss[loss=0.275, simple_loss=0.3186, pruned_loss=0.1157, over 16828.00 frames. ], batch size: 83, lr: 7.04e-03, grad_scale: 8.0 2023-04-29 03:56:45,260 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 03:56:50,626 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([6.2231, 6.5609, 6.2712, 6.4436, 6.0686, 6.0549, 6.1363, 6.5454], device='cuda:2'), covar=tensor([0.0672, 0.0575, 0.0561, 0.0437, 0.0623, 0.0247, 0.0664, 0.0606], device='cuda:2'), in_proj_covar=tensor([0.0467, 0.0596, 0.0485, 0.0407, 0.0373, 0.0388, 0.0496, 0.0443], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:56:52,887 INFO [train.py:938] (2/8) Epoch 10, validation: loss=0.158, simple_loss=0.2614, pruned_loss=0.02732, over 944034.00 frames. 2023-04-29 03:56:52,887 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 03:57:03,893 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1889, 5.7196, 5.8804, 5.5949, 5.7263, 6.1950, 5.7289, 5.5141], device='cuda:2'), covar=tensor([0.0685, 0.1577, 0.1458, 0.1661, 0.1987, 0.0774, 0.1161, 0.1969], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0438, 0.0465, 0.0385, 0.0501, 0.0488, 0.0381, 0.0504], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 03:57:38,011 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1379, 5.6843, 5.8260, 5.6099, 5.7991, 6.1992, 5.7259, 5.5849], device='cuda:2'), covar=tensor([0.0761, 0.1704, 0.1705, 0.1860, 0.2456, 0.0915, 0.1297, 0.2157], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0447, 0.0472, 0.0393, 0.0513, 0.0497, 0.0387, 0.0516], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 03:58:02,502 INFO [train.py:904] (2/8) Epoch 10, batch 50, loss[loss=0.2244, simple_loss=0.2834, pruned_loss=0.08267, over 16922.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2944, pruned_loss=0.06698, over 740592.86 frames. ], batch size: 109, lr: 7.04e-03, grad_scale: 2.0 2023-04-29 03:58:39,941 INFO [optim.py:368] (2/8) 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 03:59:08,803 INFO [train.py:904] (2/8) Epoch 10, batch 100, loss[loss=0.1998, simple_loss=0.2871, pruned_loss=0.05626, over 17130.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.289, pruned_loss=0.0638, over 1315394.80 frames. ], batch size: 48, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 03:59:37,971 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-29 03:59:58,513 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8009, 4.7776, 5.3957, 5.3692, 5.3448, 4.9364, 4.9536, 4.6701], device='cuda:2'), covar=tensor([0.0313, 0.0407, 0.0437, 0.0445, 0.0432, 0.0367, 0.0856, 0.0398], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0303, 0.0307, 0.0293, 0.0342, 0.0325, 0.0417, 0.0261], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-29 04:00:16,887 INFO [train.py:904] (2/8) Epoch 10, batch 150, loss[loss=0.2433, simple_loss=0.2964, pruned_loss=0.09513, over 16873.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2853, pruned_loss=0.06155, over 1756638.68 frames. ], batch size: 109, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:00:22,898 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:00:56,392 INFO [optim.py:368] (2/8) 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 04:01:12,436 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:01:26,562 INFO [train.py:904] (2/8) Epoch 10, batch 200, loss[loss=0.2327, simple_loss=0.2975, pruned_loss=0.084, over 16730.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2847, pruned_loss=0.06146, over 2090386.64 frames. ], batch size: 124, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:01:28,066 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:02:34,675 INFO [train.py:904] (2/8) Epoch 10, batch 250, loss[loss=0.2173, simple_loss=0.2812, pruned_loss=0.07671, over 16707.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2812, pruned_loss=0.06072, over 2361030.33 frames. ], batch size: 83, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:02:36,456 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:03:09,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6610, 4.6468, 4.5566, 4.0699, 4.5653, 1.7016, 4.3125, 4.3508], device='cuda:2'), covar=tensor([0.0086, 0.0068, 0.0124, 0.0257, 0.0070, 0.2185, 0.0109, 0.0158], device='cuda:2'), in_proj_covar=tensor([0.0119, 0.0106, 0.0153, 0.0141, 0.0122, 0.0173, 0.0140, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:03:11,342 INFO [optim.py:368] (2/8) 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 04:03:28,031 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:03:42,033 INFO [train.py:904] (2/8) Epoch 10, batch 300, loss[loss=0.1974, simple_loss=0.2713, pruned_loss=0.06175, over 16868.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2773, pruned_loss=0.05802, over 2574774.76 frames. ], batch size: 102, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:04:30,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9112, 5.2505, 4.9771, 5.0038, 4.6798, 4.7102, 4.6646, 5.3605], device='cuda:2'), covar=tensor([0.1008, 0.0899, 0.1029, 0.0635, 0.0837, 0.0808, 0.1008, 0.0856], device='cuda:2'), in_proj_covar=tensor([0.0510, 0.0645, 0.0528, 0.0441, 0.0404, 0.0418, 0.0541, 0.0478], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:04:51,283 INFO [train.py:904] (2/8) Epoch 10, batch 350, loss[loss=0.2234, simple_loss=0.2791, pruned_loss=0.08387, over 16812.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2748, pruned_loss=0.05719, over 2751174.69 frames. ], batch size: 124, lr: 7.02e-03, grad_scale: 1.0 2023-04-29 04:05:25,725 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-29 04:05:28,619 INFO [optim.py:368] (2/8) 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 04:05:59,422 INFO [train.py:904] (2/8) Epoch 10, batch 400, loss[loss=0.1845, simple_loss=0.275, pruned_loss=0.04696, over 16657.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2721, pruned_loss=0.05569, over 2877814.26 frames. ], batch size: 57, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:06:07,121 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6104, 3.9774, 4.3379, 1.9417, 4.4181, 4.5802, 3.2213, 3.3997], device='cuda:2'), covar=tensor([0.0849, 0.0182, 0.0165, 0.1262, 0.0089, 0.0114, 0.0429, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0098, 0.0084, 0.0141, 0.0069, 0.0098, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 04:07:11,415 INFO [train.py:904] (2/8) Epoch 10, batch 450, loss[loss=0.1807, simple_loss=0.2609, pruned_loss=0.05021, over 16902.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2705, pruned_loss=0.05546, over 2975780.64 frames. ], batch size: 96, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:50,759 INFO [optim.py:368] (2/8) 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 04:08:10,240 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8245, 4.0108, 4.3446, 2.1395, 4.5299, 4.5561, 3.1985, 3.5457], device='cuda:2'), covar=tensor([0.0675, 0.0177, 0.0143, 0.1090, 0.0060, 0.0102, 0.0380, 0.0332], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0098, 0.0084, 0.0141, 0.0069, 0.0098, 0.0120, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 04:08:20,249 INFO [train.py:904] (2/8) Epoch 10, batch 500, loss[loss=0.1985, simple_loss=0.2873, pruned_loss=0.05486, over 17037.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2687, pruned_loss=0.05442, over 3043086.77 frames. ], batch size: 55, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:08:42,895 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 04:09:15,097 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:09:23,880 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:09:29,465 INFO [train.py:904] (2/8) Epoch 10, batch 550, loss[loss=0.2074, simple_loss=0.2733, pruned_loss=0.0707, over 16444.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2685, pruned_loss=0.05409, over 3112113.65 frames. ], batch size: 146, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:10:07,829 INFO [optim.py:368] (2/8) 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 04:10:22,710 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:37,404 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 04:10:38,342 INFO [train.py:904] (2/8) Epoch 10, batch 600, loss[loss=0.1791, simple_loss=0.2537, pruned_loss=0.05223, over 15528.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.268, pruned_loss=0.05451, over 3168840.06 frames. ], batch size: 190, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:10:38,887 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3423, 4.2327, 2.6143, 4.7999, 3.2152, 4.7464, 2.6488, 3.6221], device='cuda:2'), covar=tensor([0.0134, 0.0273, 0.1289, 0.0143, 0.0682, 0.0378, 0.1256, 0.0472], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0163, 0.0187, 0.0121, 0.0167, 0.0202, 0.0193, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:10:38,909 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:42,794 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:55,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0200, 4.5298, 3.4320, 2.4110, 3.0794, 2.5049, 4.8779, 3.9718], device='cuda:2'), covar=tensor([0.2459, 0.0607, 0.1331, 0.2181, 0.2400, 0.1785, 0.0294, 0.1077], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0253, 0.0279, 0.0271, 0.0267, 0.0216, 0.0264, 0.0288], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:11:18,659 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:11:30,053 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:51,294 INFO [train.py:904] (2/8) Epoch 10, batch 650, loss[loss=0.1971, simple_loss=0.2655, pruned_loss=0.06433, over 12435.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2675, pruned_loss=0.05398, over 3199814.84 frames. ], batch size: 247, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:12:12,107 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:12:22,151 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 04:12:31,800 INFO [optim.py:368] (2/8) 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 04:12:49,469 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:12:55,646 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1337, 5.1536, 4.9787, 4.5818, 4.4614, 5.0852, 5.0977, 4.6419], device='cuda:2'), covar=tensor([0.0649, 0.0317, 0.0300, 0.0302, 0.1328, 0.0357, 0.0286, 0.0683], device='cuda:2'), in_proj_covar=tensor([0.0244, 0.0298, 0.0284, 0.0262, 0.0309, 0.0298, 0.0193, 0.0334], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 04:13:03,073 INFO [train.py:904] (2/8) Epoch 10, batch 700, loss[loss=0.2178, simple_loss=0.2832, pruned_loss=0.07624, over 16913.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2671, pruned_loss=0.05388, over 3216139.46 frames. ], batch size: 116, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:13:53,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1927, 4.0440, 4.2151, 4.4187, 4.5072, 4.0940, 4.2402, 4.4766], device='cuda:2'), covar=tensor([0.1255, 0.0939, 0.1314, 0.0632, 0.0526, 0.1012, 0.1791, 0.0552], device='cuda:2'), in_proj_covar=tensor([0.0522, 0.0637, 0.0777, 0.0643, 0.0490, 0.0492, 0.0514, 0.0577], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:14:09,143 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8165, 1.7193, 2.2668, 2.5967, 2.6621, 2.5905, 1.6210, 2.7511], device='cuda:2'), covar=tensor([0.0100, 0.0319, 0.0200, 0.0170, 0.0164, 0.0180, 0.0367, 0.0097], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0167, 0.0152, 0.0155, 0.0163, 0.0119, 0.0168, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 04:14:12,201 INFO [train.py:904] (2/8) Epoch 10, batch 750, loss[loss=0.2262, simple_loss=0.2948, pruned_loss=0.07879, over 16727.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2689, pruned_loss=0.05507, over 3241121.38 frames. ], batch size: 124, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:14:52,014 INFO [optim.py:368] (2/8) 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 04:15:22,687 INFO [train.py:904] (2/8) Epoch 10, batch 800, loss[loss=0.1741, simple_loss=0.2625, pruned_loss=0.0428, over 17247.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.269, pruned_loss=0.05465, over 3260841.15 frames. ], batch size: 45, lr: 7.01e-03, grad_scale: 4.0 2023-04-29 04:16:11,790 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-29 04:16:27,562 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:16:32,818 INFO [train.py:904] (2/8) Epoch 10, batch 850, loss[loss=0.19, simple_loss=0.269, pruned_loss=0.0555, over 16470.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2678, pruned_loss=0.05419, over 3273898.11 frames. ], batch size: 68, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:17:10,136 INFO [optim.py:368] (2/8) 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 04:17:33,334 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:17:34,417 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:17:41,170 INFO [train.py:904] (2/8) Epoch 10, batch 900, loss[loss=0.1876, simple_loss=0.2803, pruned_loss=0.04742, over 17111.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2668, pruned_loss=0.05288, over 3289523.69 frames. ], batch size: 53, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:17:45,758 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5898, 2.5816, 2.2373, 2.4099, 2.9380, 2.7616, 3.4702, 3.1812], device='cuda:2'), covar=tensor([0.0073, 0.0290, 0.0350, 0.0347, 0.0180, 0.0256, 0.0181, 0.0179], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0197, 0.0192, 0.0192, 0.0194, 0.0197, 0.0199, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:17:49,467 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8051, 5.0259, 5.1151, 4.9501, 5.0632, 5.6189, 5.1660, 4.8217], device='cuda:2'), covar=tensor([0.1090, 0.1844, 0.1964, 0.1859, 0.2977, 0.1105, 0.1541, 0.2528], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0491, 0.0517, 0.0424, 0.0566, 0.0538, 0.0417, 0.0566], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 04:17:58,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8430, 4.6201, 4.8665, 5.0846, 5.2270, 4.6821, 5.2241, 5.1709], device='cuda:2'), covar=tensor([0.1314, 0.1000, 0.1414, 0.0576, 0.0465, 0.0734, 0.0460, 0.0472], device='cuda:2'), in_proj_covar=tensor([0.0526, 0.0637, 0.0784, 0.0650, 0.0491, 0.0493, 0.0516, 0.0580], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:18:43,177 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 04:18:50,837 INFO [train.py:904] (2/8) Epoch 10, batch 950, loss[loss=0.1878, simple_loss=0.2866, pruned_loss=0.04451, over 17000.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2675, pruned_loss=0.05368, over 3302027.16 frames. ], batch size: 50, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:19:04,203 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:19:27,841 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5194, 3.5831, 1.9958, 3.7344, 2.6708, 3.6969, 2.0526, 2.8312], device='cuda:2'), covar=tensor([0.0196, 0.0345, 0.1512, 0.0212, 0.0720, 0.0706, 0.1357, 0.0597], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0164, 0.0188, 0.0124, 0.0166, 0.0203, 0.0192, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:19:29,762 INFO [optim.py:368] (2/8) 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 04:19:39,591 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:19:58,791 INFO [train.py:904] (2/8) Epoch 10, batch 1000, loss[loss=0.1898, simple_loss=0.2702, pruned_loss=0.05475, over 16826.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2661, pruned_loss=0.05291, over 3316668.49 frames. ], batch size: 62, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:20:47,726 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 04:20:59,345 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1640, 5.1146, 4.8981, 4.3332, 4.9656, 1.7435, 4.6602, 4.8980], device='cuda:2'), covar=tensor([0.0065, 0.0058, 0.0140, 0.0371, 0.0077, 0.2458, 0.0119, 0.0174], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0111, 0.0162, 0.0151, 0.0131, 0.0179, 0.0149, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:21:03,442 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9493, 4.0588, 2.3331, 4.5366, 2.8050, 4.4710, 2.2694, 3.1119], device='cuda:2'), covar=tensor([0.0178, 0.0316, 0.1414, 0.0184, 0.0805, 0.0407, 0.1483, 0.0628], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0163, 0.0186, 0.0123, 0.0164, 0.0202, 0.0191, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:21:09,131 INFO [train.py:904] (2/8) Epoch 10, batch 1050, loss[loss=0.1626, simple_loss=0.2428, pruned_loss=0.04115, over 16935.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2649, pruned_loss=0.05207, over 3321705.80 frames. ], batch size: 41, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:21:48,359 INFO [optim.py:368] (2/8) 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 04:22:18,523 INFO [train.py:904] (2/8) Epoch 10, batch 1100, loss[loss=0.1589, simple_loss=0.2393, pruned_loss=0.03926, over 15782.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2646, pruned_loss=0.05181, over 3331472.24 frames. ], batch size: 35, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:23:28,210 INFO [train.py:904] (2/8) Epoch 10, batch 1150, loss[loss=0.1841, simple_loss=0.2628, pruned_loss=0.05265, over 16592.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2634, pruned_loss=0.05132, over 3326533.12 frames. ], batch size: 68, lr: 6.99e-03, grad_scale: 4.0 2023-04-29 04:24:08,393 INFO [optim.py:368] (2/8) 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 04:24:32,773 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:24:38,818 INFO [train.py:904] (2/8) Epoch 10, batch 1200, loss[loss=0.1919, simple_loss=0.2654, pruned_loss=0.05922, over 12302.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2624, pruned_loss=0.05106, over 3307993.89 frames. ], batch size: 246, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:25:06,498 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2193, 3.3790, 3.6105, 3.5681, 3.5548, 3.3658, 3.3694, 3.4393], device='cuda:2'), covar=tensor([0.0403, 0.0669, 0.0445, 0.0463, 0.0501, 0.0451, 0.0752, 0.0441], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0339, 0.0338, 0.0326, 0.0380, 0.0356, 0.0463, 0.0286], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 04:25:11,112 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4101, 3.4061, 3.7747, 2.5651, 3.4153, 3.7098, 3.6606, 2.1919], device='cuda:2'), covar=tensor([0.0345, 0.0137, 0.0034, 0.0258, 0.0079, 0.0073, 0.0050, 0.0333], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0070, 0.0068, 0.0125, 0.0077, 0.0084, 0.0075, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 04:25:39,051 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:25:47,783 INFO [train.py:904] (2/8) Epoch 10, batch 1250, loss[loss=0.1997, simple_loss=0.2671, pruned_loss=0.0661, over 16894.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2635, pruned_loss=0.05167, over 3308657.12 frames. ], batch size: 96, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:26:01,768 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:20,439 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:27,271 INFO [optim.py:368] (2/8) 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 04:26:37,469 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:26:58,798 INFO [train.py:904] (2/8) Epoch 10, batch 1300, loss[loss=0.1817, simple_loss=0.2729, pruned_loss=0.04521, over 17126.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2624, pruned_loss=0.05107, over 3317881.85 frames. ], batch size: 47, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:27:07,401 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:27:44,065 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:27:45,292 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:07,573 INFO [train.py:904] (2/8) Epoch 10, batch 1350, loss[loss=0.1983, simple_loss=0.2729, pruned_loss=0.06189, over 16647.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2626, pruned_loss=0.05099, over 3314611.84 frames. ], batch size: 76, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:28:34,069 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:47,042 INFO [optim.py:368] (2/8) 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 04:29:09,509 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 04:29:18,974 INFO [train.py:904] (2/8) Epoch 10, batch 1400, loss[loss=0.1677, simple_loss=0.2648, pruned_loss=0.03523, over 17111.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2631, pruned_loss=0.05138, over 3305526.53 frames. ], batch size: 48, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:29:25,145 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 04:29:37,187 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7842, 3.4447, 3.0444, 5.1825, 4.4517, 4.7780, 1.5838, 3.5324], device='cuda:2'), covar=tensor([0.1343, 0.0561, 0.0998, 0.0127, 0.0249, 0.0309, 0.1514, 0.0651], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0155, 0.0178, 0.0136, 0.0196, 0.0212, 0.0178, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:30:00,632 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:30:14,786 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5753, 3.6128, 2.0516, 3.8782, 2.6509, 3.7572, 2.0497, 2.7642], device='cuda:2'), covar=tensor([0.0185, 0.0356, 0.1311, 0.0164, 0.0709, 0.0576, 0.1364, 0.0613], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0167, 0.0189, 0.0125, 0.0167, 0.0206, 0.0195, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:30:25,870 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-29 04:30:28,851 INFO [train.py:904] (2/8) Epoch 10, batch 1450, loss[loss=0.2015, simple_loss=0.2651, pruned_loss=0.06896, over 16759.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2624, pruned_loss=0.05121, over 3303971.34 frames. ], batch size: 102, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:31:02,368 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9055, 4.6487, 4.8560, 5.1540, 5.3357, 4.5988, 5.3168, 5.2439], device='cuda:2'), covar=tensor([0.1383, 0.1103, 0.1765, 0.0668, 0.0482, 0.0825, 0.0450, 0.0535], device='cuda:2'), in_proj_covar=tensor([0.0538, 0.0660, 0.0809, 0.0674, 0.0503, 0.0507, 0.0528, 0.0597], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:31:07,981 INFO [optim.py:368] (2/8) 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 04:31:09,643 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:31:38,224 INFO [train.py:904] (2/8) Epoch 10, batch 1500, loss[loss=0.2201, simple_loss=0.2956, pruned_loss=0.07226, over 16571.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2632, pruned_loss=0.05146, over 3310852.46 frames. ], batch size: 68, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:32:34,075 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:45,600 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:48,539 INFO [train.py:904] (2/8) Epoch 10, batch 1550, loss[loss=0.1671, simple_loss=0.2557, pruned_loss=0.03923, over 17240.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2655, pruned_loss=0.0531, over 3307450.20 frames. ], batch size: 45, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:33:26,143 INFO [optim.py:368] (2/8) 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 04:33:27,724 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:33:56,442 INFO [train.py:904] (2/8) Epoch 10, batch 1600, loss[loss=0.1803, simple_loss=0.2758, pruned_loss=0.04239, over 17151.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.267, pruned_loss=0.0537, over 3319897.63 frames. ], batch size: 49, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:34:07,748 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:29,001 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:37,284 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:52,586 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:06,231 INFO [train.py:904] (2/8) Epoch 10, batch 1650, loss[loss=0.1664, simple_loss=0.2507, pruned_loss=0.04105, over 17217.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2687, pruned_loss=0.05404, over 3320945.85 frames. ], batch size: 45, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:35:18,291 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:35,076 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7937, 2.1080, 2.1760, 4.5414, 1.9961, 2.7198, 2.3140, 2.3628], device='cuda:2'), covar=tensor([0.0848, 0.3549, 0.2305, 0.0325, 0.3954, 0.2280, 0.2993, 0.3318], device='cuda:2'), in_proj_covar=tensor([0.0358, 0.0383, 0.0325, 0.0323, 0.0409, 0.0435, 0.0345, 0.0452], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:35:45,200 INFO [optim.py:368] (2/8) 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 04:35:53,074 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:53,168 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:08,059 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 04:36:15,632 INFO [train.py:904] (2/8) Epoch 10, batch 1700, loss[loss=0.1683, simple_loss=0.2576, pruned_loss=0.03946, over 16893.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2693, pruned_loss=0.05415, over 3318429.27 frames. ], batch size: 96, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:36:42,602 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:46,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 04:36:48,265 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:37:17,837 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:37:23,690 INFO [train.py:904] (2/8) Epoch 10, batch 1750, loss[loss=0.2262, simple_loss=0.323, pruned_loss=0.06465, over 17040.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2703, pruned_loss=0.05413, over 3323414.12 frames. ], batch size: 53, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:37:28,763 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0540, 4.2420, 2.6020, 4.7312, 3.2107, 4.7439, 2.5742, 3.5186], device='cuda:2'), covar=tensor([0.0236, 0.0296, 0.1291, 0.0173, 0.0703, 0.0379, 0.1345, 0.0584], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0167, 0.0187, 0.0126, 0.0167, 0.0207, 0.0194, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:37:34,233 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-29 04:37:42,813 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:37:51,346 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8637, 3.3259, 2.9710, 5.1643, 4.2490, 4.6807, 1.7468, 3.2074], device='cuda:2'), covar=tensor([0.1291, 0.0627, 0.1048, 0.0136, 0.0251, 0.0314, 0.1441, 0.0779], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0154, 0.0177, 0.0136, 0.0195, 0.0211, 0.0176, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 04:38:01,415 INFO [optim.py:368] (2/8) 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 04:38:23,180 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7527, 3.7296, 2.7597, 2.1615, 2.5172, 2.2184, 3.6972, 3.4302], device='cuda:2'), covar=tensor([0.2249, 0.0591, 0.1523, 0.2281, 0.2180, 0.1747, 0.0525, 0.1153], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0256, 0.0280, 0.0272, 0.0274, 0.0218, 0.0264, 0.0294], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:38:32,562 INFO [train.py:904] (2/8) Epoch 10, batch 1800, loss[loss=0.2209, simple_loss=0.3025, pruned_loss=0.06959, over 16408.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2711, pruned_loss=0.05411, over 3325766.69 frames. ], batch size: 75, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:04,795 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1998, 4.1910, 4.6590, 4.6274, 4.6333, 4.3123, 4.3430, 4.1070], device='cuda:2'), covar=tensor([0.0335, 0.0537, 0.0351, 0.0414, 0.0448, 0.0334, 0.0729, 0.0560], device='cuda:2'), in_proj_covar=tensor([0.0333, 0.0337, 0.0339, 0.0325, 0.0382, 0.0356, 0.0465, 0.0285], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 04:39:06,679 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:09,710 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0420, 4.9893, 5.5900, 5.5802, 5.5602, 5.1632, 5.1137, 4.8026], device='cuda:2'), covar=tensor([0.0298, 0.0461, 0.0338, 0.0368, 0.0454, 0.0294, 0.0858, 0.0420], device='cuda:2'), in_proj_covar=tensor([0.0333, 0.0337, 0.0339, 0.0325, 0.0382, 0.0356, 0.0465, 0.0285], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 04:39:20,429 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:42,537 INFO [train.py:904] (2/8) Epoch 10, batch 1850, loss[loss=0.2297, simple_loss=0.3107, pruned_loss=0.07433, over 11842.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.272, pruned_loss=0.05433, over 3316590.57 frames. ], batch size: 246, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:56,509 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 04:40:06,246 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 04:40:21,084 INFO [optim.py:368] (2/8) 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 04:40:46,257 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 04:40:48,296 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-29 04:40:52,060 INFO [train.py:904] (2/8) Epoch 10, batch 1900, loss[loss=0.1937, simple_loss=0.2976, pruned_loss=0.04492, over 17048.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2713, pruned_loss=0.0534, over 3311499.99 frames. ], batch size: 53, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:40:55,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9839, 4.2202, 2.5330, 4.6388, 3.2550, 4.6539, 2.5085, 3.4358], device='cuda:2'), covar=tensor([0.0212, 0.0254, 0.1227, 0.0164, 0.0685, 0.0353, 0.1218, 0.0499], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0165, 0.0184, 0.0125, 0.0164, 0.0204, 0.0191, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:40:56,692 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:22,768 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0041, 3.3009, 2.8556, 5.0823, 4.2219, 4.5761, 1.7372, 3.3459], device='cuda:2'), covar=tensor([0.1198, 0.0556, 0.1070, 0.0146, 0.0235, 0.0361, 0.1383, 0.0692], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0156, 0.0179, 0.0137, 0.0196, 0.0212, 0.0177, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 04:41:33,305 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:41,655 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:42:02,732 INFO [train.py:904] (2/8) Epoch 10, batch 1950, loss[loss=0.1367, simple_loss=0.2206, pruned_loss=0.02641, over 16836.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2709, pruned_loss=0.05259, over 3322519.29 frames. ], batch size: 39, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:42:40,592 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:42:41,417 INFO [optim.py:368] (2/8) 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 04:42:42,924 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:00,127 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7568, 2.5530, 2.1353, 2.6054, 2.8862, 2.8489, 3.5275, 3.2582], device='cuda:2'), covar=tensor([0.0060, 0.0324, 0.0378, 0.0302, 0.0204, 0.0264, 0.0167, 0.0178], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0202, 0.0198, 0.0196, 0.0198, 0.0200, 0.0209, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:43:12,557 INFO [train.py:904] (2/8) Epoch 10, batch 2000, loss[loss=0.2189, simple_loss=0.2942, pruned_loss=0.07183, over 15371.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2692, pruned_loss=0.05216, over 3323420.51 frames. ], batch size: 190, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:43:31,503 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:44,541 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:44:08,367 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:44:21,491 INFO [train.py:904] (2/8) Epoch 10, batch 2050, loss[loss=0.181, simple_loss=0.2809, pruned_loss=0.0405, over 17163.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2695, pruned_loss=0.05231, over 3310638.86 frames. ], batch size: 48, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:44:51,424 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:44:59,167 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6789, 3.9153, 2.1546, 4.3797, 2.6405, 4.3860, 2.2332, 3.0997], device='cuda:2'), covar=tensor([0.0246, 0.0294, 0.1508, 0.0195, 0.0915, 0.0326, 0.1421, 0.0668], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0167, 0.0187, 0.0128, 0.0167, 0.0207, 0.0194, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:45:00,990 INFO [optim.py:368] (2/8) 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 04:45:29,945 INFO [train.py:904] (2/8) Epoch 10, batch 2100, loss[loss=0.1797, simple_loss=0.2753, pruned_loss=0.04207, over 17033.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2702, pruned_loss=0.05263, over 3315960.64 frames. ], batch size: 50, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:45:33,606 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-29 04:45:56,870 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:18,626 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:18,722 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5818, 3.7607, 4.0417, 1.9886, 4.3384, 4.3911, 3.0366, 3.2792], device='cuda:2'), covar=tensor([0.0755, 0.0184, 0.0209, 0.1094, 0.0060, 0.0126, 0.0403, 0.0356], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0098, 0.0087, 0.0139, 0.0070, 0.0101, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 04:46:28,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6692, 3.7864, 2.2069, 4.0260, 2.8264, 3.9797, 2.1211, 2.9425], device='cuda:2'), covar=tensor([0.0186, 0.0263, 0.1309, 0.0151, 0.0625, 0.0526, 0.1395, 0.0544], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0165, 0.0185, 0.0127, 0.0166, 0.0206, 0.0192, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 04:46:40,125 INFO [train.py:904] (2/8) Epoch 10, batch 2150, loss[loss=0.1763, simple_loss=0.2747, pruned_loss=0.03896, over 17147.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2714, pruned_loss=0.05333, over 3310398.88 frames. ], batch size: 47, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:47:18,311 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:47:46,702 INFO [train.py:904] (2/8) Epoch 10, batch 2200, loss[loss=0.199, simple_loss=0.2949, pruned_loss=0.05158, over 17146.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2726, pruned_loss=0.05417, over 3314671.44 frames. ], batch size: 49, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:47:51,953 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:35,495 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:48,892 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2359, 1.5092, 1.9258, 2.1633, 2.3519, 2.2273, 1.5747, 2.2536], device='cuda:2'), covar=tensor([0.0144, 0.0312, 0.0180, 0.0190, 0.0147, 0.0163, 0.0304, 0.0084], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0170, 0.0156, 0.0158, 0.0166, 0.0122, 0.0170, 0.0111], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 04:48:54,025 INFO [train.py:904] (2/8) Epoch 10, batch 2250, loss[loss=0.1611, simple_loss=0.2415, pruned_loss=0.04034, over 16799.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2736, pruned_loss=0.05485, over 3311331.68 frames. ], batch size: 39, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:48:55,410 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:33,830 INFO [optim.py:368] (2/8) 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 04:49:34,796 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93631.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:39,246 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:43,179 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 04:50:04,001 INFO [train.py:904] (2/8) Epoch 10, batch 2300, loss[loss=0.2172, simple_loss=0.2914, pruned_loss=0.07154, over 16780.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2742, pruned_loss=0.05511, over 3302981.88 frames. ], batch size: 124, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:50:17,527 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9095, 5.2508, 5.3425, 5.1489, 5.1166, 5.7802, 5.2626, 5.0133], device='cuda:2'), covar=tensor([0.1046, 0.1933, 0.1845, 0.1769, 0.2724, 0.1009, 0.1451, 0.2334], device='cuda:2'), in_proj_covar=tensor([0.0346, 0.0497, 0.0519, 0.0427, 0.0574, 0.0549, 0.0419, 0.0576], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 04:50:22,468 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:39,710 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:57,915 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:09,771 INFO [train.py:904] (2/8) Epoch 10, batch 2350, loss[loss=0.2073, simple_loss=0.2791, pruned_loss=0.06773, over 16753.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2742, pruned_loss=0.05501, over 3309199.61 frames. ], batch size: 124, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:51:27,699 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:49,909 INFO [optim.py:368] (2/8) 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 04:52:03,150 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:17,488 INFO [train.py:904] (2/8) Epoch 10, batch 2400, loss[loss=0.204, simple_loss=0.2785, pruned_loss=0.06479, over 16379.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.275, pruned_loss=0.05515, over 3313247.66 frames. ], batch size: 145, lr: 6.95e-03, grad_scale: 8.0 2023-04-29 04:52:41,473 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:42,518 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:53:26,815 INFO [train.py:904] (2/8) Epoch 10, batch 2450, loss[loss=0.1826, simple_loss=0.2582, pruned_loss=0.05357, over 16904.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2754, pruned_loss=0.05498, over 3309690.62 frames. ], batch size: 90, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:53:49,967 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:05,072 INFO [optim.py:368] (2/8) 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 04:54:05,494 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:26,482 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2518, 5.2622, 5.0012, 4.4793, 5.1014, 1.9018, 4.8288, 5.0722], device='cuda:2'), covar=tensor([0.0061, 0.0049, 0.0132, 0.0338, 0.0072, 0.2284, 0.0108, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0117, 0.0168, 0.0158, 0.0137, 0.0180, 0.0155, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 04:54:33,766 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0137, 5.0190, 5.5173, 5.5056, 5.5369, 5.1244, 5.1543, 4.8615], device='cuda:2'), covar=tensor([0.0282, 0.0413, 0.0407, 0.0458, 0.0378, 0.0315, 0.0834, 0.0386], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0341, 0.0344, 0.0327, 0.0383, 0.0359, 0.0466, 0.0289], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 04:54:34,538 INFO [train.py:904] (2/8) Epoch 10, batch 2500, loss[loss=0.1512, simple_loss=0.2381, pruned_loss=0.03211, over 16948.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2743, pruned_loss=0.0543, over 3318047.77 frames. ], batch size: 41, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:55:43,656 INFO [train.py:904] (2/8) Epoch 10, batch 2550, loss[loss=0.1946, simple_loss=0.2876, pruned_loss=0.0508, over 17092.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2746, pruned_loss=0.05447, over 3319652.77 frames. ], batch size: 47, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:55:47,618 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 04:55:51,409 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 04:56:23,962 INFO [optim.py:368] (2/8) 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 04:56:52,858 INFO [train.py:904] (2/8) Epoch 10, batch 2600, loss[loss=0.207, simple_loss=0.2869, pruned_loss=0.06354, over 16833.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2743, pruned_loss=0.05395, over 3325510.06 frames. ], batch size: 96, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:57:32,250 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:58:03,850 INFO [train.py:904] (2/8) Epoch 10, batch 2650, loss[loss=0.1747, simple_loss=0.2708, pruned_loss=0.03935, over 17228.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.274, pruned_loss=0.05357, over 3330831.71 frames. ], batch size: 45, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:58:43,800 INFO [optim.py:368] (2/8) 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 04:58:50,821 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-04-29 04:59:00,334 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:59:13,552 INFO [train.py:904] (2/8) Epoch 10, batch 2700, loss[loss=0.1865, simple_loss=0.277, pruned_loss=0.04804, over 16658.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2744, pruned_loss=0.05308, over 3327765.75 frames. ], batch size: 62, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:59:53,891 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 05:00:23,305 INFO [train.py:904] (2/8) Epoch 10, batch 2750, loss[loss=0.1891, simple_loss=0.2913, pruned_loss=0.04345, over 17030.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2748, pruned_loss=0.05298, over 3325922.48 frames. ], batch size: 53, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:00:55,392 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:01:01,089 INFO [optim.py:368] (2/8) 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 05:01:16,881 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3445, 2.0297, 2.1973, 3.9947, 2.0158, 2.5181, 2.0741, 2.2289], device='cuda:2'), covar=tensor([0.0957, 0.3299, 0.2102, 0.0382, 0.3410, 0.2070, 0.3199, 0.2716], device='cuda:2'), in_proj_covar=tensor([0.0362, 0.0387, 0.0325, 0.0324, 0.0409, 0.0440, 0.0347, 0.0456], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:01:29,685 INFO [train.py:904] (2/8) Epoch 10, batch 2800, loss[loss=0.1717, simple_loss=0.2582, pruned_loss=0.04259, over 16816.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.274, pruned_loss=0.05232, over 3328742.80 frames. ], batch size: 42, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:01:45,599 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8706, 4.1441, 2.4429, 4.7683, 2.9384, 4.6555, 2.3673, 3.2272], device='cuda:2'), covar=tensor([0.0234, 0.0287, 0.1366, 0.0182, 0.0799, 0.0389, 0.1472, 0.0677], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0167, 0.0187, 0.0131, 0.0167, 0.0211, 0.0196, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 05:02:30,989 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7411, 1.2899, 1.6333, 1.6401, 1.7957, 1.9510, 1.5298, 1.7838], device='cuda:2'), covar=tensor([0.0154, 0.0249, 0.0131, 0.0168, 0.0155, 0.0113, 0.0230, 0.0069], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0170, 0.0154, 0.0159, 0.0166, 0.0122, 0.0170, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 05:02:39,399 INFO [train.py:904] (2/8) Epoch 10, batch 2850, loss[loss=0.1981, simple_loss=0.2748, pruned_loss=0.06074, over 16854.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2738, pruned_loss=0.05241, over 3319280.53 frames. ], batch size: 90, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:03:09,000 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:03:20,115 INFO [optim.py:368] (2/8) 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 05:03:49,048 INFO [train.py:904] (2/8) Epoch 10, batch 2900, loss[loss=0.1666, simple_loss=0.2545, pruned_loss=0.03937, over 16997.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2727, pruned_loss=0.0533, over 3319335.47 frames. ], batch size: 41, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:04:21,251 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0857, 5.0776, 4.8586, 4.3777, 4.9771, 1.9107, 4.7371, 4.9176], device='cuda:2'), covar=tensor([0.0088, 0.0061, 0.0161, 0.0359, 0.0078, 0.2254, 0.0121, 0.0169], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0119, 0.0168, 0.0161, 0.0137, 0.0179, 0.0155, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:04:33,842 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:04:58,066 INFO [train.py:904] (2/8) Epoch 10, batch 2950, loss[loss=0.211, simple_loss=0.2809, pruned_loss=0.07059, over 16786.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2729, pruned_loss=0.05451, over 3312220.13 frames. ], batch size: 124, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:05:39,543 INFO [optim.py:368] (2/8) 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 05:05:49,195 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:03,098 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:08,006 INFO [train.py:904] (2/8) Epoch 10, batch 3000, loss[loss=0.1847, simple_loss=0.2619, pruned_loss=0.05375, over 16781.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.273, pruned_loss=0.05505, over 3317960.55 frames. ], batch size: 83, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:06:08,006 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 05:06:17,134 INFO [train.py:938] (2/8) Epoch 10, validation: loss=0.1426, simple_loss=0.2488, pruned_loss=0.01818, over 944034.00 frames. 2023-04-29 05:06:17,134 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 05:07:26,696 INFO [train.py:904] (2/8) Epoch 10, batch 3050, loss[loss=0.1798, simple_loss=0.2572, pruned_loss=0.05117, over 15948.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2733, pruned_loss=0.05537, over 3319508.20 frames. ], batch size: 35, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:07:36,881 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:07:50,834 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 05:07:57,835 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:08:05,349 INFO [optim.py:368] (2/8) 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 05:08:33,248 INFO [train.py:904] (2/8) Epoch 10, batch 3100, loss[loss=0.1875, simple_loss=0.2795, pruned_loss=0.04778, over 17248.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2729, pruned_loss=0.05563, over 3308238.32 frames. ], batch size: 45, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:08:45,861 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-29 05:09:04,482 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:09:43,505 INFO [train.py:904] (2/8) Epoch 10, batch 3150, loss[loss=0.2057, simple_loss=0.2755, pruned_loss=0.06797, over 16393.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2718, pruned_loss=0.05476, over 3316271.22 frames. ], batch size: 146, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:10:10,839 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3422, 3.9865, 4.4673, 2.0159, 4.7763, 4.7907, 3.4327, 3.7074], device='cuda:2'), covar=tensor([0.0489, 0.0211, 0.0189, 0.1039, 0.0044, 0.0086, 0.0285, 0.0320], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0100, 0.0089, 0.0138, 0.0070, 0.0103, 0.0122, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 05:10:23,683 INFO [optim.py:368] (2/8) 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 05:10:52,169 INFO [train.py:904] (2/8) Epoch 10, batch 3200, loss[loss=0.1655, simple_loss=0.2606, pruned_loss=0.03519, over 17096.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2706, pruned_loss=0.05324, over 3322257.34 frames. ], batch size: 47, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:11:13,556 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 05:11:32,213 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:12:04,563 INFO [train.py:904] (2/8) Epoch 10, batch 3250, loss[loss=0.1865, simple_loss=0.2659, pruned_loss=0.05352, over 16719.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2707, pruned_loss=0.05305, over 3325288.54 frames. ], batch size: 83, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:12:32,408 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-04-29 05:12:44,901 INFO [optim.py:368] (2/8) 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 05:12:53,027 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:13:14,618 INFO [train.py:904] (2/8) Epoch 10, batch 3300, loss[loss=0.1951, simple_loss=0.2781, pruned_loss=0.05606, over 16696.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2716, pruned_loss=0.05375, over 3322870.21 frames. ], batch size: 89, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:13:49,128 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1511, 3.2056, 3.5071, 2.2764, 3.1556, 3.5702, 3.3874, 1.9870], device='cuda:2'), covar=tensor([0.0394, 0.0104, 0.0041, 0.0312, 0.0080, 0.0071, 0.0056, 0.0350], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0069, 0.0069, 0.0123, 0.0078, 0.0084, 0.0075, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 05:13:52,647 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4854, 3.5827, 2.0684, 3.8362, 2.6139, 3.8145, 2.0079, 2.8010], device='cuda:2'), covar=tensor([0.0206, 0.0356, 0.1486, 0.0197, 0.0765, 0.0613, 0.1491, 0.0641], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0167, 0.0186, 0.0131, 0.0167, 0.0209, 0.0194, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 05:14:01,715 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:14:18,234 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-29 05:14:24,568 INFO [train.py:904] (2/8) Epoch 10, batch 3350, loss[loss=0.1669, simple_loss=0.2559, pruned_loss=0.03894, over 17120.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2724, pruned_loss=0.0537, over 3321905.56 frames. ], batch size: 47, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:28,545 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:15:04,278 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0840, 4.8141, 5.0887, 5.3204, 5.5248, 4.7698, 5.4598, 5.4331], device='cuda:2'), covar=tensor([0.1389, 0.0942, 0.1494, 0.0581, 0.0402, 0.0707, 0.0435, 0.0449], device='cuda:2'), in_proj_covar=tensor([0.0543, 0.0664, 0.0821, 0.0681, 0.0506, 0.0524, 0.0535, 0.0600], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:15:05,006 INFO [optim.py:368] (2/8) 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 05:15:33,171 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6449, 6.0692, 5.7169, 5.7963, 5.2736, 5.2934, 5.4854, 6.1442], device='cuda:2'), covar=tensor([0.1054, 0.0672, 0.0916, 0.0657, 0.0820, 0.0573, 0.0946, 0.0779], device='cuda:2'), in_proj_covar=tensor([0.0542, 0.0689, 0.0566, 0.0471, 0.0432, 0.0438, 0.0569, 0.0524], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:15:35,790 INFO [train.py:904] (2/8) Epoch 10, batch 3400, loss[loss=0.1535, simple_loss=0.2415, pruned_loss=0.0327, over 17243.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2715, pruned_loss=0.05314, over 3323658.29 frames. ], batch size: 44, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:16:16,631 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7686, 2.1791, 2.2621, 4.5464, 2.1126, 2.6659, 2.3107, 2.4427], device='cuda:2'), covar=tensor([0.0813, 0.3215, 0.2224, 0.0343, 0.3790, 0.2321, 0.2939, 0.3215], device='cuda:2'), in_proj_covar=tensor([0.0365, 0.0388, 0.0326, 0.0327, 0.0410, 0.0443, 0.0349, 0.0458], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:16:19,444 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1418, 5.1023, 4.9784, 4.5941, 4.5604, 5.0318, 4.9413, 4.6022], device='cuda:2'), covar=tensor([0.0579, 0.0443, 0.0273, 0.0290, 0.1140, 0.0405, 0.0278, 0.0738], device='cuda:2'), in_proj_covar=tensor([0.0257, 0.0318, 0.0302, 0.0280, 0.0330, 0.0317, 0.0208, 0.0354], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 05:16:44,908 INFO [train.py:904] (2/8) Epoch 10, batch 3450, loss[loss=0.196, simple_loss=0.2911, pruned_loss=0.05047, over 17130.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2705, pruned_loss=0.05286, over 3310312.65 frames. ], batch size: 49, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:17:26,304 INFO [optim.py:368] (2/8) 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 05:17:29,194 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3597, 3.6160, 3.3663, 2.0446, 2.9538, 2.5516, 3.7977, 3.7413], device='cuda:2'), covar=tensor([0.0188, 0.0633, 0.0570, 0.1595, 0.0694, 0.0839, 0.0416, 0.0629], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0146, 0.0156, 0.0143, 0.0135, 0.0124, 0.0137, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 05:17:56,543 INFO [train.py:904] (2/8) Epoch 10, batch 3500, loss[loss=0.1875, simple_loss=0.2648, pruned_loss=0.05511, over 16789.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.27, pruned_loss=0.05277, over 3303695.71 frames. ], batch size: 83, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:18:14,732 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7446, 1.7117, 1.5295, 1.5417, 1.8763, 1.6442, 1.7588, 1.9676], device='cuda:2'), covar=tensor([0.0104, 0.0177, 0.0272, 0.0255, 0.0146, 0.0194, 0.0126, 0.0133], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0202, 0.0197, 0.0197, 0.0200, 0.0203, 0.0212, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:18:35,809 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:19:06,951 INFO [train.py:904] (2/8) Epoch 10, batch 3550, loss[loss=0.1926, simple_loss=0.2672, pruned_loss=0.05902, over 15536.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2695, pruned_loss=0.0523, over 3295502.86 frames. ], batch size: 190, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:19:13,013 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 2023-04-29 05:19:13,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2788, 3.3195, 3.5008, 1.8123, 3.6797, 3.6194, 2.9167, 2.7243], device='cuda:2'), covar=tensor([0.0713, 0.0176, 0.0165, 0.1032, 0.0064, 0.0148, 0.0368, 0.0397], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0100, 0.0088, 0.0138, 0.0069, 0.0102, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 05:19:41,448 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:19:47,556 INFO [optim.py:368] (2/8) 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 05:20:01,882 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9203, 4.9378, 5.4975, 5.4611, 5.4311, 5.0734, 5.0375, 4.7608], device='cuda:2'), covar=tensor([0.0320, 0.0474, 0.0329, 0.0366, 0.0519, 0.0300, 0.0886, 0.0432], device='cuda:2'), in_proj_covar=tensor([0.0346, 0.0354, 0.0356, 0.0338, 0.0400, 0.0372, 0.0481, 0.0296], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 05:20:17,532 INFO [train.py:904] (2/8) Epoch 10, batch 3600, loss[loss=0.1986, simple_loss=0.2812, pruned_loss=0.05798, over 17111.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2676, pruned_loss=0.05206, over 3295108.22 frames. ], batch size: 53, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:01,060 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-29 05:21:28,700 INFO [train.py:904] (2/8) Epoch 10, batch 3650, loss[loss=0.1918, simple_loss=0.2637, pruned_loss=0.05989, over 16326.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2661, pruned_loss=0.05273, over 3296026.88 frames. ], batch size: 165, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:32,931 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:21:57,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4477, 4.1058, 4.0742, 2.1604, 3.1923, 2.6674, 3.6843, 4.0956], device='cuda:2'), covar=tensor([0.0301, 0.0581, 0.0456, 0.1683, 0.0761, 0.0857, 0.0692, 0.0762], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0147, 0.0157, 0.0144, 0.0136, 0.0125, 0.0138, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 05:21:58,544 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9656, 4.0328, 3.1937, 2.4160, 3.0106, 2.5400, 4.2812, 3.8321], device='cuda:2'), covar=tensor([0.2149, 0.0575, 0.1365, 0.1887, 0.2031, 0.1559, 0.0401, 0.0845], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0258, 0.0279, 0.0274, 0.0283, 0.0218, 0.0267, 0.0299], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:22:10,213 INFO [optim.py:368] (2/8) 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 05:22:36,822 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7997, 1.3647, 1.6741, 1.6233, 1.7384, 1.8515, 1.5069, 1.7155], device='cuda:2'), covar=tensor([0.0153, 0.0259, 0.0140, 0.0187, 0.0163, 0.0134, 0.0244, 0.0088], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0172, 0.0158, 0.0162, 0.0169, 0.0124, 0.0172, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 05:22:43,025 INFO [train.py:904] (2/8) Epoch 10, batch 3700, loss[loss=0.1719, simple_loss=0.2403, pruned_loss=0.05171, over 16898.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2657, pruned_loss=0.05436, over 3267665.85 frames. ], batch size: 96, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:22:43,300 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:23:56,092 INFO [train.py:904] (2/8) Epoch 10, batch 3750, loss[loss=0.2207, simple_loss=0.2893, pruned_loss=0.07608, over 11346.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2669, pruned_loss=0.05608, over 3249021.31 frames. ], batch size: 248, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:24:00,757 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:24:07,120 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6948, 4.6200, 4.7254, 4.9054, 5.0844, 4.4887, 4.9313, 4.9860], device='cuda:2'), covar=tensor([0.1616, 0.1257, 0.1515, 0.0898, 0.0551, 0.0973, 0.1186, 0.1005], device='cuda:2'), in_proj_covar=tensor([0.0541, 0.0664, 0.0815, 0.0679, 0.0507, 0.0525, 0.0534, 0.0603], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:24:38,276 INFO [optim.py:368] (2/8) 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 05:25:07,897 INFO [train.py:904] (2/8) Epoch 10, batch 3800, loss[loss=0.1978, simple_loss=0.271, pruned_loss=0.06229, over 16298.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2673, pruned_loss=0.0571, over 3242145.51 frames. ], batch size: 68, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:25:28,697 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:25:51,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1379, 3.3997, 3.5127, 2.3200, 3.1744, 3.5161, 3.3821, 2.1141], device='cuda:2'), covar=tensor([0.0382, 0.0069, 0.0031, 0.0284, 0.0066, 0.0069, 0.0049, 0.0296], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0069, 0.0068, 0.0122, 0.0077, 0.0085, 0.0075, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 05:26:02,151 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2810, 3.4370, 3.3149, 2.0772, 2.9198, 2.3778, 3.6379, 3.7239], device='cuda:2'), covar=tensor([0.0202, 0.0671, 0.0589, 0.1584, 0.0761, 0.0913, 0.0459, 0.0675], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0145, 0.0155, 0.0142, 0.0135, 0.0124, 0.0136, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 05:26:20,911 INFO [train.py:904] (2/8) Epoch 10, batch 3850, loss[loss=0.2019, simple_loss=0.2813, pruned_loss=0.06123, over 16573.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2674, pruned_loss=0.05735, over 3252727.37 frames. ], batch size: 62, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:26:59,694 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-29 05:27:00,982 INFO [optim.py:368] (2/8) 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 05:27:31,949 INFO [train.py:904] (2/8) Epoch 10, batch 3900, loss[loss=0.1775, simple_loss=0.256, pruned_loss=0.04947, over 16589.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2671, pruned_loss=0.05784, over 3255802.71 frames. ], batch size: 62, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:27:43,981 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:27:56,120 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:28:45,286 INFO [train.py:904] (2/8) Epoch 10, batch 3950, loss[loss=0.1907, simple_loss=0.2617, pruned_loss=0.05988, over 16737.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2667, pruned_loss=0.05781, over 3252540.69 frames. ], batch size: 124, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:29:12,912 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3272, 4.3031, 4.2364, 4.0402, 3.9345, 4.2729, 3.9679, 4.0059], device='cuda:2'), covar=tensor([0.0571, 0.0491, 0.0252, 0.0263, 0.0806, 0.0396, 0.0740, 0.0571], device='cuda:2'), in_proj_covar=tensor([0.0246, 0.0306, 0.0290, 0.0269, 0.0316, 0.0306, 0.0199, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 05:29:12,947 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:23,132 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7843, 3.7429, 4.2484, 2.1147, 4.3779, 4.3934, 3.1804, 3.1632], device='cuda:2'), covar=tensor([0.0694, 0.0190, 0.0104, 0.1108, 0.0044, 0.0090, 0.0327, 0.0408], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0101, 0.0087, 0.0139, 0.0070, 0.0101, 0.0122, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 05:29:23,168 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:25,657 INFO [optim.py:368] (2/8) 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 05:29:38,104 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 10, batch 4000, loss[loss=0.2065, simple_loss=0.2739, pruned_loss=0.06951, over 16892.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2668, pruned_loss=0.05833, over 3255079.68 frames. ], batch size: 109, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:31:05,661 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:31:07,588 INFO [train.py:904] (2/8) Epoch 10, batch 4050, loss[loss=0.1764, simple_loss=0.259, pruned_loss=0.04686, over 16979.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.267, pruned_loss=0.05724, over 3257865.55 frames. ], batch size: 41, lr: 6.89e-03, grad_scale: 16.0 2023-04-29 05:31:21,250 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-04-29 05:31:49,146 INFO [optim.py:368] (2/8) 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 05:32:20,029 INFO [train.py:904] (2/8) Epoch 10, batch 4100, loss[loss=0.2296, simple_loss=0.2994, pruned_loss=0.07989, over 11996.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2688, pruned_loss=0.05699, over 3240185.43 frames. ], batch size: 246, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:32:34,843 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:33:20,091 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0579, 3.3815, 3.5605, 3.4939, 3.5096, 3.2982, 3.3393, 3.3766], device='cuda:2'), covar=tensor([0.0421, 0.0516, 0.0382, 0.0468, 0.0484, 0.0433, 0.0785, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0337, 0.0340, 0.0325, 0.0385, 0.0361, 0.0465, 0.0286], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 05:33:26,041 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9636, 3.8426, 3.9730, 4.1257, 4.2179, 3.7998, 4.1685, 4.2478], device='cuda:2'), covar=tensor([0.1105, 0.0850, 0.1350, 0.0538, 0.0488, 0.1522, 0.0572, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0520, 0.0640, 0.0788, 0.0652, 0.0489, 0.0499, 0.0513, 0.0581], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:33:33,930 INFO [train.py:904] (2/8) Epoch 10, batch 4150, loss[loss=0.2261, simple_loss=0.3129, pruned_loss=0.06963, over 16663.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.276, pruned_loss=0.05932, over 3228240.28 frames. ], batch size: 134, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:34:17,107 INFO [optim.py:368] (2/8) 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 05:34:49,630 INFO [train.py:904] (2/8) Epoch 10, batch 4200, loss[loss=0.2543, simple_loss=0.3247, pruned_loss=0.09194, over 11508.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2841, pruned_loss=0.0619, over 3201817.65 frames. ], batch size: 248, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:35:12,280 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1009, 5.0644, 4.9185, 4.6627, 4.5279, 4.9923, 4.8201, 4.5774], device='cuda:2'), covar=tensor([0.0473, 0.0419, 0.0209, 0.0232, 0.0936, 0.0341, 0.0302, 0.0599], device='cuda:2'), in_proj_covar=tensor([0.0241, 0.0299, 0.0286, 0.0261, 0.0310, 0.0297, 0.0196, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:35:42,516 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 05:35:47,601 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 05:36:04,061 INFO [train.py:904] (2/8) Epoch 10, batch 4250, loss[loss=0.1872, simple_loss=0.2826, pruned_loss=0.04586, over 16327.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2863, pruned_loss=0.06096, over 3201470.47 frames. ], batch size: 146, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:36:24,740 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:26,027 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:27,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4634, 4.5320, 4.3045, 4.0546, 3.9575, 4.4385, 4.1022, 4.0765], device='cuda:2'), covar=tensor([0.0505, 0.0324, 0.0256, 0.0291, 0.0787, 0.0325, 0.0602, 0.0597], device='cuda:2'), in_proj_covar=tensor([0.0239, 0.0297, 0.0284, 0.0260, 0.0307, 0.0293, 0.0195, 0.0330], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:36:37,902 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:47,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8653, 1.3512, 1.5611, 1.6852, 1.8893, 1.9521, 1.4911, 1.8441], device='cuda:2'), covar=tensor([0.0181, 0.0278, 0.0152, 0.0215, 0.0175, 0.0113, 0.0261, 0.0081], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0169, 0.0153, 0.0158, 0.0165, 0.0121, 0.0168, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 05:36:49,133 INFO [optim.py:368] (2/8) 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 05:37:19,477 INFO [train.py:904] (2/8) Epoch 10, batch 4300, loss[loss=0.2073, simple_loss=0.3006, pruned_loss=0.05698, over 16770.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2879, pruned_loss=0.06021, over 3195347.33 frames. ], batch size: 124, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:37:22,064 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 05:37:32,832 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9680, 1.7668, 2.2889, 2.8678, 2.8521, 3.1852, 1.8056, 3.1286], device='cuda:2'), covar=tensor([0.0131, 0.0352, 0.0249, 0.0198, 0.0172, 0.0107, 0.0396, 0.0075], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0171, 0.0154, 0.0159, 0.0166, 0.0122, 0.0170, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 05:37:59,690 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:16,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3966, 3.7982, 3.7587, 1.8736, 3.0093, 2.4531, 3.6808, 3.8322], device='cuda:2'), covar=tensor([0.0252, 0.0577, 0.0447, 0.1810, 0.0762, 0.0830, 0.0637, 0.0809], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0142, 0.0154, 0.0140, 0.0134, 0.0122, 0.0134, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 05:38:24,181 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:33,664 INFO [train.py:904] (2/8) Epoch 10, batch 4350, loss[loss=0.231, simple_loss=0.3179, pruned_loss=0.07203, over 16866.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2916, pruned_loss=0.06188, over 3188989.38 frames. ], batch size: 116, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:38:39,565 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 05:39:18,685 INFO [optim.py:368] (2/8) 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 05:39:49,467 INFO [train.py:904] (2/8) Epoch 10, batch 4400, loss[loss=0.2471, simple_loss=0.3308, pruned_loss=0.08173, over 16467.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2949, pruned_loss=0.06341, over 3188766.62 frames. ], batch size: 75, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:40:02,134 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:01,592 INFO [train.py:904] (2/8) Epoch 10, batch 4450, loss[loss=0.2168, simple_loss=0.3084, pruned_loss=0.06257, over 17232.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2973, pruned_loss=0.06371, over 3198781.60 frames. ], batch size: 52, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:41:12,949 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:41:17,805 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:41:46,152 INFO [optim.py:368] (2/8) 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 05:42:08,851 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:42:15,000 INFO [train.py:904] (2/8) Epoch 10, batch 4500, loss[loss=0.2101, simple_loss=0.2932, pruned_loss=0.06349, over 16816.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2969, pruned_loss=0.06396, over 3198943.65 frames. ], batch size: 39, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:42:45,973 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:43:20,296 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9533, 4.9505, 5.3601, 5.3369, 5.3799, 4.9581, 4.9708, 4.6268], device='cuda:2'), covar=tensor([0.0232, 0.0308, 0.0267, 0.0330, 0.0335, 0.0251, 0.0704, 0.0358], device='cuda:2'), in_proj_covar=tensor([0.0318, 0.0321, 0.0325, 0.0311, 0.0370, 0.0344, 0.0447, 0.0276], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 05:43:27,209 INFO [train.py:904] (2/8) Epoch 10, batch 4550, loss[loss=0.2109, simple_loss=0.2884, pruned_loss=0.06667, over 12306.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2973, pruned_loss=0.06446, over 3210327.97 frames. ], batch size: 246, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:43:35,512 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:43:47,801 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:43:59,552 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:44:10,340 INFO [optim.py:368] (2/8) 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 05:44:39,215 INFO [train.py:904] (2/8) Epoch 10, batch 4600, loss[loss=0.2044, simple_loss=0.2894, pruned_loss=0.05963, over 16588.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2978, pruned_loss=0.06424, over 3228747.77 frames. ], batch size: 35, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:44:57,859 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:09,542 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:11,439 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:20,990 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:35,577 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 05:45:43,253 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:55,174 INFO [train.py:904] (2/8) Epoch 10, batch 4650, loss[loss=0.1831, simple_loss=0.2734, pruned_loss=0.04642, over 16842.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2958, pruned_loss=0.06368, over 3223137.11 frames. ], batch size: 96, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:46:20,291 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0191, 2.3310, 2.3070, 2.6848, 2.2684, 3.1395, 1.7824, 2.7393], device='cuda:2'), covar=tensor([0.1036, 0.0562, 0.0951, 0.0135, 0.0152, 0.0308, 0.1210, 0.0626], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0156, 0.0177, 0.0137, 0.0201, 0.0205, 0.0176, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 05:46:40,627 INFO [optim.py:368] (2/8) 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 05:46:55,309 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96041.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:46:58,142 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:47:05,501 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6004, 2.6122, 1.6005, 2.7232, 2.1194, 2.7540, 1.8343, 2.2632], device='cuda:2'), covar=tensor([0.0235, 0.0366, 0.1375, 0.0165, 0.0684, 0.0394, 0.1105, 0.0555], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0164, 0.0186, 0.0121, 0.0165, 0.0202, 0.0190, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 05:47:10,297 INFO [train.py:904] (2/8) Epoch 10, batch 4700, loss[loss=0.1847, simple_loss=0.2724, pruned_loss=0.0485, over 16909.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2932, pruned_loss=0.06251, over 3216669.66 frames. ], batch size: 96, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:47:52,211 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:48:23,998 INFO [train.py:904] (2/8) Epoch 10, batch 4750, loss[loss=0.184, simple_loss=0.2664, pruned_loss=0.05083, over 17000.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2894, pruned_loss=0.06047, over 3214599.24 frames. ], batch size: 55, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:08,956 INFO [optim.py:368] (2/8) 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 05:49:22,638 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:49:38,024 INFO [train.py:904] (2/8) Epoch 10, batch 4800, loss[loss=0.2081, simple_loss=0.3031, pruned_loss=0.05649, over 15256.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2869, pruned_loss=0.0591, over 3198327.83 frames. ], batch size: 191, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:38,357 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3778, 4.6742, 4.4338, 4.4450, 4.2100, 4.0855, 4.2228, 4.6608], device='cuda:2'), covar=tensor([0.1040, 0.0716, 0.0876, 0.0619, 0.0681, 0.1407, 0.0926, 0.0843], device='cuda:2'), in_proj_covar=tensor([0.0506, 0.0641, 0.0531, 0.0439, 0.0401, 0.0414, 0.0530, 0.0488], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:49:58,800 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:50:03,011 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:50:19,741 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8761, 3.7650, 3.9708, 4.1175, 4.2082, 3.8222, 4.1429, 4.2440], device='cuda:2'), covar=tensor([0.1206, 0.0992, 0.1171, 0.0522, 0.0459, 0.1203, 0.0654, 0.0504], device='cuda:2'), in_proj_covar=tensor([0.0494, 0.0612, 0.0747, 0.0623, 0.0465, 0.0479, 0.0488, 0.0551], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 05:50:54,658 INFO [train.py:904] (2/8) Epoch 10, batch 4850, loss[loss=0.2156, simple_loss=0.3015, pruned_loss=0.06488, over 15392.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2875, pruned_loss=0.05883, over 3192513.93 frames. ], batch size: 191, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:50:56,264 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:51:32,228 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96226.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:51:40,157 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 10, batch 4900, loss[loss=0.1835, simple_loss=0.2798, pruned_loss=0.0436, over 16363.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2867, pruned_loss=0.05765, over 3166455.94 frames. ], batch size: 146, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:52:42,394 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:24,548 INFO [train.py:904] (2/8) Epoch 10, batch 4950, loss[loss=0.1965, simple_loss=0.2883, pruned_loss=0.05233, over 16687.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2867, pruned_loss=0.05757, over 3175599.22 frames. ], batch size: 134, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:53:51,587 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:56,282 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:05,872 INFO [optim.py:368] (2/8) 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 05:54:12,441 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:33,261 INFO [train.py:904] (2/8) Epoch 10, batch 5000, loss[loss=0.2292, simple_loss=0.3088, pruned_loss=0.07481, over 17019.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2884, pruned_loss=0.05759, over 3196924.85 frames. ], batch size: 55, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:55:21,016 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:27,996 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:44,813 INFO [train.py:904] (2/8) Epoch 10, batch 5050, loss[loss=0.2518, simple_loss=0.3249, pruned_loss=0.0893, over 12050.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2888, pruned_loss=0.05754, over 3209941.04 frames. ], batch size: 247, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:56:02,196 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 05:56:27,889 INFO [optim.py:368] (2/8) 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 05:56:33,445 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:54,580 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:55,807 INFO [train.py:904] (2/8) Epoch 10, batch 5100, loss[loss=0.1788, simple_loss=0.2671, pruned_loss=0.04524, over 16516.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2865, pruned_loss=0.05671, over 3203825.25 frames. ], batch size: 68, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:57:20,195 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:57:58,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3130, 4.2532, 4.6828, 4.6585, 4.6603, 4.3633, 4.3608, 4.2761], device='cuda:2'), covar=tensor([0.0266, 0.0515, 0.0298, 0.0305, 0.0357, 0.0279, 0.0705, 0.0365], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0322, 0.0327, 0.0310, 0.0373, 0.0347, 0.0446, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 05:58:08,615 INFO [train.py:904] (2/8) Epoch 10, batch 5150, loss[loss=0.215, simple_loss=0.3069, pruned_loss=0.06157, over 16907.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2867, pruned_loss=0.05597, over 3193029.67 frames. ], batch size: 116, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:58:11,276 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:29,578 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:58:36,739 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:52,021 INFO [optim.py:368] (2/8) 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 05:59:21,921 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:59:22,747 INFO [train.py:904] (2/8) Epoch 10, batch 5200, loss[loss=0.1979, simple_loss=0.2839, pruned_loss=0.05595, over 16914.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2862, pruned_loss=0.05603, over 3190107.66 frames. ], batch size: 109, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:59:40,493 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3240, 4.0809, 4.1637, 2.4875, 3.6118, 4.0558, 3.7392, 2.1807], device='cuda:2'), covar=tensor([0.0432, 0.0021, 0.0024, 0.0328, 0.0060, 0.0060, 0.0048, 0.0381], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0066, 0.0068, 0.0123, 0.0076, 0.0085, 0.0074, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 05:59:55,762 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 06:00:35,342 INFO [train.py:904] (2/8) Epoch 10, batch 5250, loss[loss=0.1847, simple_loss=0.2702, pruned_loss=0.04964, over 16742.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2831, pruned_loss=0.05535, over 3208408.34 frames. ], batch size: 57, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:00:38,872 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6721, 2.5720, 2.2289, 3.8595, 2.8224, 3.8926, 1.3597, 2.7336], device='cuda:2'), covar=tensor([0.1211, 0.0651, 0.1209, 0.0101, 0.0190, 0.0333, 0.1495, 0.0794], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0154, 0.0175, 0.0134, 0.0197, 0.0203, 0.0175, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 06:01:21,023 INFO [optim.py:368] (2/8) 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 06:01:26,348 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:01:48,984 INFO [train.py:904] (2/8) Epoch 10, batch 5300, loss[loss=0.1703, simple_loss=0.2512, pruned_loss=0.04473, over 17218.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2792, pruned_loss=0.05382, over 3221817.17 frames. ], batch size: 44, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:02:17,488 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6058, 4.5302, 4.5014, 2.8490, 3.8053, 4.4018, 3.9869, 2.4253], device='cuda:2'), covar=tensor([0.0356, 0.0013, 0.0019, 0.0262, 0.0060, 0.0048, 0.0044, 0.0315], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0065, 0.0067, 0.0122, 0.0076, 0.0084, 0.0074, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:02:30,153 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:02:38,075 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5844, 4.8563, 5.0222, 4.9299, 4.9046, 5.4292, 4.9783, 4.7221], device='cuda:2'), covar=tensor([0.1096, 0.1409, 0.1575, 0.1504, 0.2273, 0.0931, 0.1242, 0.2307], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0460, 0.0486, 0.0403, 0.0534, 0.0523, 0.0398, 0.0550], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:03:03,544 INFO [train.py:904] (2/8) Epoch 10, batch 5350, loss[loss=0.1917, simple_loss=0.2845, pruned_loss=0.04942, over 16713.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2775, pruned_loss=0.05283, over 3225955.17 frames. ], batch size: 83, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:03:03,907 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9359, 5.6420, 5.8468, 5.5486, 5.5620, 6.1135, 5.7012, 5.5170], device='cuda:2'), covar=tensor([0.0751, 0.1466, 0.1421, 0.1559, 0.2300, 0.0811, 0.1100, 0.2071], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0461, 0.0486, 0.0403, 0.0536, 0.0523, 0.0399, 0.0550], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:03:48,671 INFO [optim.py:368] (2/8) 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 06:03:52,797 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:08,499 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96746.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:15,624 INFO [train.py:904] (2/8) Epoch 10, batch 5400, loss[loss=0.1801, simple_loss=0.2766, pruned_loss=0.04185, over 16696.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2799, pruned_loss=0.05297, over 3239813.91 frames. ], batch size: 89, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:04:51,432 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:02,543 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:31,405 INFO [train.py:904] (2/8) Epoch 10, batch 5450, loss[loss=0.2614, simple_loss=0.3275, pruned_loss=0.09771, over 12356.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2834, pruned_loss=0.05503, over 3225463.74 frames. ], batch size: 247, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:05:41,639 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 06:05:50,157 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 06:06:02,205 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:20,032 INFO [optim.py:368] (2/8) 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 06:06:27,174 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:49,558 INFO [train.py:904] (2/8) Epoch 10, batch 5500, loss[loss=0.2431, simple_loss=0.3151, pruned_loss=0.08553, over 16396.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2915, pruned_loss=0.06105, over 3165512.14 frames. ], batch size: 146, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:06:55,149 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7872, 1.7282, 1.4392, 1.3801, 1.8126, 1.5551, 1.6715, 1.8797], device='cuda:2'), covar=tensor([0.0101, 0.0201, 0.0305, 0.0278, 0.0157, 0.0203, 0.0138, 0.0141], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0195, 0.0193, 0.0191, 0.0194, 0.0196, 0.0197, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:07:17,928 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:08:08,494 INFO [train.py:904] (2/8) Epoch 10, batch 5550, loss[loss=0.2056, simple_loss=0.2942, pruned_loss=0.05847, over 16821.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2993, pruned_loss=0.06727, over 3116471.82 frames. ], batch size: 102, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:08:56,582 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7452, 3.6337, 3.8218, 3.6270, 3.7873, 4.1362, 3.8991, 3.7116], device='cuda:2'), covar=tensor([0.2066, 0.2187, 0.2164, 0.2452, 0.2856, 0.1915, 0.1428, 0.2370], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0469, 0.0496, 0.0409, 0.0546, 0.0534, 0.0403, 0.0558], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:09:01,353 INFO [optim.py:368] (2/8) 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 06:09:21,020 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8259, 3.0294, 2.5655, 4.6390, 3.6500, 4.3227, 1.5701, 3.0282], device='cuda:2'), covar=tensor([0.1253, 0.0600, 0.1145, 0.0139, 0.0334, 0.0350, 0.1513, 0.0803], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0156, 0.0178, 0.0136, 0.0200, 0.0206, 0.0178, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 06:09:28,005 INFO [train.py:904] (2/8) Epoch 10, batch 5600, loss[loss=0.2316, simple_loss=0.3133, pruned_loss=0.07498, over 17207.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3041, pruned_loss=0.07137, over 3094594.99 frames. ], batch size: 45, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:09:34,187 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 06:10:15,501 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:51,173 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:53,953 INFO [train.py:904] (2/8) Epoch 10, batch 5650, loss[loss=0.2671, simple_loss=0.3272, pruned_loss=0.1035, over 11393.00 frames. ], tot_loss[loss=0.231, simple_loss=0.31, pruned_loss=0.07606, over 3082773.87 frames. ], batch size: 247, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:10:59,000 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8833, 4.1348, 3.8805, 3.9499, 3.6264, 3.7285, 3.8253, 4.0872], device='cuda:2'), covar=tensor([0.0942, 0.0876, 0.1073, 0.0788, 0.0800, 0.1648, 0.0900, 0.1094], device='cuda:2'), in_proj_covar=tensor([0.0499, 0.0636, 0.0529, 0.0440, 0.0399, 0.0409, 0.0526, 0.0486], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:11:03,436 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4183, 2.0045, 2.1444, 4.0602, 1.9376, 2.4310, 2.1122, 2.2786], device='cuda:2'), covar=tensor([0.0861, 0.3033, 0.2102, 0.0374, 0.3600, 0.2095, 0.2779, 0.2896], device='cuda:2'), in_proj_covar=tensor([0.0356, 0.0383, 0.0320, 0.0318, 0.0403, 0.0436, 0.0343, 0.0448], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:11:34,449 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 06:12:02,234 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:12:11,111 INFO [train.py:904] (2/8) Epoch 10, batch 5700, loss[loss=0.2509, simple_loss=0.3375, pruned_loss=0.08218, over 15398.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3119, pruned_loss=0.07787, over 3085323.87 frames. ], batch size: 191, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:12:25,235 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:12:42,892 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-29 06:13:16,337 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:29,504 INFO [train.py:904] (2/8) Epoch 10, batch 5750, loss[loss=0.2185, simple_loss=0.3009, pruned_loss=0.06806, over 16788.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3147, pruned_loss=0.07983, over 3037526.62 frames. ], batch size: 83, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:13:32,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7805, 1.7696, 1.5474, 1.5386, 1.8489, 1.6317, 1.6514, 1.9187], device='cuda:2'), covar=tensor([0.0098, 0.0162, 0.0273, 0.0239, 0.0121, 0.0179, 0.0127, 0.0127], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0194, 0.0192, 0.0191, 0.0193, 0.0196, 0.0197, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:13:54,873 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6009, 5.9531, 5.6697, 5.7778, 5.3087, 5.1963, 5.4624, 6.1520], device='cuda:2'), covar=tensor([0.0938, 0.0718, 0.0924, 0.0712, 0.0696, 0.0576, 0.0942, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0499, 0.0631, 0.0527, 0.0436, 0.0397, 0.0408, 0.0522, 0.0481], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:14:17,842 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:14:22,108 INFO [optim.py:368] (2/8) 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 06:14:49,725 INFO [train.py:904] (2/8) Epoch 10, batch 5800, loss[loss=0.2184, simple_loss=0.3107, pruned_loss=0.063, over 17118.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3147, pruned_loss=0.07901, over 3033208.28 frames. ], batch size: 48, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:15:47,678 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6397, 4.9186, 5.0462, 4.9067, 4.8857, 5.4643, 5.0247, 4.8372], device='cuda:2'), covar=tensor([0.1006, 0.1723, 0.1751, 0.1715, 0.2271, 0.0943, 0.1293, 0.2227], device='cuda:2'), in_proj_covar=tensor([0.0343, 0.0473, 0.0499, 0.0411, 0.0549, 0.0537, 0.0409, 0.0564], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:16:07,864 INFO [train.py:904] (2/8) Epoch 10, batch 5850, loss[loss=0.2197, simple_loss=0.304, pruned_loss=0.06768, over 16676.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3123, pruned_loss=0.07688, over 3041242.40 frames. ], batch size: 134, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:16:10,403 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5231, 3.7900, 3.8103, 1.9960, 3.0753, 2.4865, 3.7890, 3.9377], device='cuda:2'), covar=tensor([0.0222, 0.0600, 0.0469, 0.1828, 0.0727, 0.0833, 0.0611, 0.0841], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0141, 0.0157, 0.0142, 0.0134, 0.0125, 0.0135, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 06:17:00,834 INFO [optim.py:368] (2/8) 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 06:17:14,103 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 06:17:28,941 INFO [train.py:904] (2/8) Epoch 10, batch 5900, loss[loss=0.2134, simple_loss=0.3021, pruned_loss=0.06232, over 16905.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.312, pruned_loss=0.07645, over 3047542.03 frames. ], batch size: 109, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:17:37,278 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2039, 4.2377, 4.4450, 4.2577, 4.3210, 4.7924, 4.3411, 4.1594], device='cuda:2'), covar=tensor([0.1690, 0.1777, 0.1673, 0.1825, 0.2483, 0.1069, 0.1460, 0.2383], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0476, 0.0502, 0.0414, 0.0552, 0.0539, 0.0411, 0.0566], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:17:45,029 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:18:49,537 INFO [train.py:904] (2/8) Epoch 10, batch 5950, loss[loss=0.2118, simple_loss=0.2914, pruned_loss=0.06607, over 16622.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3119, pruned_loss=0.07472, over 3063717.94 frames. ], batch size: 62, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:19:03,969 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8052, 3.8240, 4.2696, 4.2370, 4.2166, 3.9357, 3.9766, 3.9002], device='cuda:2'), covar=tensor([0.0346, 0.0562, 0.0349, 0.0395, 0.0466, 0.0418, 0.0828, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0319, 0.0323, 0.0310, 0.0372, 0.0346, 0.0442, 0.0278], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 06:19:14,595 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0105, 2.5842, 2.5869, 1.8646, 2.7691, 2.8336, 2.3886, 2.3882], device='cuda:2'), covar=tensor([0.0664, 0.0195, 0.0188, 0.0867, 0.0079, 0.0159, 0.0390, 0.0387], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0098, 0.0086, 0.0137, 0.0068, 0.0096, 0.0120, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 06:19:20,390 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:19:31,590 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 06:19:41,579 INFO [optim.py:368] (2/8) 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 06:19:52,696 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:01,356 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 10, batch 6000, loss[loss=0.2024, simple_loss=0.2871, pruned_loss=0.05889, over 16732.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.31, pruned_loss=0.07318, over 3088591.42 frames. ], batch size: 124, lr: 6.82e-03, grad_scale: 4.0 2023-04-29 06:20:09,122 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 06:20:23,723 INFO [train.py:938] (2/8) Epoch 10, validation: loss=0.165, simple_loss=0.2783, pruned_loss=0.02583, over 944034.00 frames. 2023-04-29 06:20:23,724 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 06:20:30,443 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:57,535 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:21:35,172 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7937, 5.1997, 5.4255, 5.1152, 5.1207, 5.7867, 5.2994, 5.0644], device='cuda:2'), covar=tensor([0.1023, 0.1852, 0.1694, 0.1919, 0.2504, 0.0951, 0.1309, 0.2172], device='cuda:2'), in_proj_covar=tensor([0.0342, 0.0473, 0.0501, 0.0410, 0.0545, 0.0535, 0.0408, 0.0561], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:21:42,482 INFO [train.py:904] (2/8) Epoch 10, batch 6050, loss[loss=0.2149, simple_loss=0.3021, pruned_loss=0.06381, over 16744.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3083, pruned_loss=0.07213, over 3099572.07 frames. ], batch size: 124, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:21:44,931 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 06:21:52,137 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:29,743 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:35,029 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:35,672 INFO [optim.py:368] (2/8) 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 06:22:50,667 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1936, 1.9731, 1.6087, 1.7106, 2.2899, 1.9547, 2.1559, 2.3760], device='cuda:2'), covar=tensor([0.0108, 0.0246, 0.0335, 0.0308, 0.0160, 0.0270, 0.0139, 0.0150], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0194, 0.0192, 0.0193, 0.0194, 0.0197, 0.0198, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:23:02,108 INFO [train.py:904] (2/8) Epoch 10, batch 6100, loss[loss=0.2436, simple_loss=0.3093, pruned_loss=0.08894, over 11582.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3076, pruned_loss=0.07151, over 3076676.63 frames. ], batch size: 247, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:23:48,674 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:23:57,640 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2253, 2.0674, 2.6919, 3.0210, 3.0757, 3.5480, 2.0679, 3.4275], device='cuda:2'), covar=tensor([0.0116, 0.0313, 0.0193, 0.0168, 0.0161, 0.0098, 0.0339, 0.0084], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0165, 0.0150, 0.0154, 0.0162, 0.0118, 0.0169, 0.0109], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 06:24:23,704 INFO [train.py:904] (2/8) Epoch 10, batch 6150, loss[loss=0.2018, simple_loss=0.2902, pruned_loss=0.05664, over 16405.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3063, pruned_loss=0.07143, over 3066984.34 frames. ], batch size: 146, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:24:53,098 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7961, 3.6955, 3.9295, 3.6616, 3.8285, 4.1945, 3.8774, 3.6836], device='cuda:2'), covar=tensor([0.2226, 0.1947, 0.1865, 0.2487, 0.2848, 0.1792, 0.1402, 0.2469], device='cuda:2'), in_proj_covar=tensor([0.0345, 0.0476, 0.0506, 0.0413, 0.0551, 0.0539, 0.0412, 0.0565], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:25:17,520 INFO [optim.py:368] (2/8) 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 06:25:41,181 INFO [train.py:904] (2/8) Epoch 10, batch 6200, loss[loss=0.2363, simple_loss=0.3139, pruned_loss=0.07934, over 15253.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3045, pruned_loss=0.0708, over 3083498.49 frames. ], batch size: 190, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:46,657 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:26:07,072 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7842, 3.9147, 4.3331, 1.9070, 4.6097, 4.5389, 3.2003, 3.3124], device='cuda:2'), covar=tensor([0.0696, 0.0182, 0.0130, 0.1194, 0.0036, 0.0076, 0.0320, 0.0390], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0098, 0.0086, 0.0139, 0.0068, 0.0096, 0.0120, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 06:26:57,938 INFO [train.py:904] (2/8) Epoch 10, batch 6250, loss[loss=0.2259, simple_loss=0.3104, pruned_loss=0.07073, over 16650.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3035, pruned_loss=0.06937, over 3115160.24 frames. ], batch size: 134, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:27:17,434 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 06:27:18,739 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:18,919 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:31,885 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5799, 2.3174, 2.3067, 4.4383, 2.1042, 2.7346, 2.2733, 2.4581], device='cuda:2'), covar=tensor([0.0864, 0.3072, 0.2084, 0.0285, 0.3581, 0.2101, 0.2969, 0.2758], device='cuda:2'), in_proj_covar=tensor([0.0358, 0.0382, 0.0319, 0.0314, 0.0405, 0.0436, 0.0344, 0.0446], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:27:47,814 INFO [optim.py:368] (2/8) 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 06:28:11,794 INFO [train.py:904] (2/8) Epoch 10, batch 6300, loss[loss=0.2542, simple_loss=0.3182, pruned_loss=0.0951, over 11519.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3033, pruned_loss=0.06864, over 3126022.60 frames. ], batch size: 246, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:28:18,581 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:28:48,991 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:28:49,149 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.37 vs. limit=5.0 2023-04-29 06:29:24,485 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:29:26,264 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 06:29:30,239 INFO [train.py:904] (2/8) Epoch 10, batch 6350, loss[loss=0.2125, simple_loss=0.3005, pruned_loss=0.06229, over 16820.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3043, pruned_loss=0.07035, over 3102174.38 frames. ], batch size: 102, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:29:31,921 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:33,046 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:30:13,656 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:30:21,622 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:30:22,846 INFO [optim.py:368] (2/8) 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 06:30:46,249 INFO [train.py:904] (2/8) Epoch 10, batch 6400, loss[loss=0.2152, simple_loss=0.2994, pruned_loss=0.06552, over 16693.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.3048, pruned_loss=0.07147, over 3093905.19 frames. ], batch size: 134, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:31:18,744 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6606, 3.8074, 1.9131, 4.1454, 2.7331, 4.1231, 2.1064, 3.0053], device='cuda:2'), covar=tensor([0.0176, 0.0275, 0.1687, 0.0117, 0.0689, 0.0393, 0.1510, 0.0580], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0160, 0.0184, 0.0118, 0.0163, 0.0199, 0.0188, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 06:32:00,790 INFO [train.py:904] (2/8) Epoch 10, batch 6450, loss[loss=0.2246, simple_loss=0.3118, pruned_loss=0.06869, over 16318.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3047, pruned_loss=0.07045, over 3106441.51 frames. ], batch size: 35, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:07,202 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0332, 2.7893, 2.8430, 2.0662, 2.6599, 2.1806, 2.7282, 3.0232], device='cuda:2'), covar=tensor([0.0273, 0.0714, 0.0420, 0.1514, 0.0684, 0.0840, 0.0472, 0.0548], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0141, 0.0156, 0.0141, 0.0135, 0.0124, 0.0136, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 06:32:52,375 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 06:32:57,254 INFO [optim.py:368] (2/8) 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 06:33:21,800 INFO [train.py:904] (2/8) Epoch 10, batch 6500, loss[loss=0.2235, simple_loss=0.304, pruned_loss=0.07144, over 16698.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3031, pruned_loss=0.07034, over 3104805.51 frames. ], batch size: 89, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:17,032 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:34:41,150 INFO [train.py:904] (2/8) Epoch 10, batch 6550, loss[loss=0.2645, simple_loss=0.3206, pruned_loss=0.1042, over 11440.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3054, pruned_loss=0.07102, over 3120301.90 frames. ], batch size: 246, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:55,678 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:04,976 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:35,376 INFO [optim.py:368] (2/8) 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 06:35:42,216 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1492, 4.1537, 4.3511, 4.1308, 4.2723, 4.7676, 4.3487, 4.0750], device='cuda:2'), covar=tensor([0.1817, 0.1968, 0.1912, 0.1959, 0.2517, 0.1144, 0.1431, 0.2497], device='cuda:2'), in_proj_covar=tensor([0.0344, 0.0481, 0.0510, 0.0414, 0.0554, 0.0544, 0.0416, 0.0570], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:35:56,855 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:36:00,714 INFO [train.py:904] (2/8) Epoch 10, batch 6600, loss[loss=0.2072, simple_loss=0.296, pruned_loss=0.05922, over 16474.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3077, pruned_loss=0.07172, over 3107225.58 frames. ], batch size: 68, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:36:19,346 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:14,619 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 10, batch 6650, loss[loss=0.2818, simple_loss=0.3387, pruned_loss=0.1125, over 11835.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3073, pruned_loss=0.0719, over 3113509.75 frames. ], batch size: 247, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:37:24,849 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:05,361 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 06:38:05,388 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:14,889 INFO [optim.py:368] (2/8) 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 06:38:31,002 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:31,350 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 06:38:38,424 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:38,453 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7826, 3.7525, 3.8880, 3.7028, 3.7724, 4.1931, 3.9229, 3.6674], device='cuda:2'), covar=tensor([0.1839, 0.2079, 0.1960, 0.2472, 0.2980, 0.1592, 0.1427, 0.2836], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0472, 0.0505, 0.0408, 0.0547, 0.0537, 0.0412, 0.0563], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 06:38:39,791 INFO [train.py:904] (2/8) Epoch 10, batch 6700, loss[loss=0.2711, simple_loss=0.3333, pruned_loss=0.1045, over 11715.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3064, pruned_loss=0.07258, over 3093210.85 frames. ], batch size: 250, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:39:20,426 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:39:57,931 INFO [train.py:904] (2/8) Epoch 10, batch 6750, loss[loss=0.1767, simple_loss=0.2684, pruned_loss=0.0425, over 16921.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3051, pruned_loss=0.07188, over 3110539.54 frames. ], batch size: 96, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:40:04,368 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:40:04,625 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 06:40:49,800 INFO [optim.py:368] (2/8) 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 06:41:15,033 INFO [train.py:904] (2/8) Epoch 10, batch 6800, loss[loss=0.2443, simple_loss=0.3213, pruned_loss=0.08364, over 15348.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3055, pruned_loss=0.07214, over 3098296.77 frames. ], batch size: 191, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:41:39,123 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:42:33,969 INFO [train.py:904] (2/8) Epoch 10, batch 6850, loss[loss=0.2221, simple_loss=0.3138, pruned_loss=0.06522, over 16620.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3062, pruned_loss=0.07221, over 3102320.99 frames. ], batch size: 134, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:42:47,599 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:42:48,170 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-29 06:43:24,606 INFO [optim.py:368] (2/8) 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 06:43:37,276 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:43:49,036 INFO [train.py:904] (2/8) Epoch 10, batch 6900, loss[loss=0.2383, simple_loss=0.3205, pruned_loss=0.07803, over 16726.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3085, pruned_loss=0.07185, over 3114637.80 frames. ], batch size: 124, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:44:01,152 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:44:05,090 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 06:45:09,157 INFO [train.py:904] (2/8) Epoch 10, batch 6950, loss[loss=0.198, simple_loss=0.2877, pruned_loss=0.05418, over 16467.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3094, pruned_loss=0.07245, over 3126018.14 frames. ], batch size: 75, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:45:54,257 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 06:46:01,761 INFO [optim.py:368] (2/8) 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 06:46:04,487 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 06:46:27,403 INFO [train.py:904] (2/8) Epoch 10, batch 7000, loss[loss=0.1924, simple_loss=0.2996, pruned_loss=0.04261, over 17048.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.31, pruned_loss=0.07244, over 3117008.59 frames. ], batch size: 50, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:03,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4832, 3.3954, 2.7685, 2.0477, 2.4076, 2.2287, 3.5650, 3.1903], device='cuda:2'), covar=tensor([0.2536, 0.0777, 0.1485, 0.2239, 0.2115, 0.1759, 0.0464, 0.1047], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0255, 0.0280, 0.0275, 0.0279, 0.0215, 0.0264, 0.0287], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:47:08,046 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:47:43,137 INFO [train.py:904] (2/8) Epoch 10, batch 7050, loss[loss=0.2214, simple_loss=0.3024, pruned_loss=0.07025, over 16675.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3108, pruned_loss=0.07253, over 3097541.71 frames. ], batch size: 62, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:56,614 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:48:34,408 INFO [optim.py:368] (2/8) 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 06:48:59,553 INFO [train.py:904] (2/8) Epoch 10, batch 7100, loss[loss=0.207, simple_loss=0.2923, pruned_loss=0.06083, over 16798.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3085, pruned_loss=0.07158, over 3105880.79 frames. ], batch size: 124, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:49:14,091 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:49:27,996 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:50:12,812 INFO [train.py:904] (2/8) Epoch 10, batch 7150, loss[loss=0.2686, simple_loss=0.3395, pruned_loss=0.09885, over 15346.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3073, pruned_loss=0.07219, over 3089476.36 frames. ], batch size: 191, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:51:03,784 INFO [optim.py:368] (2/8) 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 06:51:15,729 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:16,706 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:29,374 INFO [train.py:904] (2/8) Epoch 10, batch 7200, loss[loss=0.2049, simple_loss=0.2832, pruned_loss=0.06332, over 12005.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3049, pruned_loss=0.07018, over 3093877.16 frames. ], batch size: 248, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:52:32,649 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 10, batch 7250, loss[loss=0.1952, simple_loss=0.2786, pruned_loss=0.05591, over 16205.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3026, pruned_loss=0.06891, over 3084815.32 frames. ], batch size: 165, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:52:53,580 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:53:17,032 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 06:53:38,987 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9171, 2.9851, 2.9504, 4.8297, 3.7578, 4.4646, 1.7324, 2.9380], device='cuda:2'), covar=tensor([0.1238, 0.0645, 0.0958, 0.0146, 0.0319, 0.0302, 0.1360, 0.0838], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0156, 0.0179, 0.0135, 0.0201, 0.0205, 0.0177, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 06:53:45,148 INFO [optim.py:368] (2/8) 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 06:54:05,881 INFO [train.py:904] (2/8) Epoch 10, batch 7300, loss[loss=0.2156, simple_loss=0.3028, pruned_loss=0.06422, over 16416.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3015, pruned_loss=0.06857, over 3083780.55 frames. ], batch size: 146, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:54:08,853 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8246, 1.6900, 1.5108, 1.4191, 1.8462, 1.5230, 1.6813, 1.8546], device='cuda:2'), covar=tensor([0.0100, 0.0189, 0.0254, 0.0245, 0.0134, 0.0201, 0.0132, 0.0122], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0195, 0.0194, 0.0192, 0.0193, 0.0198, 0.0198, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 06:55:23,864 INFO [train.py:904] (2/8) Epoch 10, batch 7350, loss[loss=0.1933, simple_loss=0.2853, pruned_loss=0.05065, over 16785.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3024, pruned_loss=0.06998, over 3042239.87 frames. ], batch size: 124, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:19,126 INFO [optim.py:368] (2/8) 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 06:56:41,002 INFO [train.py:904] (2/8) Epoch 10, batch 7400, loss[loss=0.2363, simple_loss=0.3155, pruned_loss=0.0785, over 16890.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3036, pruned_loss=0.07049, over 3052568.75 frames. ], batch size: 116, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:57,651 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:03,069 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:59,089 INFO [train.py:904] (2/8) Epoch 10, batch 7450, loss[loss=0.2373, simple_loss=0.3251, pruned_loss=0.07474, over 16317.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3059, pruned_loss=0.07257, over 3020745.37 frames. ], batch size: 165, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:58:14,250 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:58:57,288 INFO [optim.py:368] (2/8) 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 06:59:19,906 INFO [train.py:904] (2/8) Epoch 10, batch 7500, loss[loss=0.2332, simple_loss=0.3123, pruned_loss=0.07704, over 16959.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3061, pruned_loss=0.07164, over 3027863.90 frames. ], batch size: 109, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 07:00:01,911 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4373, 2.1007, 1.6831, 1.8213, 2.3738, 2.1370, 2.4195, 2.6272], device='cuda:2'), covar=tensor([0.0093, 0.0289, 0.0402, 0.0372, 0.0171, 0.0258, 0.0162, 0.0164], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0193, 0.0191, 0.0191, 0.0192, 0.0195, 0.0195, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:00:20,547 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:27,376 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:34,223 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:38,279 INFO [train.py:904] (2/8) Epoch 10, batch 7550, loss[loss=0.2285, simple_loss=0.3076, pruned_loss=0.0747, over 15552.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3057, pruned_loss=0.07203, over 3025867.38 frames. ], batch size: 191, lr: 6.76e-03, grad_scale: 2.0 2023-04-29 07:00:38,835 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2932, 3.3595, 1.7921, 3.5736, 2.4151, 3.6148, 1.9429, 2.5580], device='cuda:2'), covar=tensor([0.0225, 0.0333, 0.1721, 0.0116, 0.0877, 0.0496, 0.1596, 0.0787], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0160, 0.0185, 0.0117, 0.0163, 0.0201, 0.0189, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 07:01:32,282 INFO [optim.py:368] (2/8) 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 07:01:53,309 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:01:53,973 INFO [train.py:904] (2/8) Epoch 10, batch 7600, loss[loss=0.2206, simple_loss=0.3023, pruned_loss=0.06945, over 16741.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3048, pruned_loss=0.07191, over 3045150.56 frames. ], batch size: 124, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:02:01,607 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:02:16,956 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 07:02:24,326 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:03:11,982 INFO [train.py:904] (2/8) Epoch 10, batch 7650, loss[loss=0.2106, simple_loss=0.2952, pruned_loss=0.06299, over 15195.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3057, pruned_loss=0.07297, over 3046220.83 frames. ], batch size: 190, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:03:59,277 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:08,571 INFO [optim.py:368] (2/8) 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 07:04:28,291 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 10, batch 7700, loss[loss=0.2402, simple_loss=0.3193, pruned_loss=0.08057, over 16254.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3056, pruned_loss=0.07344, over 3048801.49 frames. ], batch size: 35, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:04:50,085 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:05:28,817 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3126, 1.9758, 1.5461, 1.7525, 2.2622, 2.0645, 2.2803, 2.5259], device='cuda:2'), covar=tensor([0.0103, 0.0252, 0.0376, 0.0331, 0.0154, 0.0250, 0.0119, 0.0143], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0194, 0.0193, 0.0192, 0.0192, 0.0197, 0.0196, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:05:43,695 INFO [train.py:904] (2/8) Epoch 10, batch 7750, loss[loss=0.2742, simple_loss=0.3314, pruned_loss=0.1085, over 11597.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3059, pruned_loss=0.07311, over 3048506.20 frames. ], batch size: 248, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:05:59,743 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:06:02,448 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:06:38,055 INFO [optim.py:368] (2/8) 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 07:06:49,419 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7372, 3.6545, 3.8115, 3.9244, 3.9975, 3.5990, 3.9711, 4.0240], device='cuda:2'), covar=tensor([0.1590, 0.1031, 0.1334, 0.0682, 0.0680, 0.1670, 0.0814, 0.0690], device='cuda:2'), in_proj_covar=tensor([0.0496, 0.0615, 0.0755, 0.0633, 0.0483, 0.0477, 0.0505, 0.0564], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:06:59,329 INFO [train.py:904] (2/8) Epoch 10, batch 7800, loss[loss=0.2374, simple_loss=0.3122, pruned_loss=0.08129, over 16634.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3061, pruned_loss=0.07317, over 3071117.65 frames. ], batch size: 57, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:08:05,625 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9198, 3.1513, 3.0735, 2.0957, 2.8122, 3.0044, 3.0360, 1.8802], device='cuda:2'), covar=tensor([0.0412, 0.0039, 0.0048, 0.0333, 0.0084, 0.0099, 0.0065, 0.0351], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0065, 0.0069, 0.0125, 0.0076, 0.0087, 0.0074, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 07:08:12,877 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:08:16,575 INFO [train.py:904] (2/8) Epoch 10, batch 7850, loss[loss=0.196, simple_loss=0.2897, pruned_loss=0.05114, over 16690.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3073, pruned_loss=0.07345, over 3078459.36 frames. ], batch size: 89, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:08:57,398 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:10,220 INFO [optim.py:368] (2/8) 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 07:09:17,159 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 07:09:22,396 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:23,678 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:29,725 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:30,436 INFO [train.py:904] (2/8) Epoch 10, batch 7900, loss[loss=0.2304, simple_loss=0.3157, pruned_loss=0.07248, over 16876.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3066, pruned_loss=0.07301, over 3073974.40 frames. ], batch size: 116, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:09:35,942 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1139, 1.4667, 1.7881, 2.1089, 2.2415, 2.2841, 1.6191, 2.1682], device='cuda:2'), covar=tensor([0.0148, 0.0328, 0.0198, 0.0208, 0.0172, 0.0131, 0.0300, 0.0091], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0166, 0.0149, 0.0153, 0.0163, 0.0117, 0.0168, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 07:09:55,621 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:10:31,502 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99290.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:10:49,021 INFO [train.py:904] (2/8) Epoch 10, batch 7950, loss[loss=0.2712, simple_loss=0.3269, pruned_loss=0.1077, over 11244.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3068, pruned_loss=0.07382, over 3048942.63 frames. ], batch size: 248, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:10:53,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6521, 2.1766, 1.7719, 1.9745, 2.5290, 2.2936, 2.6462, 2.7780], device='cuda:2'), covar=tensor([0.0100, 0.0254, 0.0344, 0.0309, 0.0154, 0.0235, 0.0139, 0.0154], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0193, 0.0192, 0.0191, 0.0191, 0.0195, 0.0195, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:11:17,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4769, 3.5296, 3.1891, 3.0981, 3.1120, 3.4082, 3.2835, 3.1563], device='cuda:2'), covar=tensor([0.0510, 0.0506, 0.0227, 0.0226, 0.0521, 0.0391, 0.1093, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0234, 0.0296, 0.0272, 0.0253, 0.0294, 0.0287, 0.0188, 0.0318], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:11:22,466 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:27,241 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:30,516 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:33,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0619, 2.5766, 2.5978, 1.8829, 2.7913, 2.8664, 2.3293, 2.4278], device='cuda:2'), covar=tensor([0.0634, 0.0200, 0.0178, 0.0835, 0.0081, 0.0156, 0.0386, 0.0355], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0101, 0.0085, 0.0138, 0.0069, 0.0097, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 07:11:43,788 INFO [optim.py:368] (2/8) 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 07:11:56,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5478, 3.5630, 2.8870, 2.0537, 2.4614, 2.1561, 3.7846, 3.3913], device='cuda:2'), covar=tensor([0.2515, 0.0704, 0.1458, 0.2325, 0.2351, 0.1810, 0.0516, 0.1012], device='cuda:2'), in_proj_covar=tensor([0.0304, 0.0255, 0.0280, 0.0276, 0.0281, 0.0217, 0.0265, 0.0289], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:12:06,027 INFO [train.py:904] (2/8) Epoch 10, batch 8000, loss[loss=0.1897, simple_loss=0.2768, pruned_loss=0.05132, over 16701.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3075, pruned_loss=0.07461, over 3039453.57 frames. ], batch size: 57, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:12:57,228 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:21,633 INFO [train.py:904] (2/8) Epoch 10, batch 8050, loss[loss=0.2664, simple_loss=0.326, pruned_loss=0.1034, over 11814.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3068, pruned_loss=0.07418, over 3038937.01 frames. ], batch size: 246, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:13:29,868 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:50,409 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:14:01,649 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0521, 4.0281, 2.3156, 4.8108, 3.0012, 4.7631, 2.5842, 3.0805], device='cuda:2'), covar=tensor([0.0186, 0.0325, 0.1594, 0.0084, 0.0768, 0.0360, 0.1388, 0.0692], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0162, 0.0187, 0.0120, 0.0166, 0.0205, 0.0193, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 07:14:18,153 INFO [optim.py:368] (2/8) 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 07:14:39,575 INFO [train.py:904] (2/8) Epoch 10, batch 8100, loss[loss=0.21, simple_loss=0.2892, pruned_loss=0.06544, over 17216.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3057, pruned_loss=0.07298, over 3061107.98 frames. ], batch size: 52, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:15:03,928 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:15:23,135 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99481.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:15:54,070 INFO [train.py:904] (2/8) Epoch 10, batch 8150, loss[loss=0.2082, simple_loss=0.288, pruned_loss=0.06414, over 16425.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3033, pruned_loss=0.07198, over 3072264.25 frames. ], batch size: 35, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:16:35,896 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:16:49,692 INFO [optim.py:368] (2/8) 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 07:17:00,890 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:08,937 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:09,684 INFO [train.py:904] (2/8) Epoch 10, batch 8200, loss[loss=0.2046, simple_loss=0.2913, pruned_loss=0.05892, over 16900.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3006, pruned_loss=0.07073, over 3085840.28 frames. ], batch size: 109, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:17:35,553 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:51,239 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 07:18:04,664 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:11,208 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:20,129 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:28,353 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:32,973 INFO [train.py:904] (2/8) Epoch 10, batch 8250, loss[loss=0.1885, simple_loss=0.2748, pruned_loss=0.05112, over 16662.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2995, pruned_loss=0.06779, over 3085394.54 frames. ], batch size: 62, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:19:10,184 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:15,979 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:19,853 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:36,138 INFO [optim.py:368] (2/8) 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 07:19:55,542 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:56,723 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1201, 1.3579, 1.7559, 2.0450, 2.2258, 2.2371, 1.7101, 2.2334], device='cuda:2'), covar=tensor([0.0160, 0.0355, 0.0194, 0.0187, 0.0190, 0.0144, 0.0308, 0.0093], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0167, 0.0148, 0.0152, 0.0163, 0.0117, 0.0168, 0.0107], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 07:19:57,961 INFO [train.py:904] (2/8) Epoch 10, batch 8300, loss[loss=0.1841, simple_loss=0.279, pruned_loss=0.04466, over 15320.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2966, pruned_loss=0.06501, over 3057435.49 frames. ], batch size: 191, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:20:37,233 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,556 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,674 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 10, batch 8350, loss[loss=0.231, simple_loss=0.3043, pruned_loss=0.07885, over 12100.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2962, pruned_loss=0.06285, over 3070650.57 frames. ], batch size: 247, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:21:25,089 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4713, 2.5564, 1.9816, 2.1970, 2.9173, 2.6532, 3.2781, 3.2042], device='cuda:2'), covar=tensor([0.0067, 0.0264, 0.0379, 0.0333, 0.0190, 0.0228, 0.0150, 0.0144], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0190, 0.0190, 0.0187, 0.0190, 0.0192, 0.0192, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:21:29,860 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:21,849 INFO [optim.py:368] (2/8) 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 07:22:23,114 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3653, 4.2192, 4.4525, 4.5888, 4.7354, 4.2791, 4.7148, 4.7107], device='cuda:2'), covar=tensor([0.1522, 0.0959, 0.1285, 0.0584, 0.0494, 0.0930, 0.0468, 0.0537], device='cuda:2'), in_proj_covar=tensor([0.0488, 0.0610, 0.0740, 0.0626, 0.0476, 0.0477, 0.0497, 0.0560], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:22:26,052 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:43,905 INFO [train.py:904] (2/8) Epoch 10, batch 8400, loss[loss=0.1876, simple_loss=0.2767, pruned_loss=0.04927, over 15373.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2933, pruned_loss=0.06064, over 3055577.36 frames. ], batch size: 191, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:22:49,359 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:23:25,431 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:24:06,248 INFO [train.py:904] (2/8) Epoch 10, batch 8450, loss[loss=0.1846, simple_loss=0.279, pruned_loss=0.04504, over 16791.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2912, pruned_loss=0.05862, over 3051805.51 frames. ], batch size: 83, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:24:27,957 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7815, 4.9907, 5.2833, 5.0790, 5.0935, 5.6309, 5.2059, 4.9471], device='cuda:2'), covar=tensor([0.0763, 0.1791, 0.1705, 0.1654, 0.2179, 0.0860, 0.1265, 0.2156], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0453, 0.0488, 0.0396, 0.0521, 0.0517, 0.0398, 0.0539], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 07:24:42,070 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:25:06,072 INFO [optim.py:368] (2/8) 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 07:25:25,941 INFO [train.py:904] (2/8) Epoch 10, batch 8500, loss[loss=0.1936, simple_loss=0.2813, pruned_loss=0.05301, over 16794.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2867, pruned_loss=0.05561, over 3053781.13 frames. ], batch size: 116, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:26:09,291 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 07:26:21,532 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:26:50,575 INFO [train.py:904] (2/8) Epoch 10, batch 8550, loss[loss=0.1864, simple_loss=0.2708, pruned_loss=0.05095, over 12237.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2843, pruned_loss=0.05477, over 3008177.92 frames. ], batch size: 247, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:27:34,852 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:34,917 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:50,548 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:02,240 INFO [optim.py:368] (2/8) 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 07:28:16,656 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:29,553 INFO [train.py:904] (2/8) Epoch 10, batch 8600, loss[loss=0.2002, simple_loss=0.2974, pruned_loss=0.05147, over 16186.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2847, pruned_loss=0.05394, over 3007302.63 frames. ], batch size: 165, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:29:10,994 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:30:11,016 INFO [train.py:904] (2/8) Epoch 10, batch 8650, loss[loss=0.1821, simple_loss=0.2779, pruned_loss=0.04315, over 16859.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2825, pruned_loss=0.05238, over 3000780.69 frames. ], batch size: 116, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:30:24,854 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:30:29,213 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9310, 3.4343, 3.5461, 2.0987, 2.8938, 2.2310, 3.4073, 3.6232], device='cuda:2'), covar=tensor([0.0253, 0.0565, 0.0418, 0.1610, 0.0703, 0.0907, 0.0663, 0.0774], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0134, 0.0150, 0.0136, 0.0130, 0.0120, 0.0131, 0.0142], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 07:31:10,431 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:31:27,975 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:31:31,866 INFO [optim.py:368] (2/8) 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 07:31:56,447 INFO [train.py:904] (2/8) Epoch 10, batch 8700, loss[loss=0.1876, simple_loss=0.2811, pruned_loss=0.04709, over 16678.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2801, pruned_loss=0.05105, over 3030266.25 frames. ], batch size: 134, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:32:27,989 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:32:42,795 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:33:18,063 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5084, 3.3226, 2.7971, 2.0878, 2.1901, 2.2062, 3.5255, 3.1344], device='cuda:2'), covar=tensor([0.2592, 0.0704, 0.1506, 0.2427, 0.2344, 0.1828, 0.0439, 0.1087], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0245, 0.0271, 0.0265, 0.0264, 0.0212, 0.0254, 0.0274], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:33:35,354 INFO [train.py:904] (2/8) Epoch 10, batch 8750, loss[loss=0.2085, simple_loss=0.3013, pruned_loss=0.05787, over 16808.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2796, pruned_loss=0.04983, over 3051438.32 frames. ], batch size: 124, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:34:32,195 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:34:32,342 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:35:02,703 INFO [optim.py:368] (2/8) 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 07:35:29,749 INFO [train.py:904] (2/8) Epoch 10, batch 8800, loss[loss=0.1967, simple_loss=0.2858, pruned_loss=0.05381, over 15501.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2779, pruned_loss=0.04883, over 3048209.80 frames. ], batch size: 192, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:36:11,836 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:36:23,693 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2023-04-29 07:36:33,742 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-29 07:37:15,875 INFO [train.py:904] (2/8) Epoch 10, batch 8850, loss[loss=0.1651, simple_loss=0.2538, pruned_loss=0.03819, over 12456.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.28, pruned_loss=0.04815, over 3035611.21 frames. ], batch size: 248, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:38:03,389 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:24,187 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5930, 3.7476, 2.8002, 2.1729, 2.5044, 2.2483, 3.9651, 3.4465], device='cuda:2'), covar=tensor([0.2519, 0.0615, 0.1496, 0.2012, 0.2057, 0.1778, 0.0352, 0.0907], device='cuda:2'), in_proj_covar=tensor([0.0290, 0.0244, 0.0270, 0.0263, 0.0260, 0.0210, 0.0253, 0.0273], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:38:34,765 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:36,934 INFO [optim.py:368] (2/8) 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 07:38:48,400 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:39:01,669 INFO [train.py:904] (2/8) Epoch 10, batch 8900, loss[loss=0.1873, simple_loss=0.2736, pruned_loss=0.05049, over 12980.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2797, pruned_loss=0.04697, over 3046426.93 frames. ], batch size: 250, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:39:43,460 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:40:45,778 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:01,123 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:05,918 INFO [train.py:904] (2/8) Epoch 10, batch 8950, loss[loss=0.163, simple_loss=0.2574, pruned_loss=0.03431, over 16541.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2789, pruned_loss=0.04711, over 3057223.55 frames. ], batch size: 68, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:42:21,876 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:42:29,137 INFO [optim.py:368] (2/8) 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 07:42:55,053 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3447, 4.4263, 4.2562, 3.9872, 3.8736, 4.3209, 4.0905, 4.0210], device='cuda:2'), covar=tensor([0.0557, 0.0443, 0.0274, 0.0260, 0.0863, 0.0455, 0.0468, 0.0637], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0282, 0.0263, 0.0244, 0.0285, 0.0279, 0.0182, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:42:55,809 INFO [train.py:904] (2/8) Epoch 10, batch 9000, loss[loss=0.1661, simple_loss=0.2576, pruned_loss=0.03729, over 16675.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2756, pruned_loss=0.04551, over 3089196.86 frames. ], batch size: 134, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:42:55,810 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 07:43:05,464 INFO [train.py:938] (2/8) Epoch 10, validation: loss=0.1565, simple_loss=0.2604, pruned_loss=0.02634, over 944034.00 frames. 2023-04-29 07:43:05,464 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 07:43:21,440 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 07:43:24,731 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0521, 3.5841, 3.5416, 2.2615, 3.2411, 3.5752, 3.3955, 1.8299], device='cuda:2'), covar=tensor([0.0436, 0.0026, 0.0034, 0.0332, 0.0064, 0.0054, 0.0047, 0.0411], device='cuda:2'), in_proj_covar=tensor([0.0123, 0.0062, 0.0066, 0.0121, 0.0075, 0.0083, 0.0073, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 07:43:30,276 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:44:12,990 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:44:50,483 INFO [train.py:904] (2/8) Epoch 10, batch 9050, loss[loss=0.1915, simple_loss=0.2711, pruned_loss=0.05599, over 16803.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2769, pruned_loss=0.04644, over 3074553.74 frames. ], batch size: 124, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:44:51,442 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:45:35,729 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3057, 3.3835, 3.6565, 3.6418, 3.6395, 3.4053, 3.4589, 3.4725], device='cuda:2'), covar=tensor([0.0331, 0.0638, 0.0426, 0.0417, 0.0425, 0.0455, 0.0715, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0306, 0.0309, 0.0297, 0.0356, 0.0328, 0.0418, 0.0268], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-29 07:46:02,033 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2031, 4.2098, 4.0587, 3.6021, 4.1261, 1.6491, 3.9482, 3.7997], device='cuda:2'), covar=tensor([0.0083, 0.0080, 0.0139, 0.0204, 0.0073, 0.2369, 0.0109, 0.0167], device='cuda:2'), in_proj_covar=tensor([0.0121, 0.0107, 0.0151, 0.0142, 0.0124, 0.0172, 0.0139, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:46:08,563 INFO [optim.py:368] (2/8) 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 07:46:37,387 INFO [train.py:904] (2/8) Epoch 10, batch 9100, loss[loss=0.1666, simple_loss=0.2553, pruned_loss=0.03891, over 12230.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2765, pruned_loss=0.04689, over 3084743.03 frames. ], batch size: 248, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:46:58,854 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:48:34,506 INFO [train.py:904] (2/8) Epoch 10, batch 9150, loss[loss=0.1799, simple_loss=0.2735, pruned_loss=0.04315, over 17000.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2767, pruned_loss=0.04659, over 3073107.38 frames. ], batch size: 55, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:48:43,197 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-04-29 07:48:52,353 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3664, 3.2265, 3.1633, 3.4278, 3.4540, 3.2724, 3.4282, 3.5121], device='cuda:2'), covar=tensor([0.1074, 0.0920, 0.1594, 0.0879, 0.0911, 0.2280, 0.1137, 0.0986], device='cuda:2'), in_proj_covar=tensor([0.0471, 0.0590, 0.0705, 0.0606, 0.0458, 0.0459, 0.0478, 0.0539], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:49:54,593 INFO [optim.py:368] (2/8) 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 07:50:13,980 INFO [train.py:904] (2/8) Epoch 10, batch 9200, loss[loss=0.1837, simple_loss=0.2627, pruned_loss=0.05231, over 11833.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2724, pruned_loss=0.04567, over 3075347.42 frames. ], batch size: 248, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:50:18,445 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1409, 1.4349, 1.7447, 2.0294, 2.1697, 2.2147, 1.6741, 2.1717], device='cuda:2'), covar=tensor([0.0142, 0.0342, 0.0210, 0.0228, 0.0190, 0.0164, 0.0311, 0.0097], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0165, 0.0148, 0.0149, 0.0160, 0.0113, 0.0166, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 07:51:35,108 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:51:50,960 INFO [train.py:904] (2/8) Epoch 10, batch 9250, loss[loss=0.1832, simple_loss=0.278, pruned_loss=0.0442, over 16268.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2728, pruned_loss=0.0458, over 3082411.71 frames. ], batch size: 165, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:52:57,518 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5449, 4.8208, 4.6008, 4.6136, 4.2850, 4.3653, 4.3131, 4.8694], device='cuda:2'), covar=tensor([0.0913, 0.0803, 0.0928, 0.0621, 0.0820, 0.1101, 0.0984, 0.0803], device='cuda:2'), in_proj_covar=tensor([0.0484, 0.0613, 0.0504, 0.0425, 0.0384, 0.0407, 0.0514, 0.0469], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:53:14,010 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 10, batch 9300, loss[loss=0.1751, simple_loss=0.2606, pruned_loss=0.04475, over 16178.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2707, pruned_loss=0.04504, over 3076136.01 frames. ], batch size: 165, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:54:09,183 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:54:28,321 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3800, 1.9951, 2.2608, 3.9451, 1.9719, 2.4221, 2.2054, 2.1733], device='cuda:2'), covar=tensor([0.0799, 0.3404, 0.2121, 0.0388, 0.3781, 0.2257, 0.2826, 0.3263], device='cuda:2'), in_proj_covar=tensor([0.0339, 0.0371, 0.0312, 0.0306, 0.0399, 0.0414, 0.0333, 0.0430], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:55:29,731 INFO [train.py:904] (2/8) Epoch 10, batch 9350, loss[loss=0.1898, simple_loss=0.2803, pruned_loss=0.04968, over 16418.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2707, pruned_loss=0.04494, over 3093789.00 frames. ], batch size: 146, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:55:50,100 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:56:28,465 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:56:31,908 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4759, 3.3821, 2.7857, 2.0673, 2.2038, 2.2322, 3.5204, 3.1934], device='cuda:2'), covar=tensor([0.2585, 0.0734, 0.1517, 0.2331, 0.2357, 0.1789, 0.0409, 0.1012], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0246, 0.0274, 0.0266, 0.0259, 0.0213, 0.0256, 0.0275], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:56:48,376 INFO [optim.py:368] (2/8) 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 07:57:12,447 INFO [train.py:904] (2/8) Epoch 10, batch 9400, loss[loss=0.1997, simple_loss=0.302, pruned_loss=0.04865, over 16281.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2716, pruned_loss=0.04476, over 3094083.40 frames. ], batch size: 146, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:57:16,013 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 07:57:25,167 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:57:32,822 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 07:57:38,097 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3609, 2.9311, 2.6062, 2.1611, 2.1443, 2.1888, 2.9185, 2.8733], device='cuda:2'), covar=tensor([0.2153, 0.0855, 0.1351, 0.2277, 0.2072, 0.1751, 0.0460, 0.1080], device='cuda:2'), in_proj_covar=tensor([0.0290, 0.0244, 0.0271, 0.0264, 0.0256, 0.0211, 0.0254, 0.0273], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 07:58:33,075 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:58:52,903 INFO [train.py:904] (2/8) Epoch 10, batch 9450, loss[loss=0.1799, simple_loss=0.2653, pruned_loss=0.04728, over 12652.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2734, pruned_loss=0.04515, over 3072606.41 frames. ], batch size: 248, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:59:35,306 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3008, 3.2329, 3.1922, 3.4443, 3.4279, 3.2495, 3.4387, 3.5090], device='cuda:2'), covar=tensor([0.1293, 0.0982, 0.1522, 0.0830, 0.0918, 0.2517, 0.1137, 0.0999], device='cuda:2'), in_proj_covar=tensor([0.0470, 0.0588, 0.0708, 0.0607, 0.0459, 0.0460, 0.0476, 0.0539], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:00:10,871 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 10, batch 9500, loss[loss=0.1563, simple_loss=0.2553, pruned_loss=0.02859, over 16945.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2728, pruned_loss=0.04486, over 3076560.27 frames. ], batch size: 96, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:01:07,301 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 08:02:03,023 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 10, batch 9550, loss[loss=0.1976, simple_loss=0.294, pruned_loss=0.05063, over 16674.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2726, pruned_loss=0.04466, over 3099311.06 frames. ], batch size: 134, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:03:16,116 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3084, 2.6629, 2.6163, 5.0727, 2.4623, 3.0761, 2.7386, 2.9254], device='cuda:2'), covar=tensor([0.0632, 0.3066, 0.1995, 0.0227, 0.3573, 0.1955, 0.2642, 0.2943], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0370, 0.0313, 0.0305, 0.0400, 0.0412, 0.0334, 0.0430], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:03:40,249 INFO [optim.py:368] (2/8) 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 08:03:45,436 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:04:02,626 INFO [train.py:904] (2/8) Epoch 10, batch 9600, loss[loss=0.1945, simple_loss=0.2897, pruned_loss=0.04961, over 16929.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2738, pruned_loss=0.04543, over 3082558.68 frames. ], batch size: 109, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:05:52,878 INFO [train.py:904] (2/8) Epoch 10, batch 9650, loss[loss=0.1838, simple_loss=0.2803, pruned_loss=0.04365, over 16871.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2759, pruned_loss=0.04568, over 3082139.61 frames. ], batch size: 102, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:05:54,282 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7055, 4.6673, 5.1135, 5.1259, 5.0869, 4.8457, 4.7318, 4.5286], device='cuda:2'), covar=tensor([0.0244, 0.0414, 0.0286, 0.0306, 0.0429, 0.0263, 0.0820, 0.0363], device='cuda:2'), in_proj_covar=tensor([0.0295, 0.0303, 0.0306, 0.0292, 0.0351, 0.0322, 0.0410, 0.0264], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-29 08:06:14,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4897, 5.8042, 5.4962, 5.5783, 5.2008, 5.2277, 5.2318, 5.8853], device='cuda:2'), covar=tensor([0.0830, 0.0772, 0.1044, 0.0633, 0.0751, 0.0617, 0.0911, 0.0800], device='cuda:2'), in_proj_covar=tensor([0.0485, 0.0605, 0.0499, 0.0422, 0.0383, 0.0401, 0.0508, 0.0464], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:07:15,509 INFO [optim.py:368] (2/8) 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 08:07:32,660 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2275, 1.9449, 2.0907, 3.7626, 1.9504, 2.3074, 2.0627, 2.1460], device='cuda:2'), covar=tensor([0.0920, 0.3467, 0.2246, 0.0406, 0.3887, 0.2161, 0.3003, 0.3215], device='cuda:2'), in_proj_covar=tensor([0.0343, 0.0371, 0.0315, 0.0305, 0.0401, 0.0414, 0.0336, 0.0432], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:07:41,458 INFO [train.py:904] (2/8) Epoch 10, batch 9700, loss[loss=0.1943, simple_loss=0.2823, pruned_loss=0.0531, over 16917.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2749, pruned_loss=0.04566, over 3083763.51 frames. ], batch size: 116, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:07:52,614 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:08:17,292 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8247, 1.2321, 1.5917, 1.6976, 1.7474, 1.8409, 1.5614, 1.7516], device='cuda:2'), covar=tensor([0.0201, 0.0303, 0.0158, 0.0191, 0.0219, 0.0165, 0.0318, 0.0098], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0165, 0.0148, 0.0149, 0.0161, 0.0113, 0.0166, 0.0105], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 08:08:54,360 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 08:09:22,977 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 08:09:25,204 INFO [train.py:904] (2/8) Epoch 10, batch 9750, loss[loss=0.1855, simple_loss=0.278, pruned_loss=0.04649, over 15264.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.274, pruned_loss=0.04589, over 3086549.88 frames. ], batch size: 191, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:09:32,958 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:09:34,820 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:10:35,772 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 08:10:45,034 INFO [optim.py:368] (2/8) 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 08:11:05,030 INFO [train.py:904] (2/8) Epoch 10, batch 9800, loss[loss=0.1678, simple_loss=0.2508, pruned_loss=0.0424, over 12054.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2738, pruned_loss=0.04529, over 3080485.67 frames. ], batch size: 250, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:11:34,931 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-29 08:11:36,093 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:12:49,193 INFO [train.py:904] (2/8) Epoch 10, batch 9850, loss[loss=0.1989, simple_loss=0.2888, pruned_loss=0.05451, over 16676.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.275, pruned_loss=0.04499, over 3084347.47 frames. ], batch size: 62, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:13:00,247 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 08:13:24,207 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2451, 3.3437, 3.6011, 3.5725, 3.5909, 3.3888, 3.4117, 3.4647], device='cuda:2'), covar=tensor([0.0367, 0.0685, 0.0425, 0.0415, 0.0485, 0.0505, 0.0776, 0.0444], device='cuda:2'), in_proj_covar=tensor([0.0295, 0.0303, 0.0306, 0.0292, 0.0351, 0.0323, 0.0409, 0.0263], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-29 08:14:17,926 INFO [optim.py:368] (2/8) 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 08:14:41,618 INFO [train.py:904] (2/8) Epoch 10, batch 9900, loss[loss=0.1862, simple_loss=0.2744, pruned_loss=0.04897, over 16321.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2748, pruned_loss=0.04503, over 3065545.94 frames. ], batch size: 35, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:16:40,042 INFO [train.py:904] (2/8) Epoch 10, batch 9950, loss[loss=0.1579, simple_loss=0.2602, pruned_loss=0.02781, over 16554.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2758, pruned_loss=0.04504, over 3041082.61 frames. ], batch size: 68, lr: 6.68e-03, grad_scale: 4.0 2023-04-29 08:18:13,670 INFO [optim.py:368] (2/8) 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 08:18:42,224 INFO [train.py:904] (2/8) Epoch 10, batch 10000, loss[loss=0.1818, simple_loss=0.2823, pruned_loss=0.04063, over 15463.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2745, pruned_loss=0.04444, over 3066393.31 frames. ], batch size: 191, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:18:42,871 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0445, 4.0327, 3.9357, 3.4848, 3.9688, 1.6534, 3.7867, 3.6600], device='cuda:2'), covar=tensor([0.0077, 0.0077, 0.0110, 0.0180, 0.0072, 0.2364, 0.0091, 0.0179], device='cuda:2'), in_proj_covar=tensor([0.0122, 0.0108, 0.0151, 0.0138, 0.0124, 0.0173, 0.0138, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:19:34,657 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9677, 2.0086, 2.2551, 3.2055, 2.1053, 2.2047, 2.1682, 2.0842], device='cuda:2'), covar=tensor([0.0813, 0.3129, 0.1877, 0.0459, 0.3479, 0.2181, 0.2943, 0.3107], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0367, 0.0312, 0.0302, 0.0396, 0.0411, 0.0333, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:19:54,018 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:20:23,593 INFO [train.py:904] (2/8) Epoch 10, batch 10050, loss[loss=0.1971, simple_loss=0.2857, pruned_loss=0.05419, over 12096.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2744, pruned_loss=0.04423, over 3066116.69 frames. ], batch size: 247, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:21:09,849 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3576, 2.3480, 1.9465, 2.2059, 2.6859, 2.4015, 3.0676, 3.0028], device='cuda:2'), covar=tensor([0.0081, 0.0319, 0.0426, 0.0331, 0.0228, 0.0314, 0.0178, 0.0189], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0195, 0.0193, 0.0190, 0.0191, 0.0194, 0.0188, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:21:25,328 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:21:36,390 INFO [optim.py:368] (2/8) 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 08:21:56,134 INFO [train.py:904] (2/8) Epoch 10, batch 10100, loss[loss=0.1871, simple_loss=0.2767, pruned_loss=0.04874, over 16375.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2756, pruned_loss=0.04482, over 3066324.95 frames. ], batch size: 146, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:22:16,047 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:23:38,425 INFO [train.py:904] (2/8) Epoch 11, batch 0, loss[loss=0.2728, simple_loss=0.3404, pruned_loss=0.1027, over 16517.00 frames. ], tot_loss[loss=0.2728, simple_loss=0.3404, pruned_loss=0.1027, over 16517.00 frames. ], batch size: 68, lr: 6.37e-03, grad_scale: 8.0 2023-04-29 08:23:38,426 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 08:23:45,823 INFO [train.py:938] (2/8) Epoch 11, validation: loss=0.1557, simple_loss=0.2595, pruned_loss=0.02591, over 944034.00 frames. 2023-04-29 08:23:45,824 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 08:24:43,127 INFO [optim.py:368] (2/8) 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 08:24:55,123 INFO [train.py:904] (2/8) Epoch 11, batch 50, loss[loss=0.1941, simple_loss=0.2875, pruned_loss=0.05035, over 17020.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2887, pruned_loss=0.06477, over 750410.75 frames. ], batch size: 55, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:25:12,393 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0660, 4.5433, 3.3624, 2.4465, 2.9075, 2.4928, 4.9044, 3.8204], device='cuda:2'), covar=tensor([0.2225, 0.0557, 0.1351, 0.2063, 0.2410, 0.1767, 0.0298, 0.1000], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0245, 0.0274, 0.0265, 0.0256, 0.0212, 0.0256, 0.0277], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:25:36,824 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6061, 3.7273, 1.9735, 3.8763, 2.6170, 3.8703, 2.1827, 2.8908], device='cuda:2'), covar=tensor([0.0241, 0.0326, 0.1608, 0.0258, 0.0826, 0.0481, 0.1354, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0156, 0.0183, 0.0119, 0.0163, 0.0194, 0.0192, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 08:25:42,915 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 08:26:05,598 INFO [train.py:904] (2/8) Epoch 11, batch 100, loss[loss=0.1748, simple_loss=0.26, pruned_loss=0.04482, over 17224.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2822, pruned_loss=0.05914, over 1332589.51 frames. ], batch size: 44, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:26:36,221 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0981, 5.0151, 4.9259, 4.5110, 4.4858, 4.9076, 4.9960, 4.5800], device='cuda:2'), covar=tensor([0.0539, 0.0427, 0.0260, 0.0307, 0.1078, 0.0413, 0.0230, 0.0725], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0289, 0.0268, 0.0249, 0.0290, 0.0284, 0.0185, 0.0316], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:27:03,352 INFO [optim.py:368] (2/8) 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 08:27:12,713 INFO [train.py:904] (2/8) Epoch 11, batch 150, loss[loss=0.221, simple_loss=0.2894, pruned_loss=0.07631, over 16854.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.281, pruned_loss=0.05802, over 1761268.81 frames. ], batch size: 116, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:27:58,586 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:28:23,297 INFO [train.py:904] (2/8) Epoch 11, batch 200, loss[loss=0.1981, simple_loss=0.291, pruned_loss=0.05267, over 16675.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2821, pruned_loss=0.05924, over 2105868.45 frames. ], batch size: 57, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:21,756 INFO [optim.py:368] (2/8) 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 08:29:23,306 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:29:31,793 INFO [train.py:904] (2/8) Epoch 11, batch 250, loss[loss=0.2061, simple_loss=0.2942, pruned_loss=0.05901, over 17058.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2805, pruned_loss=0.05859, over 2378505.88 frames. ], batch size: 50, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:46,204 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:30:37,960 INFO [train.py:904] (2/8) Epoch 11, batch 300, loss[loss=0.1877, simple_loss=0.2763, pruned_loss=0.0496, over 16745.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2774, pruned_loss=0.05706, over 2572506.50 frames. ], batch size: 62, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:30:51,058 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:31:35,542 INFO [optim.py:368] (2/8) 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 08:31:47,602 INFO [train.py:904] (2/8) Epoch 11, batch 350, loss[loss=0.1737, simple_loss=0.2629, pruned_loss=0.04228, over 17129.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.275, pruned_loss=0.05642, over 2740323.21 frames. ], batch size: 49, lr: 6.36e-03, grad_scale: 1.0 2023-04-29 08:32:01,966 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0058, 1.8165, 2.4531, 2.8151, 2.6687, 3.1313, 2.1537, 3.0748], device='cuda:2'), covar=tensor([0.0165, 0.0328, 0.0214, 0.0201, 0.0222, 0.0148, 0.0312, 0.0095], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0170, 0.0154, 0.0156, 0.0168, 0.0120, 0.0171, 0.0109], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 08:32:42,838 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5439, 4.5153, 4.9323, 4.9532, 5.0070, 4.6261, 4.6030, 4.3896], device='cuda:2'), covar=tensor([0.0323, 0.0725, 0.0423, 0.0399, 0.0419, 0.0396, 0.0913, 0.0503], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0328, 0.0332, 0.0311, 0.0373, 0.0346, 0.0444, 0.0281], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 08:32:56,331 INFO [train.py:904] (2/8) Epoch 11, batch 400, loss[loss=0.2128, simple_loss=0.2802, pruned_loss=0.07271, over 16749.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2731, pruned_loss=0.0559, over 2864788.03 frames. ], batch size: 124, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:32:56,821 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6137, 2.5657, 1.9180, 2.3523, 2.9657, 2.6573, 3.4302, 3.2427], device='cuda:2'), covar=tensor([0.0071, 0.0276, 0.0410, 0.0316, 0.0175, 0.0277, 0.0156, 0.0162], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0201, 0.0199, 0.0197, 0.0200, 0.0202, 0.0201, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:33:22,130 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:33:54,607 INFO [optim.py:368] (2/8) 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 08:33:56,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2430, 3.7076, 3.8282, 1.9685, 3.1528, 2.3170, 3.7149, 3.7239], device='cuda:2'), covar=tensor([0.0264, 0.0654, 0.0453, 0.1709, 0.0699, 0.0959, 0.0568, 0.0912], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0138, 0.0157, 0.0142, 0.0135, 0.0125, 0.0135, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 08:34:05,294 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 08:34:06,160 INFO [train.py:904] (2/8) Epoch 11, batch 450, loss[loss=0.1417, simple_loss=0.2208, pruned_loss=0.03135, over 16991.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2713, pruned_loss=0.05442, over 2975127.25 frames. ], batch size: 41, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:34:46,986 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4245, 3.4851, 1.7291, 3.6156, 2.5393, 3.6861, 1.9491, 2.7838], device='cuda:2'), covar=tensor([0.0225, 0.0314, 0.1626, 0.0260, 0.0742, 0.0521, 0.1463, 0.0574], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0163, 0.0188, 0.0127, 0.0168, 0.0204, 0.0197, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 08:34:47,003 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 08:34:54,619 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:35:19,214 INFO [train.py:904] (2/8) Epoch 11, batch 500, loss[loss=0.164, simple_loss=0.2602, pruned_loss=0.03389, over 17104.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2688, pruned_loss=0.05307, over 3056855.66 frames. ], batch size: 48, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:35:36,855 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8608, 2.6895, 2.5672, 2.0082, 2.6145, 2.6613, 2.5988, 1.8990], device='cuda:2'), covar=tensor([0.0338, 0.0068, 0.0049, 0.0283, 0.0083, 0.0078, 0.0079, 0.0308], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0068, 0.0069, 0.0125, 0.0077, 0.0086, 0.0076, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 08:36:10,824 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 08:36:13,475 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:19,094 INFO [optim.py:368] (2/8) 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 08:36:24,797 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:29,128 INFO [train.py:904] (2/8) Epoch 11, batch 550, loss[loss=0.1823, simple_loss=0.2669, pruned_loss=0.04881, over 17211.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2674, pruned_loss=0.05251, over 3115258.27 frames. ], batch size: 45, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:37:40,172 INFO [train.py:904] (2/8) Epoch 11, batch 600, loss[loss=0.1881, simple_loss=0.2521, pruned_loss=0.06206, over 16865.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2662, pruned_loss=0.05308, over 3168448.63 frames. ], batch size: 90, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:37:43,152 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5655, 3.4709, 2.7831, 2.1184, 2.5479, 2.2735, 3.6197, 3.3158], device='cuda:2'), covar=tensor([0.2464, 0.0763, 0.1389, 0.2355, 0.2169, 0.1766, 0.0525, 0.1326], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0253, 0.0281, 0.0273, 0.0273, 0.0221, 0.0266, 0.0291], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:38:38,908 INFO [optim.py:368] (2/8) 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 08:38:49,712 INFO [train.py:904] (2/8) Epoch 11, batch 650, loss[loss=0.1533, simple_loss=0.2401, pruned_loss=0.03322, over 17212.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2654, pruned_loss=0.05239, over 3209325.53 frames. ], batch size: 46, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:39:48,008 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 08:39:54,770 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8473, 4.2921, 4.3591, 3.3401, 3.7390, 4.3360, 3.9488, 2.5385], device='cuda:2'), covar=tensor([0.0347, 0.0040, 0.0033, 0.0224, 0.0066, 0.0066, 0.0062, 0.0345], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0070, 0.0070, 0.0127, 0.0078, 0.0088, 0.0078, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 08:39:58,899 INFO [train.py:904] (2/8) Epoch 11, batch 700, loss[loss=0.1713, simple_loss=0.2491, pruned_loss=0.04674, over 16741.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2649, pruned_loss=0.05166, over 3242827.97 frames. ], batch size: 83, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:40:44,242 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6765, 2.5148, 2.2062, 2.5324, 2.9235, 2.8052, 3.4751, 3.1856], device='cuda:2'), covar=tensor([0.0072, 0.0322, 0.0389, 0.0324, 0.0198, 0.0273, 0.0165, 0.0183], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0206, 0.0202, 0.0201, 0.0204, 0.0205, 0.0208, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:40:57,196 INFO [optim.py:368] (2/8) 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 08:41:09,230 INFO [train.py:904] (2/8) Epoch 11, batch 750, loss[loss=0.2024, simple_loss=0.268, pruned_loss=0.06837, over 16784.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2648, pruned_loss=0.05208, over 3258815.70 frames. ], batch size: 124, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:41:42,360 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:42:18,047 INFO [train.py:904] (2/8) Epoch 11, batch 800, loss[loss=0.174, simple_loss=0.2687, pruned_loss=0.03964, over 17268.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2651, pruned_loss=0.05193, over 3278081.88 frames. ], batch size: 52, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:43:11,908 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:15,304 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:16,112 INFO [optim.py:368] (2/8) 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 08:43:26,679 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-29 08:43:27,548 INFO [train.py:904] (2/8) Epoch 11, batch 850, loss[loss=0.1618, simple_loss=0.2479, pruned_loss=0.03789, over 17214.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2646, pruned_loss=0.05116, over 3286626.41 frames. ], batch size: 44, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:44:17,574 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:44:37,482 INFO [train.py:904] (2/8) Epoch 11, batch 900, loss[loss=0.1942, simple_loss=0.2954, pruned_loss=0.0465, over 17051.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2644, pruned_loss=0.05086, over 3296576.72 frames. ], batch size: 53, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:45:35,217 INFO [optim.py:368] (2/8) 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 08:45:45,386 INFO [train.py:904] (2/8) Epoch 11, batch 950, loss[loss=0.1766, simple_loss=0.2695, pruned_loss=0.04182, over 17286.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.264, pruned_loss=0.05064, over 3295840.92 frames. ], batch size: 52, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:46:29,870 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 08:46:54,282 INFO [train.py:904] (2/8) Epoch 11, batch 1000, loss[loss=0.1629, simple_loss=0.2575, pruned_loss=0.03411, over 17061.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.262, pruned_loss=0.04973, over 3305119.93 frames. ], batch size: 50, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:47:52,028 INFO [optim.py:368] (2/8) 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 08:48:02,250 INFO [train.py:904] (2/8) Epoch 11, batch 1050, loss[loss=0.178, simple_loss=0.2655, pruned_loss=0.04522, over 17247.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2626, pruned_loss=0.05026, over 3310807.06 frames. ], batch size: 45, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:48:36,348 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:48:38,742 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3153, 3.5809, 3.0380, 1.9497, 2.6191, 2.1231, 3.6835, 3.9068], device='cuda:2'), covar=tensor([0.0209, 0.0559, 0.0665, 0.1918, 0.0922, 0.1142, 0.0483, 0.0655], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0142, 0.0157, 0.0143, 0.0136, 0.0125, 0.0137, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 08:48:49,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8876, 4.8886, 5.4649, 5.3909, 5.3764, 5.0311, 5.0008, 4.6918], device='cuda:2'), covar=tensor([0.0317, 0.0482, 0.0339, 0.0496, 0.0479, 0.0349, 0.0865, 0.0421], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0348, 0.0350, 0.0329, 0.0397, 0.0366, 0.0465, 0.0296], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 08:49:12,619 INFO [train.py:904] (2/8) Epoch 11, batch 1100, loss[loss=0.177, simple_loss=0.2734, pruned_loss=0.04032, over 17267.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2621, pruned_loss=0.049, over 3318063.89 frames. ], batch size: 52, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:49:20,413 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2182, 4.0487, 4.2839, 4.4507, 4.5525, 4.1011, 4.3359, 4.4921], device='cuda:2'), covar=tensor([0.1251, 0.0913, 0.1227, 0.0553, 0.0435, 0.1218, 0.1501, 0.0499], device='cuda:2'), in_proj_covar=tensor([0.0533, 0.0665, 0.0821, 0.0687, 0.0516, 0.0517, 0.0531, 0.0605], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:49:43,773 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:50:07,508 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:50:09,059 INFO [optim.py:368] (2/8) 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 08:50:16,212 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 08:50:20,435 INFO [train.py:904] (2/8) Epoch 11, batch 1150, loss[loss=0.1972, simple_loss=0.2959, pruned_loss=0.04924, over 17152.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2629, pruned_loss=0.04857, over 3323863.56 frames. ], batch size: 48, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:50:24,457 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9891, 4.0375, 4.3595, 2.0405, 4.5443, 4.6077, 3.2341, 3.5321], device='cuda:2'), covar=tensor([0.0600, 0.0172, 0.0210, 0.1128, 0.0054, 0.0110, 0.0357, 0.0357], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0096, 0.0087, 0.0135, 0.0068, 0.0101, 0.0119, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 08:50:42,144 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0419, 5.6151, 5.8177, 5.5987, 5.5781, 6.1510, 5.7825, 5.4596], device='cuda:2'), covar=tensor([0.0868, 0.1647, 0.1699, 0.1736, 0.2724, 0.0889, 0.1158, 0.2370], device='cuda:2'), in_proj_covar=tensor([0.0348, 0.0490, 0.0530, 0.0423, 0.0565, 0.0556, 0.0423, 0.0575], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 08:50:58,836 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 08:51:14,262 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:51:18,695 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 08:51:27,908 INFO [train.py:904] (2/8) Epoch 11, batch 1200, loss[loss=0.1854, simple_loss=0.2808, pruned_loss=0.04503, over 17135.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2617, pruned_loss=0.04827, over 3320601.06 frames. ], batch size: 48, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:52:27,630 INFO [optim.py:368] (2/8) 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 08:52:39,081 INFO [train.py:904] (2/8) Epoch 11, batch 1250, loss[loss=0.1785, simple_loss=0.2727, pruned_loss=0.04213, over 17105.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2616, pruned_loss=0.04914, over 3317578.93 frames. ], batch size: 48, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:53:49,400 INFO [train.py:904] (2/8) Epoch 11, batch 1300, loss[loss=0.159, simple_loss=0.2475, pruned_loss=0.03523, over 16827.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2604, pruned_loss=0.04945, over 3319363.00 frames. ], batch size: 42, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:54:01,887 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0700, 5.7535, 5.8530, 5.6139, 5.5819, 6.1236, 5.7694, 5.4425], device='cuda:2'), covar=tensor([0.0871, 0.1615, 0.1786, 0.1702, 0.2840, 0.1021, 0.1242, 0.2170], device='cuda:2'), in_proj_covar=tensor([0.0346, 0.0488, 0.0531, 0.0422, 0.0565, 0.0555, 0.0422, 0.0575], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 08:54:38,317 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1931, 2.0257, 2.2681, 3.7072, 2.0912, 2.3666, 2.1688, 2.2074], device='cuda:2'), covar=tensor([0.1062, 0.3195, 0.2114, 0.0526, 0.3377, 0.2342, 0.3052, 0.2967], device='cuda:2'), in_proj_covar=tensor([0.0361, 0.0388, 0.0328, 0.0324, 0.0408, 0.0443, 0.0352, 0.0459], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:54:46,485 INFO [optim.py:368] (2/8) 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 08:54:57,207 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:54:58,763 INFO [train.py:904] (2/8) Epoch 11, batch 1350, loss[loss=0.1935, simple_loss=0.2831, pruned_loss=0.05192, over 16723.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2605, pruned_loss=0.04967, over 3310576.61 frames. ], batch size: 57, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:55:45,154 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:56:05,623 INFO [train.py:904] (2/8) Epoch 11, batch 1400, loss[loss=0.1795, simple_loss=0.2749, pruned_loss=0.04205, over 17042.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2605, pruned_loss=0.04938, over 3311280.19 frames. ], batch size: 50, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:56:20,008 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 08:57:10,184 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:57:15,710 INFO [train.py:904] (2/8) Epoch 11, batch 1450, loss[loss=0.1877, simple_loss=0.2512, pruned_loss=0.0621, over 16901.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2599, pruned_loss=0.04936, over 3309032.66 frames. ], batch size: 109, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:02,089 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0918, 5.0468, 4.7900, 3.5450, 4.8723, 1.6434, 4.5798, 4.6578], device='cuda:2'), covar=tensor([0.0128, 0.0098, 0.0205, 0.0699, 0.0115, 0.3358, 0.0167, 0.0329], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0119, 0.0169, 0.0157, 0.0137, 0.0182, 0.0155, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 08:58:16,888 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-29 08:58:25,000 INFO [train.py:904] (2/8) Epoch 11, batch 1500, loss[loss=0.2298, simple_loss=0.2922, pruned_loss=0.08373, over 16768.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2592, pruned_loss=0.04958, over 3304066.10 frames. ], batch size: 124, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:52,460 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9301, 4.1182, 4.4557, 2.1649, 4.6646, 4.6879, 3.3846, 3.6121], device='cuda:2'), covar=tensor([0.0653, 0.0146, 0.0154, 0.1021, 0.0044, 0.0097, 0.0312, 0.0326], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0098, 0.0089, 0.0138, 0.0070, 0.0103, 0.0122, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 08:59:24,563 INFO [optim.py:368] (2/8) 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 08:59:35,064 INFO [train.py:904] (2/8) Epoch 11, batch 1550, loss[loss=0.1545, simple_loss=0.2399, pruned_loss=0.03449, over 17198.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2606, pruned_loss=0.05058, over 3304343.13 frames. ], batch size: 44, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 09:00:27,149 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0937, 5.4306, 5.2066, 5.2339, 4.8910, 4.8397, 4.8747, 5.5453], device='cuda:2'), covar=tensor([0.1052, 0.0786, 0.0851, 0.0648, 0.0728, 0.0793, 0.0957, 0.0705], device='cuda:2'), in_proj_covar=tensor([0.0547, 0.0685, 0.0565, 0.0478, 0.0429, 0.0441, 0.0573, 0.0523], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:00:39,946 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:00:43,355 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 09:00:45,009 INFO [train.py:904] (2/8) Epoch 11, batch 1600, loss[loss=0.1949, simple_loss=0.2812, pruned_loss=0.05429, over 16749.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2631, pruned_loss=0.05148, over 3303122.06 frames. ], batch size: 57, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:01:00,320 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-29 09:01:26,034 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:01:42,624 INFO [optim.py:368] (2/8) 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 09:01:53,523 INFO [train.py:904] (2/8) Epoch 11, batch 1650, loss[loss=0.1748, simple_loss=0.2672, pruned_loss=0.04121, over 17260.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2646, pruned_loss=0.05227, over 3306202.30 frames. ], batch size: 52, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:02:03,297 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:02:50,068 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:02:52,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9674, 2.3324, 2.3927, 4.8018, 2.2909, 2.9014, 2.3758, 2.6035], device='cuda:2'), covar=tensor([0.0826, 0.3239, 0.2240, 0.0314, 0.3790, 0.2081, 0.3026, 0.3359], device='cuda:2'), in_proj_covar=tensor([0.0366, 0.0392, 0.0331, 0.0325, 0.0412, 0.0447, 0.0355, 0.0463], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:03:02,605 INFO [train.py:904] (2/8) Epoch 11, batch 1700, loss[loss=0.1944, simple_loss=0.2719, pruned_loss=0.05846, over 16688.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2669, pruned_loss=0.05249, over 3306459.12 frames. ], batch size: 134, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:03:10,659 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:58,996 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 09:04:01,647 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:04:02,559 INFO [optim.py:368] (2/8) 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 09:04:13,825 INFO [train.py:904] (2/8) Epoch 11, batch 1750, loss[loss=0.2273, simple_loss=0.3015, pruned_loss=0.07657, over 16332.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2678, pruned_loss=0.05289, over 3301561.65 frames. ], batch size: 165, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:04:33,926 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7421, 3.8134, 2.0729, 4.0897, 2.7129, 4.0243, 2.2009, 2.8553], device='cuda:2'), covar=tensor([0.0200, 0.0303, 0.1489, 0.0199, 0.0786, 0.0527, 0.1337, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0167, 0.0187, 0.0133, 0.0169, 0.0210, 0.0195, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 09:04:50,426 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-04-29 09:05:22,484 INFO [train.py:904] (2/8) Epoch 11, batch 1800, loss[loss=0.2138, simple_loss=0.284, pruned_loss=0.0718, over 16737.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2689, pruned_loss=0.05314, over 3297263.24 frames. ], batch size: 124, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:21,365 INFO [optim.py:368] (2/8) 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 09:06:31,957 INFO [train.py:904] (2/8) Epoch 11, batch 1850, loss[loss=0.1649, simple_loss=0.2561, pruned_loss=0.03687, over 17197.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2689, pruned_loss=0.05227, over 3297235.86 frames. ], batch size: 46, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:34,744 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8139, 5.1904, 4.9882, 4.9799, 4.6364, 4.5869, 4.6032, 5.3142], device='cuda:2'), covar=tensor([0.1147, 0.0854, 0.0897, 0.0690, 0.0743, 0.0886, 0.0990, 0.0742], device='cuda:2'), in_proj_covar=tensor([0.0546, 0.0685, 0.0566, 0.0479, 0.0428, 0.0444, 0.0574, 0.0521], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:06:50,859 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:07:39,145 INFO [train.py:904] (2/8) Epoch 11, batch 1900, loss[loss=0.186, simple_loss=0.2731, pruned_loss=0.04949, over 17036.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2686, pruned_loss=0.05155, over 3300101.38 frames. ], batch size: 50, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:07:45,038 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4045, 2.1599, 1.6202, 1.7862, 2.5321, 2.3217, 2.7664, 2.6850], device='cuda:2'), covar=tensor([0.0165, 0.0362, 0.0518, 0.0460, 0.0214, 0.0329, 0.0222, 0.0253], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0205, 0.0201, 0.0202, 0.0205, 0.0205, 0.0212, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:08:16,174 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:08:40,923 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 11, batch 1950, loss[loss=0.1554, simple_loss=0.2507, pruned_loss=0.03004, over 17234.00 frames. ], tot_loss[loss=0.186, simple_loss=0.269, pruned_loss=0.05153, over 3309809.44 frames. ], batch size: 45, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:08:53,827 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 09:08:55,092 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:09:19,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9707, 4.9682, 5.5735, 5.5211, 5.5057, 5.1507, 5.1139, 4.8330], device='cuda:2'), covar=tensor([0.0295, 0.0404, 0.0269, 0.0383, 0.0430, 0.0289, 0.0855, 0.0388], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0352, 0.0353, 0.0331, 0.0401, 0.0370, 0.0471, 0.0300], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 09:09:41,369 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:00,144 INFO [train.py:904] (2/8) Epoch 11, batch 2000, loss[loss=0.2126, simple_loss=0.2961, pruned_loss=0.06456, over 12240.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2688, pruned_loss=0.05135, over 3305922.73 frames. ], batch size: 247, lr: 6.31e-03, grad_scale: 8.0 2023-04-29 09:10:07,778 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:18,215 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8048, 3.7922, 2.9375, 2.2409, 2.5098, 2.2613, 3.8409, 3.3701], device='cuda:2'), covar=tensor([0.2181, 0.0554, 0.1294, 0.2419, 0.2491, 0.1835, 0.0430, 0.1199], device='cuda:2'), in_proj_covar=tensor([0.0303, 0.0256, 0.0282, 0.0275, 0.0280, 0.0222, 0.0267, 0.0298], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:10:58,864 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:11:00,877 INFO [optim.py:368] (2/8) 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 09:11:11,326 INFO [train.py:904] (2/8) Epoch 11, batch 2050, loss[loss=0.1997, simple_loss=0.2664, pruned_loss=0.06647, over 16873.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2679, pruned_loss=0.05137, over 3309413.53 frames. ], batch size: 96, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:11:16,434 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:05,786 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:21,600 INFO [train.py:904] (2/8) Epoch 11, batch 2100, loss[loss=0.1656, simple_loss=0.2465, pruned_loss=0.04232, over 16822.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2695, pruned_loss=0.05275, over 3297466.20 frames. ], batch size: 39, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:13:22,846 INFO [optim.py:368] (2/8) 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 09:13:31,355 INFO [train.py:904] (2/8) Epoch 11, batch 2150, loss[loss=0.1651, simple_loss=0.2579, pruned_loss=0.03615, over 17041.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2709, pruned_loss=0.05317, over 3304021.77 frames. ], batch size: 55, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:14:16,063 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 09:14:42,107 INFO [train.py:904] (2/8) Epoch 11, batch 2200, loss[loss=0.1928, simple_loss=0.2639, pruned_loss=0.06085, over 16653.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2707, pruned_loss=0.05303, over 3317239.40 frames. ], batch size: 134, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:10,362 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:15:35,178 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:15:44,096 INFO [optim.py:368] (2/8) 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 09:15:51,099 INFO [train.py:904] (2/8) Epoch 11, batch 2250, loss[loss=0.1944, simple_loss=0.2795, pruned_loss=0.05469, over 17245.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2703, pruned_loss=0.05246, over 3323437.76 frames. ], batch size: 45, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:54,408 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:03,134 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-29 09:16:12,348 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:44,179 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:52,878 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0221, 3.9831, 3.8820, 3.1675, 3.9279, 1.7690, 3.7027, 3.3028], device='cuda:2'), covar=tensor([0.0107, 0.0092, 0.0176, 0.0301, 0.0083, 0.2668, 0.0118, 0.0264], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0119, 0.0168, 0.0157, 0.0137, 0.0179, 0.0154, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:16:54,319 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 09:17:01,699 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:02,364 INFO [train.py:904] (2/8) Epoch 11, batch 2300, loss[loss=0.1594, simple_loss=0.2449, pruned_loss=0.03695, over 16766.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2694, pruned_loss=0.05232, over 3319840.36 frames. ], batch size: 39, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:17:02,656 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:24,932 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 09:17:35,766 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:48,570 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:18:02,532 INFO [optim.py:368] (2/8) 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 09:18:11,650 INFO [train.py:904] (2/8) Epoch 11, batch 2350, loss[loss=0.2176, simple_loss=0.3062, pruned_loss=0.06445, over 16722.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2708, pruned_loss=0.05358, over 3321931.97 frames. ], batch size: 57, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:18:21,010 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6069, 3.6133, 2.8960, 2.1363, 2.4414, 2.1701, 3.7636, 3.3230], device='cuda:2'), covar=tensor([0.2252, 0.0548, 0.1325, 0.2213, 0.2225, 0.1742, 0.0415, 0.1135], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0255, 0.0280, 0.0275, 0.0282, 0.0221, 0.0266, 0.0298], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:18:45,244 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:19:06,008 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-29 09:19:11,008 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 09:19:19,656 INFO [train.py:904] (2/8) Epoch 11, batch 2400, loss[loss=0.1649, simple_loss=0.2577, pruned_loss=0.03603, over 17213.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2718, pruned_loss=0.05351, over 3325963.53 frames. ], batch size: 45, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:09,851 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:20:16,873 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9336, 4.9356, 5.5199, 5.4590, 5.4474, 5.0493, 5.0229, 4.7071], device='cuda:2'), covar=tensor([0.0294, 0.0449, 0.0273, 0.0362, 0.0459, 0.0330, 0.0824, 0.0411], device='cuda:2'), in_proj_covar=tensor([0.0333, 0.0347, 0.0348, 0.0326, 0.0397, 0.0367, 0.0465, 0.0294], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 09:20:17,376 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 09:20:20,789 INFO [optim.py:368] (2/8) 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 09:20:28,625 INFO [train.py:904] (2/8) Epoch 11, batch 2450, loss[loss=0.1704, simple_loss=0.2588, pruned_loss=0.041, over 17185.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2715, pruned_loss=0.0529, over 3326174.33 frames. ], batch size: 46, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:21:10,713 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0883, 5.7330, 5.8027, 5.5538, 5.5726, 6.1858, 5.7049, 5.3966], device='cuda:2'), covar=tensor([0.0825, 0.1736, 0.1941, 0.1919, 0.2735, 0.0884, 0.1292, 0.2496], device='cuda:2'), in_proj_covar=tensor([0.0359, 0.0503, 0.0547, 0.0440, 0.0579, 0.0574, 0.0430, 0.0592], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:21:12,281 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-29 09:21:22,179 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:23,588 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4211, 3.9857, 3.9796, 2.1018, 3.2333, 2.4915, 3.7667, 4.0531], device='cuda:2'), covar=tensor([0.0290, 0.0712, 0.0456, 0.1656, 0.0728, 0.0931, 0.0645, 0.0815], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0147, 0.0158, 0.0143, 0.0136, 0.0125, 0.0138, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 09:21:30,457 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0071, 4.2571, 2.5563, 4.7251, 2.9516, 4.6664, 2.7566, 3.3479], device='cuda:2'), covar=tensor([0.0179, 0.0228, 0.1181, 0.0142, 0.0700, 0.0354, 0.1133, 0.0524], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0167, 0.0187, 0.0134, 0.0167, 0.0211, 0.0193, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 09:21:31,580 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:42,828 INFO [train.py:904] (2/8) Epoch 11, batch 2500, loss[loss=0.2593, simple_loss=0.3203, pruned_loss=0.09912, over 12277.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.272, pruned_loss=0.05287, over 3317464.18 frames. ], batch size: 246, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:11,572 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:22:43,745 INFO [optim.py:368] (2/8) 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 09:22:47,571 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:49,959 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:50,607 INFO [train.py:904] (2/8) Epoch 11, batch 2550, loss[loss=0.1628, simple_loss=0.2519, pruned_loss=0.03687, over 17192.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2725, pruned_loss=0.05288, over 3318806.39 frames. ], batch size: 44, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:59,827 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:15,995 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:23:51,814 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:59,571 INFO [train.py:904] (2/8) Epoch 11, batch 2600, loss[loss=0.1605, simple_loss=0.2419, pruned_loss=0.03958, over 16837.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2719, pruned_loss=0.05267, over 3316979.73 frames. ], batch size: 39, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:24:05,894 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 09:24:11,999 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:24:26,723 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:25:01,153 INFO [optim.py:368] (2/8) 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 09:25:10,376 INFO [train.py:904] (2/8) Epoch 11, batch 2650, loss[loss=0.1932, simple_loss=0.2834, pruned_loss=0.05153, over 17048.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2723, pruned_loss=0.05207, over 3325663.77 frames. ], batch size: 53, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:25:16,194 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 09:26:18,841 INFO [train.py:904] (2/8) Epoch 11, batch 2700, loss[loss=0.1815, simple_loss=0.2616, pruned_loss=0.0507, over 16207.00 frames. ], tot_loss[loss=0.188, simple_loss=0.272, pruned_loss=0.05204, over 3326367.05 frames. ], batch size: 165, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:26:54,738 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7987, 5.0244, 5.2642, 5.0175, 5.0211, 5.6829, 5.1468, 4.8645], device='cuda:2'), covar=tensor([0.1160, 0.1923, 0.1999, 0.1969, 0.2787, 0.1069, 0.1436, 0.2324], device='cuda:2'), in_proj_covar=tensor([0.0357, 0.0503, 0.0546, 0.0437, 0.0581, 0.0572, 0.0432, 0.0586], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:27:00,749 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:27:02,251 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.62 vs. limit=5.0 2023-04-29 09:27:19,005 INFO [optim.py:368] (2/8) 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 09:27:27,257 INFO [train.py:904] (2/8) Epoch 11, batch 2750, loss[loss=0.1817, simple_loss=0.2686, pruned_loss=0.04746, over 16400.00 frames. ], tot_loss[loss=0.188, simple_loss=0.272, pruned_loss=0.05195, over 3322528.63 frames. ], batch size: 75, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:28:09,496 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 09:28:36,571 INFO [train.py:904] (2/8) Epoch 11, batch 2800, loss[loss=0.197, simple_loss=0.2748, pruned_loss=0.05966, over 16723.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2723, pruned_loss=0.0518, over 3322741.84 frames. ], batch size: 89, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:28:38,197 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 09:29:37,342 INFO [optim.py:368] (2/8) 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 09:29:37,601 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:29:44,351 INFO [train.py:904] (2/8) Epoch 11, batch 2850, loss[loss=0.1947, simple_loss=0.2709, pruned_loss=0.05927, over 16715.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2712, pruned_loss=0.05156, over 3321862.55 frames. ], batch size: 76, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:45,780 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:26,597 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0233, 2.8630, 2.7489, 2.1412, 2.6424, 2.2282, 2.7364, 2.8944], device='cuda:2'), covar=tensor([0.0295, 0.0614, 0.0466, 0.1427, 0.0631, 0.0890, 0.0517, 0.0562], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0147, 0.0159, 0.0143, 0.0136, 0.0125, 0.0137, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 09:30:44,698 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:52,870 INFO [train.py:904] (2/8) Epoch 11, batch 2900, loss[loss=0.1678, simple_loss=0.2511, pruned_loss=0.04229, over 16806.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.27, pruned_loss=0.05209, over 3319044.24 frames. ], batch size: 42, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:30:57,320 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:19,995 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:35,849 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8056, 3.9914, 2.1227, 4.5606, 2.8993, 4.4108, 2.1496, 3.1366], device='cuda:2'), covar=tensor([0.0203, 0.0283, 0.1517, 0.0151, 0.0798, 0.0454, 0.1538, 0.0603], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0168, 0.0187, 0.0136, 0.0169, 0.0214, 0.0195, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 09:31:52,195 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:55,136 INFO [optim.py:368] (2/8) 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 09:32:04,091 INFO [train.py:904] (2/8) Epoch 11, batch 2950, loss[loss=0.2146, simple_loss=0.2831, pruned_loss=0.07303, over 16494.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2702, pruned_loss=0.0528, over 3317095.18 frames. ], batch size: 146, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:32:27,832 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:33:12,380 INFO [train.py:904] (2/8) Epoch 11, batch 3000, loss[loss=0.1796, simple_loss=0.2525, pruned_loss=0.05339, over 16798.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2705, pruned_loss=0.05299, over 3323533.25 frames. ], batch size: 83, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:33:12,380 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 09:33:22,057 INFO [train.py:938] (2/8) Epoch 11, validation: loss=0.1413, simple_loss=0.2475, pruned_loss=0.01754, over 944034.00 frames. 2023-04-29 09:33:22,057 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 09:34:04,250 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:34:20,848 INFO [optim.py:368] (2/8) 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 09:34:30,328 INFO [train.py:904] (2/8) Epoch 11, batch 3050, loss[loss=0.168, simple_loss=0.2525, pruned_loss=0.04173, over 16040.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2703, pruned_loss=0.05248, over 3330011.72 frames. ], batch size: 35, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:34:38,492 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2266, 4.1906, 4.6146, 4.6145, 4.6503, 4.3176, 4.3261, 4.1163], device='cuda:2'), covar=tensor([0.0354, 0.0574, 0.0372, 0.0385, 0.0476, 0.0346, 0.0856, 0.0667], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0357, 0.0357, 0.0335, 0.0403, 0.0372, 0.0480, 0.0300], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 09:35:07,355 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:35:12,341 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8300, 2.8958, 2.6110, 4.6011, 3.7503, 4.3080, 1.5072, 3.1546], device='cuda:2'), covar=tensor([0.1231, 0.0619, 0.1016, 0.0159, 0.0249, 0.0333, 0.1433, 0.0682], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0156, 0.0178, 0.0145, 0.0196, 0.0208, 0.0175, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 09:35:35,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9211, 4.1578, 2.3985, 4.7712, 3.1222, 4.7090, 2.7014, 3.3748], device='cuda:2'), covar=tensor([0.0228, 0.0307, 0.1540, 0.0168, 0.0772, 0.0408, 0.1324, 0.0584], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0170, 0.0189, 0.0137, 0.0172, 0.0216, 0.0197, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 09:35:37,110 INFO [train.py:904] (2/8) Epoch 11, batch 3100, loss[loss=0.219, simple_loss=0.2778, pruned_loss=0.08007, over 16767.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2696, pruned_loss=0.05225, over 3324294.46 frames. ], batch size: 83, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:06,680 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 09:36:13,900 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-29 09:36:39,248 INFO [optim.py:368] (2/8) 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 09:36:39,572 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:36:47,518 INFO [train.py:904] (2/8) Epoch 11, batch 3150, loss[loss=0.1844, simple_loss=0.2796, pruned_loss=0.04458, over 16707.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2695, pruned_loss=0.05206, over 3314749.23 frames. ], batch size: 57, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:49,110 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:46,268 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:50,278 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8920, 2.4558, 1.9221, 2.2084, 2.8242, 2.6325, 3.1231, 3.0486], device='cuda:2'), covar=tensor([0.0130, 0.0253, 0.0354, 0.0307, 0.0153, 0.0241, 0.0160, 0.0158], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0209, 0.0203, 0.0204, 0.0209, 0.0206, 0.0218, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:37:56,591 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:57,461 INFO [train.py:904] (2/8) Epoch 11, batch 3200, loss[loss=0.2068, simple_loss=0.2828, pruned_loss=0.06544, over 16540.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2683, pruned_loss=0.05111, over 3316606.49 frames. ], batch size: 146, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:38:01,277 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:38:57,879 INFO [optim.py:368] (2/8) 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 09:39:06,601 INFO [train.py:904] (2/8) Epoch 11, batch 3250, loss[loss=0.1832, simple_loss=0.2624, pruned_loss=0.05204, over 16915.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2683, pruned_loss=0.05125, over 3320877.31 frames. ], batch size: 109, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:39:08,055 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:39:13,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9130, 3.9262, 4.3377, 4.3549, 4.3567, 4.0344, 4.0763, 3.9670], device='cuda:2'), covar=tensor([0.0386, 0.0605, 0.0398, 0.0402, 0.0446, 0.0402, 0.0767, 0.0545], device='cuda:2'), in_proj_covar=tensor([0.0343, 0.0359, 0.0357, 0.0337, 0.0404, 0.0374, 0.0482, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 09:39:48,588 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 09:40:15,730 INFO [train.py:904] (2/8) Epoch 11, batch 3300, loss[loss=0.163, simple_loss=0.2505, pruned_loss=0.03778, over 16818.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2687, pruned_loss=0.05096, over 3320984.41 frames. ], batch size: 42, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:40:49,983 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6347, 1.7227, 1.5682, 1.4377, 1.8007, 1.5654, 1.6017, 1.8366], device='cuda:2'), covar=tensor([0.0143, 0.0235, 0.0292, 0.0288, 0.0159, 0.0218, 0.0177, 0.0153], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0209, 0.0202, 0.0203, 0.0209, 0.0206, 0.0218, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:41:16,318 INFO [optim.py:368] (2/8) 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 09:41:24,659 INFO [train.py:904] (2/8) Epoch 11, batch 3350, loss[loss=0.1569, simple_loss=0.2456, pruned_loss=0.03413, over 17218.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2686, pruned_loss=0.05075, over 3315248.48 frames. ], batch size: 44, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:42:33,962 INFO [train.py:904] (2/8) Epoch 11, batch 3400, loss[loss=0.1731, simple_loss=0.2721, pruned_loss=0.03707, over 17066.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2687, pruned_loss=0.05053, over 3315252.50 frames. ], batch size: 50, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:43:33,846 INFO [optim.py:368] (2/8) 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 09:43:41,665 INFO [train.py:904] (2/8) Epoch 11, batch 3450, loss[loss=0.1718, simple_loss=0.2458, pruned_loss=0.04896, over 16713.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2682, pruned_loss=0.05086, over 3311968.14 frames. ], batch size: 83, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:44:08,684 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9143, 4.0952, 3.1504, 2.3258, 2.9093, 2.4338, 4.4515, 3.7660], device='cuda:2'), covar=tensor([0.2396, 0.0679, 0.1429, 0.2282, 0.2412, 0.1824, 0.0412, 0.1082], device='cuda:2'), in_proj_covar=tensor([0.0306, 0.0260, 0.0285, 0.0279, 0.0289, 0.0225, 0.0272, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:44:52,761 INFO [train.py:904] (2/8) Epoch 11, batch 3500, loss[loss=0.1969, simple_loss=0.2767, pruned_loss=0.05853, over 16810.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2672, pruned_loss=0.05037, over 3313803.72 frames. ], batch size: 124, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:45:55,145 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 11, batch 3550, loss[loss=0.1957, simple_loss=0.2714, pruned_loss=0.06004, over 16852.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2658, pruned_loss=0.05037, over 3319711.97 frames. ], batch size: 116, lr: 6.27e-03, grad_scale: 4.0 2023-04-29 09:47:12,591 INFO [train.py:904] (2/8) Epoch 11, batch 3600, loss[loss=0.1885, simple_loss=0.2766, pruned_loss=0.05019, over 16652.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2651, pruned_loss=0.04999, over 3327058.28 frames. ], batch size: 62, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:48:17,988 INFO [optim.py:368] (2/8) 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 09:48:24,356 INFO [train.py:904] (2/8) Epoch 11, batch 3650, loss[loss=0.1842, simple_loss=0.2602, pruned_loss=0.05407, over 11729.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2643, pruned_loss=0.05059, over 3305488.16 frames. ], batch size: 246, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:48:42,441 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 09:49:28,027 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8207, 4.9102, 5.0962, 4.9745, 4.9556, 5.5669, 5.1336, 4.8650], device='cuda:2'), covar=tensor([0.1161, 0.1781, 0.1712, 0.1799, 0.2571, 0.0959, 0.1292, 0.2285], device='cuda:2'), in_proj_covar=tensor([0.0353, 0.0499, 0.0538, 0.0430, 0.0571, 0.0566, 0.0428, 0.0586], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:49:37,393 INFO [train.py:904] (2/8) Epoch 11, batch 3700, loss[loss=0.202, simple_loss=0.2755, pruned_loss=0.06431, over 15673.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2639, pruned_loss=0.05246, over 3283680.44 frames. ], batch size: 191, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:49:53,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8366, 2.8462, 2.6328, 4.2914, 3.7315, 4.2184, 1.5460, 3.0433], device='cuda:2'), covar=tensor([0.1236, 0.0549, 0.0929, 0.0141, 0.0151, 0.0310, 0.1293, 0.0643], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0157, 0.0178, 0.0147, 0.0198, 0.0209, 0.0176, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 09:49:54,834 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4595, 3.7137, 4.0291, 2.7812, 3.6637, 4.0068, 3.7203, 2.5567], device='cuda:2'), covar=tensor([0.0389, 0.0128, 0.0034, 0.0283, 0.0065, 0.0070, 0.0062, 0.0294], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0071, 0.0071, 0.0126, 0.0079, 0.0090, 0.0077, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:50:41,011 INFO [optim.py:368] (2/8) 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 09:50:48,706 INFO [train.py:904] (2/8) Epoch 11, batch 3750, loss[loss=0.1676, simple_loss=0.2379, pruned_loss=0.04868, over 16882.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2649, pruned_loss=0.05428, over 3266382.75 frames. ], batch size: 96, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:50:56,816 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0642, 5.3816, 5.1041, 5.1158, 4.8477, 4.7563, 4.8395, 5.4572], device='cuda:2'), covar=tensor([0.1048, 0.0754, 0.0990, 0.0657, 0.0800, 0.0899, 0.0912, 0.0753], device='cuda:2'), in_proj_covar=tensor([0.0558, 0.0691, 0.0571, 0.0484, 0.0436, 0.0446, 0.0578, 0.0528], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:51:57,075 INFO [train.py:904] (2/8) Epoch 11, batch 3800, loss[loss=0.1796, simple_loss=0.252, pruned_loss=0.05361, over 16787.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2664, pruned_loss=0.05589, over 3262012.26 frames. ], batch size: 102, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:52:46,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9998, 4.9851, 4.8531, 4.5740, 4.5083, 4.9671, 4.7645, 4.6314], device='cuda:2'), covar=tensor([0.0547, 0.0388, 0.0226, 0.0245, 0.0821, 0.0316, 0.0328, 0.0540], device='cuda:2'), in_proj_covar=tensor([0.0262, 0.0329, 0.0307, 0.0285, 0.0330, 0.0328, 0.0209, 0.0356], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:53:00,926 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6218, 4.9188, 4.7058, 4.6769, 4.4013, 4.3721, 4.3860, 4.9747], device='cuda:2'), covar=tensor([0.1105, 0.0785, 0.0847, 0.0643, 0.0777, 0.1117, 0.0898, 0.0766], device='cuda:2'), in_proj_covar=tensor([0.0563, 0.0696, 0.0575, 0.0488, 0.0439, 0.0451, 0.0581, 0.0531], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:53:02,333 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 11, batch 3850, loss[loss=0.1891, simple_loss=0.2672, pruned_loss=0.05545, over 16824.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2666, pruned_loss=0.05647, over 3257320.93 frames. ], batch size: 96, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:53:16,080 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 09:54:00,629 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:54:19,877 INFO [train.py:904] (2/8) Epoch 11, batch 3900, loss[loss=0.1881, simple_loss=0.2646, pruned_loss=0.05575, over 15563.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2658, pruned_loss=0.05656, over 3264825.35 frames. ], batch size: 190, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:55:02,358 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7621, 4.8511, 5.0657, 4.9456, 4.9356, 5.4896, 5.0301, 4.7804], device='cuda:2'), covar=tensor([0.1083, 0.1701, 0.1816, 0.1600, 0.2335, 0.0868, 0.1283, 0.2087], device='cuda:2'), in_proj_covar=tensor([0.0349, 0.0496, 0.0533, 0.0426, 0.0563, 0.0557, 0.0423, 0.0579], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:55:25,283 INFO [optim.py:368] (2/8) 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 09:55:28,927 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:55:31,848 INFO [train.py:904] (2/8) Epoch 11, batch 3950, loss[loss=0.1881, simple_loss=0.2636, pruned_loss=0.05623, over 16240.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2642, pruned_loss=0.0563, over 3282083.24 frames. ], batch size: 165, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:55:34,149 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8124, 2.4663, 1.8861, 2.3158, 2.9222, 2.7570, 3.0255, 3.0214], device='cuda:2'), covar=tensor([0.0143, 0.0236, 0.0360, 0.0293, 0.0126, 0.0198, 0.0159, 0.0153], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0204, 0.0199, 0.0199, 0.0203, 0.0201, 0.0212, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:55:49,295 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4570, 3.3092, 2.6115, 2.1450, 2.2772, 2.0520, 3.3652, 3.0127], device='cuda:2'), covar=tensor([0.2369, 0.0648, 0.1580, 0.2264, 0.2367, 0.1929, 0.0553, 0.1204], device='cuda:2'), in_proj_covar=tensor([0.0304, 0.0258, 0.0286, 0.0279, 0.0291, 0.0224, 0.0270, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:55:53,747 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2892, 2.0552, 2.1989, 4.0869, 2.1194, 2.5032, 2.1443, 2.3263], device='cuda:2'), covar=tensor([0.1019, 0.3337, 0.2237, 0.0425, 0.3308, 0.2122, 0.3275, 0.2581], device='cuda:2'), in_proj_covar=tensor([0.0368, 0.0396, 0.0332, 0.0326, 0.0414, 0.0457, 0.0361, 0.0472], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 09:56:21,723 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:56:44,039 INFO [train.py:904] (2/8) Epoch 11, batch 4000, loss[loss=0.2294, simple_loss=0.298, pruned_loss=0.08044, over 15438.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2648, pruned_loss=0.05696, over 3265658.23 frames. ], batch size: 190, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:56:44,544 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:00,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3949, 3.4152, 3.7354, 1.6877, 3.8122, 3.8657, 3.0125, 2.8761], device='cuda:2'), covar=tensor([0.0784, 0.0247, 0.0147, 0.1266, 0.0073, 0.0143, 0.0375, 0.0432], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0101, 0.0090, 0.0140, 0.0071, 0.0106, 0.0122, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 09:57:29,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0616, 3.7026, 3.7153, 2.2663, 3.3271, 3.6610, 3.4085, 2.0144], device='cuda:2'), covar=tensor([0.0427, 0.0048, 0.0034, 0.0330, 0.0070, 0.0085, 0.0068, 0.0340], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0124, 0.0078, 0.0089, 0.0077, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:57:38,499 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4647, 4.4059, 4.4837, 2.7104, 3.7611, 4.3328, 3.8617, 2.5043], device='cuda:2'), covar=tensor([0.0407, 0.0025, 0.0024, 0.0325, 0.0060, 0.0054, 0.0046, 0.0321], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0124, 0.0078, 0.0089, 0.0077, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 09:57:48,231 INFO [optim.py:368] (2/8) 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 09:57:48,873 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:55,346 INFO [train.py:904] (2/8) Epoch 11, batch 4050, loss[loss=0.1995, simple_loss=0.2724, pruned_loss=0.06331, over 12430.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2648, pruned_loss=0.05564, over 3273439.47 frames. ], batch size: 246, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:58:11,078 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:59:08,731 INFO [train.py:904] (2/8) Epoch 11, batch 4100, loss[loss=0.2182, simple_loss=0.2856, pruned_loss=0.07539, over 11703.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2658, pruned_loss=0.05461, over 3272092.50 frames. ], batch size: 246, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:59:11,382 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 10:00:18,929 INFO [optim.py:368] (2/8) 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 10:00:26,723 INFO [train.py:904] (2/8) Epoch 11, batch 4150, loss[loss=0.2201, simple_loss=0.3062, pruned_loss=0.06695, over 17043.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2737, pruned_loss=0.05767, over 3238353.72 frames. ], batch size: 53, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:00:37,826 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9953, 2.5663, 2.6100, 1.8563, 2.7972, 2.8027, 2.4141, 2.3554], device='cuda:2'), covar=tensor([0.0699, 0.0209, 0.0206, 0.0898, 0.0095, 0.0178, 0.0430, 0.0440], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0102, 0.0091, 0.0140, 0.0071, 0.0106, 0.0123, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 10:01:44,701 INFO [train.py:904] (2/8) Epoch 11, batch 4200, loss[loss=0.2277, simple_loss=0.3161, pruned_loss=0.0696, over 15301.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2811, pruned_loss=0.05968, over 3200167.51 frames. ], batch size: 190, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:02:49,764 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:02:53,623 INFO [optim.py:368] (2/8) 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 10:02:59,807 INFO [train.py:904] (2/8) Epoch 11, batch 4250, loss[loss=0.1901, simple_loss=0.2881, pruned_loss=0.04602, over 16871.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2842, pruned_loss=0.05902, over 3202968.17 frames. ], batch size: 96, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:03:59,202 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:08,542 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:12,690 INFO [train.py:904] (2/8) Epoch 11, batch 4300, loss[loss=0.1983, simple_loss=0.2875, pruned_loss=0.0545, over 16457.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2849, pruned_loss=0.05807, over 3197645.95 frames. ], batch size: 68, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:04:55,563 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1297, 2.7673, 2.7844, 2.0139, 2.5994, 2.1822, 2.6880, 2.8972], device='cuda:2'), covar=tensor([0.0328, 0.0679, 0.0485, 0.1571, 0.0723, 0.0746, 0.0694, 0.0713], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0145, 0.0156, 0.0144, 0.0135, 0.0124, 0.0137, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 10:05:11,565 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:17,304 INFO [optim.py:368] (2/8) 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 10:05:25,482 INFO [train.py:904] (2/8) Epoch 11, batch 4350, loss[loss=0.2254, simple_loss=0.3112, pruned_loss=0.06979, over 16932.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2889, pruned_loss=0.0598, over 3183421.84 frames. ], batch size: 116, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:05:27,867 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:34,712 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:37,394 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:06:38,331 INFO [train.py:904] (2/8) Epoch 11, batch 4400, loss[loss=0.2022, simple_loss=0.2896, pruned_loss=0.05742, over 16672.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.291, pruned_loss=0.06063, over 3178128.88 frames. ], batch size: 76, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:06:51,565 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 10:07:34,285 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9411, 2.2805, 2.1793, 2.4214, 1.9987, 3.1186, 1.7516, 2.6104], device='cuda:2'), covar=tensor([0.1118, 0.0631, 0.1084, 0.0129, 0.0151, 0.0368, 0.1332, 0.0749], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0159, 0.0181, 0.0146, 0.0201, 0.0210, 0.0181, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 10:07:40,705 INFO [optim.py:368] (2/8) 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 10:07:48,892 INFO [train.py:904] (2/8) Epoch 11, batch 4450, loss[loss=0.2194, simple_loss=0.3025, pruned_loss=0.06815, over 15395.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2937, pruned_loss=0.06122, over 3204487.78 frames. ], batch size: 190, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:08:50,047 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6340, 3.1895, 2.7443, 5.1084, 4.0698, 4.4058, 1.5013, 3.1413], device='cuda:2'), covar=tensor([0.1300, 0.0623, 0.1063, 0.0128, 0.0370, 0.0323, 0.1515, 0.0779], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0159, 0.0181, 0.0146, 0.0201, 0.0210, 0.0180, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 10:09:04,901 INFO [train.py:904] (2/8) Epoch 11, batch 4500, loss[loss=0.1977, simple_loss=0.2922, pruned_loss=0.05164, over 16749.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2939, pruned_loss=0.06159, over 3199536.64 frames. ], batch size: 89, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:09:05,438 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7212, 5.6831, 5.4634, 4.8435, 5.5638, 2.1715, 5.2725, 5.3015], device='cuda:2'), covar=tensor([0.0030, 0.0028, 0.0078, 0.0250, 0.0039, 0.2027, 0.0060, 0.0093], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0116, 0.0163, 0.0156, 0.0134, 0.0175, 0.0151, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:10:04,173 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-04-29 10:10:04,875 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:10:09,391 INFO [optim.py:368] (2/8) 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 10:10:17,383 INFO [train.py:904] (2/8) Epoch 11, batch 4550, loss[loss=0.2316, simple_loss=0.3133, pruned_loss=0.07497, over 16251.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2948, pruned_loss=0.06287, over 3200672.55 frames. ], batch size: 165, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:10:53,590 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:03,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2592, 3.4194, 3.6925, 1.5749, 3.8729, 3.9229, 2.9397, 2.9038], device='cuda:2'), covar=tensor([0.0891, 0.0214, 0.0150, 0.1408, 0.0068, 0.0084, 0.0412, 0.0475], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0100, 0.0089, 0.0137, 0.0069, 0.0101, 0.0120, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 10:11:15,422 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:29,147 INFO [train.py:904] (2/8) Epoch 11, batch 4600, loss[loss=0.1907, simple_loss=0.2828, pruned_loss=0.0493, over 16906.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2954, pruned_loss=0.0628, over 3202674.42 frames. ], batch size: 116, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:12:22,219 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:26,022 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 10:12:28,013 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:35,553 INFO [optim.py:368] (2/8) 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 10:12:37,646 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:41,943 INFO [train.py:904] (2/8) Epoch 11, batch 4650, loss[loss=0.1753, simple_loss=0.2593, pruned_loss=0.04567, over 16837.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.294, pruned_loss=0.06253, over 3209218.56 frames. ], batch size: 39, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:12:47,343 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:51,756 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:37,559 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:55,735 INFO [train.py:904] (2/8) Epoch 11, batch 4700, loss[loss=0.1921, simple_loss=0.2744, pruned_loss=0.05493, over 16534.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2915, pruned_loss=0.06151, over 3204912.55 frames. ], batch size: 68, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:14:01,835 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:14:19,599 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5257, 4.5243, 4.4941, 3.7264, 4.4398, 1.5268, 4.2000, 4.2216], device='cuda:2'), covar=tensor([0.0118, 0.0114, 0.0115, 0.0440, 0.0095, 0.2399, 0.0120, 0.0195], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0116, 0.0163, 0.0156, 0.0133, 0.0175, 0.0150, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:15:01,675 INFO [optim.py:368] (2/8) 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 10:15:09,063 INFO [train.py:904] (2/8) Epoch 11, batch 4750, loss[loss=0.1894, simple_loss=0.2764, pruned_loss=0.05122, over 16683.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2878, pruned_loss=0.05982, over 3206701.05 frames. ], batch size: 62, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:15:32,274 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:16:22,052 INFO [train.py:904] (2/8) Epoch 11, batch 4800, loss[loss=0.2476, simple_loss=0.3158, pruned_loss=0.08966, over 11990.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2848, pruned_loss=0.05803, over 3199875.21 frames. ], batch size: 248, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:17:02,206 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:12,330 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:27,377 INFO [optim.py:368] (2/8) 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 10:17:35,183 INFO [train.py:904] (2/8) Epoch 11, batch 4850, loss[loss=0.2315, simple_loss=0.3188, pruned_loss=0.0721, over 11897.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2854, pruned_loss=0.05742, over 3178897.37 frames. ], batch size: 246, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:18:38,750 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:18:45,468 INFO [train.py:904] (2/8) Epoch 11, batch 4900, loss[loss=0.1741, simple_loss=0.2604, pruned_loss=0.04391, over 17189.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2842, pruned_loss=0.05606, over 3173292.49 frames. ], batch size: 46, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:19:29,984 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:19:50,194 INFO [optim.py:368] (2/8) 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 10:19:51,776 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106448.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:19:56,920 INFO [train.py:904] (2/8) Epoch 11, batch 4950, loss[loss=0.1981, simple_loss=0.2844, pruned_loss=0.05587, over 17132.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2836, pruned_loss=0.05557, over 3184393.04 frames. ], batch size: 49, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:20:01,911 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:03,058 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:52,670 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3459, 3.3949, 2.6412, 2.0368, 2.3122, 2.1455, 3.4578, 3.2294], device='cuda:2'), covar=tensor([0.2803, 0.0768, 0.1737, 0.2325, 0.2287, 0.1781, 0.0592, 0.0896], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0254, 0.0280, 0.0275, 0.0281, 0.0219, 0.0264, 0.0293], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:21:00,647 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 11, batch 5000, loss[loss=0.1853, simple_loss=0.2786, pruned_loss=0.04606, over 16752.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2855, pruned_loss=0.05574, over 3201062.13 frames. ], batch size: 89, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:21:10,911 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:28,694 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:47,685 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 10:22:14,238 INFO [optim.py:368] (2/8) 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 10:22:21,433 INFO [train.py:904] (2/8) Epoch 11, batch 5050, loss[loss=0.1838, simple_loss=0.2699, pruned_loss=0.04882, over 16667.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2858, pruned_loss=0.05563, over 3200399.89 frames. ], batch size: 62, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:23:32,362 INFO [train.py:904] (2/8) Epoch 11, batch 5100, loss[loss=0.1754, simple_loss=0.2665, pruned_loss=0.04221, over 16471.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2841, pruned_loss=0.05471, over 3196193.78 frames. ], batch size: 146, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:24:03,608 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:24:38,780 INFO [optim.py:368] (2/8) 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 10:24:44,356 INFO [train.py:904] (2/8) Epoch 11, batch 5150, loss[loss=0.2049, simple_loss=0.2932, pruned_loss=0.05833, over 16739.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2843, pruned_loss=0.0541, over 3182986.57 frames. ], batch size: 124, lr: 6.22e-03, grad_scale: 4.0 2023-04-29 10:25:12,773 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0630, 3.1245, 1.7422, 3.2625, 2.3616, 3.3347, 2.0598, 2.6556], device='cuda:2'), covar=tensor([0.0198, 0.0303, 0.1478, 0.0154, 0.0712, 0.0423, 0.1260, 0.0567], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0161, 0.0185, 0.0124, 0.0165, 0.0203, 0.0191, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 10:25:43,745 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:25:50,556 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8731, 3.9386, 2.2182, 4.3823, 2.7799, 4.3209, 2.3099, 3.0866], device='cuda:2'), covar=tensor([0.0167, 0.0237, 0.1403, 0.0113, 0.0733, 0.0310, 0.1368, 0.0609], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0162, 0.0186, 0.0125, 0.0166, 0.0204, 0.0193, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 10:25:56,050 INFO [train.py:904] (2/8) Epoch 11, batch 5200, loss[loss=0.193, simple_loss=0.2802, pruned_loss=0.05287, over 15381.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.283, pruned_loss=0.05384, over 3166093.86 frames. ], batch size: 191, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:26:42,190 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:27:04,337 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 11, batch 5250, loss[loss=0.1708, simple_loss=0.2619, pruned_loss=0.03989, over 16920.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.28, pruned_loss=0.05312, over 3183902.00 frames. ], batch size: 90, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:27:45,605 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7328, 5.0255, 4.8030, 4.7905, 4.5326, 4.5637, 4.4610, 5.1042], device='cuda:2'), covar=tensor([0.1116, 0.0824, 0.0892, 0.0676, 0.0732, 0.0939, 0.1000, 0.0749], device='cuda:2'), in_proj_covar=tensor([0.0534, 0.0661, 0.0547, 0.0455, 0.0419, 0.0427, 0.0548, 0.0509], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:27:52,545 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:28:02,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8241, 3.8688, 2.0069, 4.4815, 2.7955, 4.3908, 2.3521, 3.0572], device='cuda:2'), covar=tensor([0.0185, 0.0268, 0.1619, 0.0081, 0.0760, 0.0293, 0.1412, 0.0659], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0161, 0.0186, 0.0124, 0.0166, 0.0203, 0.0192, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 10:28:17,705 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 10:28:22,267 INFO [train.py:904] (2/8) Epoch 11, batch 5300, loss[loss=0.1803, simple_loss=0.2616, pruned_loss=0.0495, over 16871.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2761, pruned_loss=0.05143, over 3199473.18 frames. ], batch size: 116, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:28:32,916 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:29:27,199 INFO [optim.py:368] (2/8) 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 10:29:33,905 INFO [train.py:904] (2/8) Epoch 11, batch 5350, loss[loss=0.1818, simple_loss=0.2739, pruned_loss=0.04492, over 16777.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2742, pruned_loss=0.05085, over 3201278.88 frames. ], batch size: 83, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:30:26,253 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0197, 5.3048, 5.0567, 5.0593, 4.7775, 4.6993, 4.6972, 5.3906], device='cuda:2'), covar=tensor([0.1089, 0.0780, 0.0978, 0.0697, 0.0766, 0.0815, 0.0978, 0.0813], device='cuda:2'), in_proj_covar=tensor([0.0535, 0.0663, 0.0550, 0.0456, 0.0421, 0.0429, 0.0551, 0.0510], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:30:45,866 INFO [train.py:904] (2/8) Epoch 11, batch 5400, loss[loss=0.1913, simple_loss=0.2887, pruned_loss=0.04698, over 16649.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2768, pruned_loss=0.05139, over 3209222.32 frames. ], batch size: 89, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:31:18,156 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:31:52,402 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 10:31:54,542 INFO [optim.py:368] (2/8) 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 10:32:02,030 INFO [train.py:904] (2/8) Epoch 11, batch 5450, loss[loss=0.2184, simple_loss=0.2993, pruned_loss=0.06873, over 16467.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2803, pruned_loss=0.05304, over 3216510.87 frames. ], batch size: 146, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:32:34,433 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:03,978 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:19,270 INFO [train.py:904] (2/8) Epoch 11, batch 5500, loss[loss=0.28, simple_loss=0.3402, pruned_loss=0.1098, over 11673.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2886, pruned_loss=0.05847, over 3185828.49 frames. ], batch size: 248, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:34:13,637 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 10:34:18,906 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:34:23,401 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-04-29 10:34:31,621 INFO [optim.py:368] (2/8) 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 10:34:37,901 INFO [train.py:904] (2/8) Epoch 11, batch 5550, loss[loss=0.313, simple_loss=0.3574, pruned_loss=0.1343, over 11236.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2967, pruned_loss=0.06478, over 3144720.93 frames. ], batch size: 248, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:34:42,209 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 10:35:06,909 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7694, 3.6265, 3.8199, 3.6581, 3.7848, 4.1479, 3.8635, 3.6709], device='cuda:2'), covar=tensor([0.1884, 0.2061, 0.1918, 0.2286, 0.2524, 0.1646, 0.1414, 0.2482], device='cuda:2'), in_proj_covar=tensor([0.0343, 0.0474, 0.0517, 0.0413, 0.0545, 0.0542, 0.0409, 0.0562], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 10:35:38,322 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 10:35:57,900 INFO [train.py:904] (2/8) Epoch 11, batch 5600, loss[loss=0.2798, simple_loss=0.349, pruned_loss=0.1054, over 15357.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3035, pruned_loss=0.07111, over 3080325.05 frames. ], batch size: 191, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:36:12,171 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:36:12,259 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4280, 1.6407, 1.9513, 2.3846, 2.3773, 2.6540, 1.6703, 2.6018], device='cuda:2'), covar=tensor([0.0154, 0.0360, 0.0241, 0.0214, 0.0203, 0.0127, 0.0364, 0.0094], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0170, 0.0152, 0.0159, 0.0165, 0.0122, 0.0168, 0.0114], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 10:37:02,570 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:37:17,301 INFO [optim.py:368] (2/8) 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 10:37:21,711 INFO [train.py:904] (2/8) Epoch 11, batch 5650, loss[loss=0.294, simple_loss=0.3471, pruned_loss=0.1205, over 11467.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3083, pruned_loss=0.07467, over 3083566.80 frames. ], batch size: 246, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:37:22,228 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:32,230 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:48,686 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 10:38:34,219 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8560, 2.6533, 2.5223, 1.9301, 2.5153, 2.6565, 2.5373, 1.6727], device='cuda:2'), covar=tensor([0.0349, 0.0061, 0.0078, 0.0303, 0.0106, 0.0119, 0.0096, 0.0417], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0069, 0.0070, 0.0126, 0.0077, 0.0091, 0.0077, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 10:38:42,747 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:38:43,535 INFO [train.py:904] (2/8) Epoch 11, batch 5700, loss[loss=0.2488, simple_loss=0.3364, pruned_loss=0.08059, over 16713.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3105, pruned_loss=0.07704, over 3055612.77 frames. ], batch size: 124, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:39:02,921 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:39:14,831 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3857, 3.5224, 3.2394, 3.0861, 2.9576, 3.3790, 3.2133, 3.1135], device='cuda:2'), covar=tensor([0.0675, 0.0520, 0.0348, 0.0326, 0.0711, 0.0511, 0.1533, 0.0564], device='cuda:2'), in_proj_covar=tensor([0.0240, 0.0306, 0.0284, 0.0264, 0.0306, 0.0306, 0.0196, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:39:59,318 INFO [optim.py:368] (2/8) 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 10:40:04,427 INFO [train.py:904] (2/8) Epoch 11, batch 5750, loss[loss=0.2083, simple_loss=0.2982, pruned_loss=0.05919, over 16640.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3131, pruned_loss=0.07841, over 3039758.09 frames. ], batch size: 134, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:41:01,975 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 10:41:25,722 INFO [train.py:904] (2/8) Epoch 11, batch 5800, loss[loss=0.2015, simple_loss=0.2982, pruned_loss=0.05239, over 16865.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3129, pruned_loss=0.07675, over 3044959.34 frames. ], batch size: 96, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:42:35,494 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 10:42:39,025 INFO [optim.py:368] (2/8) 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 10:42:43,689 INFO [train.py:904] (2/8) Epoch 11, batch 5850, loss[loss=0.2175, simple_loss=0.2911, pruned_loss=0.07201, over 11493.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3105, pruned_loss=0.07511, over 3034812.02 frames. ], batch size: 248, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:44:00,689 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4731, 4.7834, 4.5778, 4.5597, 4.2662, 4.2865, 4.2163, 4.8756], device='cuda:2'), covar=tensor([0.0982, 0.0784, 0.0924, 0.0718, 0.0752, 0.1069, 0.0959, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0536, 0.0660, 0.0550, 0.0455, 0.0418, 0.0430, 0.0550, 0.0506], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:44:05,219 INFO [train.py:904] (2/8) Epoch 11, batch 5900, loss[loss=0.2279, simple_loss=0.3132, pruned_loss=0.0713, over 16665.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.31, pruned_loss=0.07472, over 3050671.37 frames. ], batch size: 76, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:44:48,925 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5195, 3.5331, 2.8389, 2.1351, 2.4938, 2.1976, 3.6813, 3.3863], device='cuda:2'), covar=tensor([0.2584, 0.0646, 0.1447, 0.2227, 0.2243, 0.1863, 0.0400, 0.0974], device='cuda:2'), in_proj_covar=tensor([0.0303, 0.0257, 0.0282, 0.0277, 0.0282, 0.0221, 0.0267, 0.0293], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:45:21,995 INFO [optim.py:368] (2/8) 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 10:45:26,034 INFO [train.py:904] (2/8) Epoch 11, batch 5950, loss[loss=0.2416, simple_loss=0.3328, pruned_loss=0.0752, over 16852.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3103, pruned_loss=0.07304, over 3064947.35 frames. ], batch size: 116, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:45:48,935 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0999, 3.6401, 3.4753, 1.8805, 2.8906, 2.3911, 3.5123, 3.8274], device='cuda:2'), covar=tensor([0.0269, 0.0571, 0.0561, 0.1953, 0.0790, 0.0923, 0.0646, 0.0877], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0144, 0.0158, 0.0144, 0.0136, 0.0126, 0.0137, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 10:45:53,723 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9773, 3.4361, 3.4050, 2.3364, 3.1160, 3.4288, 3.3078, 1.9323], device='cuda:2'), covar=tensor([0.0467, 0.0038, 0.0045, 0.0339, 0.0086, 0.0089, 0.0060, 0.0395], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0068, 0.0069, 0.0124, 0.0077, 0.0090, 0.0077, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 10:46:27,371 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6358, 2.7860, 2.3485, 4.1937, 2.9240, 4.0068, 1.3818, 2.6802], device='cuda:2'), covar=tensor([0.1422, 0.0726, 0.1305, 0.0162, 0.0347, 0.0403, 0.1702, 0.0993], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0158, 0.0181, 0.0143, 0.0199, 0.0208, 0.0180, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 10:46:40,552 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:46:48,947 INFO [train.py:904] (2/8) Epoch 11, batch 6000, loss[loss=0.2011, simple_loss=0.2853, pruned_loss=0.05843, over 16747.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3089, pruned_loss=0.07221, over 3085150.99 frames. ], batch size: 83, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:46:48,947 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 10:46:59,881 INFO [train.py:938] (2/8) Epoch 11, validation: loss=0.163, simple_loss=0.2761, pruned_loss=0.02492, over 944034.00 frames. 2023-04-29 10:46:59,882 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 10:47:10,239 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:47:13,538 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:47:19,748 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 10:48:12,723 INFO [optim.py:368] (2/8) 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 10:48:18,415 INFO [train.py:904] (2/8) Epoch 11, batch 6050, loss[loss=0.2028, simple_loss=0.3085, pruned_loss=0.04854, over 16723.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3074, pruned_loss=0.07076, over 3118933.82 frames. ], batch size: 89, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:48:48,419 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0585, 4.2728, 4.7713, 2.3405, 4.9816, 4.9462, 3.3260, 3.6599], device='cuda:2'), covar=tensor([0.0656, 0.0172, 0.0101, 0.1085, 0.0038, 0.0108, 0.0308, 0.0357], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0097, 0.0087, 0.0137, 0.0069, 0.0101, 0.0118, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 10:48:48,456 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:48:58,623 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7982, 2.5133, 2.2898, 3.3187, 2.4387, 3.6580, 1.4235, 2.7423], device='cuda:2'), covar=tensor([0.1223, 0.0660, 0.1157, 0.0146, 0.0171, 0.0383, 0.1565, 0.0744], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0158, 0.0181, 0.0143, 0.0199, 0.0209, 0.0180, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 10:49:11,183 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9779, 2.9403, 2.2887, 2.7459, 3.3176, 2.8678, 3.7350, 3.6076], device='cuda:2'), covar=tensor([0.0052, 0.0278, 0.0396, 0.0312, 0.0169, 0.0261, 0.0146, 0.0148], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0200, 0.0197, 0.0197, 0.0199, 0.0201, 0.0205, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:49:34,932 INFO [train.py:904] (2/8) Epoch 11, batch 6100, loss[loss=0.2125, simple_loss=0.3026, pruned_loss=0.06123, over 16804.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3065, pruned_loss=0.06943, over 3138394.03 frames. ], batch size: 83, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:50:17,291 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1182, 2.4292, 2.3477, 2.8525, 2.2172, 3.2644, 1.8029, 2.6750], device='cuda:2'), covar=tensor([0.0889, 0.0464, 0.0856, 0.0139, 0.0129, 0.0384, 0.1069, 0.0594], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0158, 0.0181, 0.0143, 0.0198, 0.0208, 0.0179, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 10:50:37,370 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0341, 4.0806, 4.4714, 4.4294, 4.4429, 4.1483, 4.1590, 4.0677], device='cuda:2'), covar=tensor([0.0307, 0.0517, 0.0347, 0.0443, 0.0462, 0.0360, 0.0907, 0.0468], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0341, 0.0343, 0.0326, 0.0388, 0.0362, 0.0463, 0.0294], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 10:50:42,794 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2254, 2.0493, 2.0784, 3.9088, 1.9388, 2.4375, 2.1917, 2.2307], device='cuda:2'), covar=tensor([0.1012, 0.3248, 0.2274, 0.0437, 0.3740, 0.2250, 0.2927, 0.3006], device='cuda:2'), in_proj_covar=tensor([0.0359, 0.0386, 0.0325, 0.0315, 0.0407, 0.0445, 0.0352, 0.0453], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:50:51,419 INFO [optim.py:368] (2/8) 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 10:50:56,529 INFO [train.py:904] (2/8) Epoch 11, batch 6150, loss[loss=0.1868, simple_loss=0.2787, pruned_loss=0.04744, over 16680.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3043, pruned_loss=0.06874, over 3144540.46 frames. ], batch size: 62, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:51:07,571 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:52:14,169 INFO [train.py:904] (2/8) Epoch 11, batch 6200, loss[loss=0.1952, simple_loss=0.2773, pruned_loss=0.05652, over 17056.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3029, pruned_loss=0.06866, over 3128519.49 frames. ], batch size: 55, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:52:42,540 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3909, 2.0174, 1.5697, 1.8200, 2.3050, 2.0552, 2.2498, 2.4922], device='cuda:2'), covar=tensor([0.0104, 0.0262, 0.0377, 0.0342, 0.0173, 0.0254, 0.0149, 0.0160], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0200, 0.0198, 0.0198, 0.0200, 0.0201, 0.0205, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:52:42,544 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:07,965 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 10:53:29,994 INFO [train.py:904] (2/8) Epoch 11, batch 6250, loss[loss=0.2215, simple_loss=0.3073, pruned_loss=0.06789, over 16245.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3022, pruned_loss=0.06817, over 3140126.19 frames. ], batch size: 35, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:36,309 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:54:38,994 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:54:45,102 INFO [train.py:904] (2/8) Epoch 11, batch 6300, loss[loss=0.2357, simple_loss=0.3132, pruned_loss=0.07906, over 16974.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3017, pruned_loss=0.06736, over 3139150.62 frames. ], batch size: 109, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:54,156 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:55:52,343 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:56:00,514 INFO [optim.py:368] (2/8) 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 10:56:03,042 INFO [train.py:904] (2/8) Epoch 11, batch 6350, loss[loss=0.2196, simple_loss=0.3057, pruned_loss=0.06672, over 17209.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3038, pruned_loss=0.06969, over 3110626.69 frames. ], batch size: 44, lr: 6.18e-03, grad_scale: 4.0 2023-04-29 10:56:10,572 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:25,827 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:59,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4965, 3.8916, 2.6397, 2.0370, 2.8052, 2.1462, 3.9125, 3.6842], device='cuda:2'), covar=tensor([0.2911, 0.0558, 0.1852, 0.2285, 0.2199, 0.1884, 0.0537, 0.0789], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0254, 0.0279, 0.0274, 0.0277, 0.0217, 0.0263, 0.0288], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 10:57:20,954 INFO [train.py:904] (2/8) Epoch 11, batch 6400, loss[loss=0.234, simple_loss=0.3117, pruned_loss=0.07812, over 15490.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3031, pruned_loss=0.07013, over 3105968.86 frames. ], batch size: 191, lr: 6.18e-03, grad_scale: 8.0 2023-04-29 10:58:35,864 INFO [optim.py:368] (2/8) 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 10:58:35,879 INFO [train.py:904] (2/8) Epoch 11, batch 6450, loss[loss=0.201, simple_loss=0.2905, pruned_loss=0.05578, over 16455.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3037, pruned_loss=0.07013, over 3084725.47 frames. ], batch size: 35, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 10:59:54,943 INFO [train.py:904] (2/8) Epoch 11, batch 6500, loss[loss=0.2148, simple_loss=0.2946, pruned_loss=0.0675, over 16372.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3009, pruned_loss=0.06925, over 3076414.39 frames. ], batch size: 35, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:00:14,832 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:01:12,920 INFO [optim.py:368] (2/8) 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 11:01:12,935 INFO [train.py:904] (2/8) Epoch 11, batch 6550, loss[loss=0.2234, simple_loss=0.3303, pruned_loss=0.05831, over 16696.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3042, pruned_loss=0.07046, over 3072024.26 frames. ], batch size: 89, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:02:13,590 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:02:15,211 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 11:02:15,223 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 11:02:25,852 INFO [train.py:904] (2/8) Epoch 11, batch 6600, loss[loss=0.2621, simple_loss=0.3239, pruned_loss=0.1001, over 11617.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3059, pruned_loss=0.07049, over 3076755.85 frames. ], batch size: 247, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:03:27,059 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0895, 3.0235, 3.1390, 1.6449, 3.3077, 3.3424, 2.5959, 2.4671], device='cuda:2'), covar=tensor([0.0781, 0.0208, 0.0173, 0.1141, 0.0074, 0.0137, 0.0427, 0.0470], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0099, 0.0088, 0.0137, 0.0069, 0.0102, 0.0119, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 11:03:41,607 INFO [optim.py:368] (2/8) 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 11:03:41,622 INFO [train.py:904] (2/8) Epoch 11, batch 6650, loss[loss=0.2142, simple_loss=0.2954, pruned_loss=0.0665, over 16730.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3062, pruned_loss=0.07139, over 3075557.49 frames. ], batch size: 134, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:04:00,014 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7946, 1.9972, 2.3250, 3.1243, 2.1694, 2.2426, 2.2531, 2.0961], device='cuda:2'), covar=tensor([0.0968, 0.2959, 0.1849, 0.0572, 0.3461, 0.2025, 0.2659, 0.3011], device='cuda:2'), in_proj_covar=tensor([0.0359, 0.0387, 0.0325, 0.0318, 0.0409, 0.0445, 0.0353, 0.0455], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:04:03,226 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:04:09,429 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6066, 4.8785, 4.6206, 4.5971, 4.3524, 4.2862, 4.3049, 4.9628], device='cuda:2'), covar=tensor([0.1110, 0.0828, 0.1042, 0.0811, 0.0880, 0.1173, 0.1102, 0.0818], device='cuda:2'), in_proj_covar=tensor([0.0545, 0.0670, 0.0558, 0.0463, 0.0423, 0.0437, 0.0559, 0.0515], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:04:56,954 INFO [train.py:904] (2/8) Epoch 11, batch 6700, loss[loss=0.2485, simple_loss=0.3254, pruned_loss=0.08579, over 16376.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3053, pruned_loss=0.07167, over 3065066.19 frames. ], batch size: 146, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:05:14,875 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:05:35,890 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5350, 5.5594, 5.2990, 4.6633, 5.3729, 2.2242, 5.1949, 5.2285], device='cuda:2'), covar=tensor([0.0052, 0.0041, 0.0137, 0.0320, 0.0057, 0.2024, 0.0086, 0.0127], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0113, 0.0161, 0.0152, 0.0132, 0.0175, 0.0147, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:05:48,424 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7427, 2.5439, 2.3697, 3.2679, 2.4061, 3.6240, 1.4972, 2.6908], device='cuda:2'), covar=tensor([0.1334, 0.0642, 0.1189, 0.0152, 0.0176, 0.0422, 0.1546, 0.0832], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0159, 0.0183, 0.0144, 0.0201, 0.0209, 0.0181, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 11:06:07,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6040, 4.3453, 4.5938, 4.7615, 4.9489, 4.4907, 4.8781, 4.9147], device='cuda:2'), covar=tensor([0.1409, 0.1201, 0.1502, 0.0663, 0.0540, 0.0869, 0.0553, 0.0605], device='cuda:2'), in_proj_covar=tensor([0.0514, 0.0636, 0.0768, 0.0645, 0.0496, 0.0497, 0.0514, 0.0578], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:06:13,521 INFO [optim.py:368] (2/8) 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 11:06:13,536 INFO [train.py:904] (2/8) Epoch 11, batch 6750, loss[loss=0.2279, simple_loss=0.3069, pruned_loss=0.07444, over 15242.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3035, pruned_loss=0.07126, over 3083010.55 frames. ], batch size: 190, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:07:28,505 INFO [train.py:904] (2/8) Epoch 11, batch 6800, loss[loss=0.2181, simple_loss=0.3037, pruned_loss=0.06626, over 16689.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3038, pruned_loss=0.07081, over 3104540.52 frames. ], batch size: 124, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:07:48,576 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:05,670 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:45,531 INFO [optim.py:368] (2/8) 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 11:08:45,547 INFO [train.py:904] (2/8) Epoch 11, batch 6850, loss[loss=0.2625, simple_loss=0.3216, pruned_loss=0.1017, over 11855.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3057, pruned_loss=0.07183, over 3084901.18 frames. ], batch size: 248, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:09:01,706 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:35,633 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:46,548 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:59,714 INFO [train.py:904] (2/8) Epoch 11, batch 6900, loss[loss=0.219, simple_loss=0.3081, pruned_loss=0.06498, over 16467.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3077, pruned_loss=0.07162, over 3085812.21 frames. ], batch size: 75, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:10:14,199 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 11:10:45,914 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:10:59,097 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:11:15,787 INFO [train.py:904] (2/8) Epoch 11, batch 6950, loss[loss=0.1926, simple_loss=0.2802, pruned_loss=0.05247, over 16820.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3097, pruned_loss=0.0736, over 3072446.29 frames. ], batch size: 102, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:11:17,884 INFO [optim.py:368] (2/8) 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 11:11:30,066 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1013, 3.7564, 3.7975, 2.4734, 3.4064, 3.6518, 3.5690, 2.0175], device='cuda:2'), covar=tensor([0.0441, 0.0031, 0.0035, 0.0318, 0.0066, 0.0099, 0.0055, 0.0384], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0069, 0.0069, 0.0127, 0.0078, 0.0091, 0.0078, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 11:12:20,777 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:12:33,547 INFO [train.py:904] (2/8) Epoch 11, batch 7000, loss[loss=0.2097, simple_loss=0.307, pruned_loss=0.05615, over 15382.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3096, pruned_loss=0.07264, over 3074840.07 frames. ], batch size: 190, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:13:15,304 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-29 11:13:22,970 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8727, 2.3317, 2.5437, 4.7155, 2.3324, 2.8590, 2.3802, 2.6816], device='cuda:2'), covar=tensor([0.0809, 0.3199, 0.2113, 0.0296, 0.3472, 0.2067, 0.2929, 0.2767], device='cuda:2'), in_proj_covar=tensor([0.0360, 0.0388, 0.0325, 0.0317, 0.0412, 0.0444, 0.0352, 0.0454], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:13:44,147 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5910, 4.5357, 4.4088, 3.7609, 4.4191, 1.5194, 4.1617, 4.2271], device='cuda:2'), covar=tensor([0.0079, 0.0077, 0.0155, 0.0344, 0.0084, 0.2438, 0.0134, 0.0172], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0113, 0.0160, 0.0151, 0.0131, 0.0175, 0.0147, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:13:52,263 INFO [train.py:904] (2/8) Epoch 11, batch 7050, loss[loss=0.2204, simple_loss=0.3047, pruned_loss=0.06805, over 16421.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3096, pruned_loss=0.07213, over 3077749.21 frames. ], batch size: 146, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:13:53,480 INFO [optim.py:368] (2/8) 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 11:14:32,886 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:14:37,533 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8265, 4.7912, 4.6095, 3.8528, 4.6502, 1.7179, 4.4343, 4.5433], device='cuda:2'), covar=tensor([0.0073, 0.0067, 0.0153, 0.0382, 0.0075, 0.2339, 0.0117, 0.0175], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0113, 0.0161, 0.0152, 0.0132, 0.0176, 0.0148, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:15:11,199 INFO [train.py:904] (2/8) Epoch 11, batch 7100, loss[loss=0.2305, simple_loss=0.3118, pruned_loss=0.07459, over 16668.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3083, pruned_loss=0.07194, over 3067534.74 frames. ], batch size: 134, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:07,260 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:16:27,132 INFO [train.py:904] (2/8) Epoch 11, batch 7150, loss[loss=0.1928, simple_loss=0.2871, pruned_loss=0.0493, over 16913.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3065, pruned_loss=0.0716, over 3061206.52 frames. ], batch size: 96, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:28,930 INFO [optim.py:368] (2/8) 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 11:16:39,210 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 11:16:49,116 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 11:17:10,533 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:17:12,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2805, 1.9904, 2.1484, 4.0037, 2.0133, 2.4543, 2.1414, 2.2250], device='cuda:2'), covar=tensor([0.1033, 0.3535, 0.2434, 0.0382, 0.3897, 0.2331, 0.3326, 0.3267], device='cuda:2'), in_proj_covar=tensor([0.0358, 0.0386, 0.0325, 0.0316, 0.0411, 0.0445, 0.0351, 0.0453], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:17:41,816 INFO [train.py:904] (2/8) Epoch 11, batch 7200, loss[loss=0.1828, simple_loss=0.2684, pruned_loss=0.04861, over 16482.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3042, pruned_loss=0.06964, over 3068669.63 frames. ], batch size: 68, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:17:54,610 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6166, 2.6183, 1.7127, 2.6796, 2.1114, 2.7998, 2.0090, 2.3405], device='cuda:2'), covar=tensor([0.0258, 0.0351, 0.1286, 0.0164, 0.0705, 0.0497, 0.1217, 0.0647], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0161, 0.0184, 0.0124, 0.0166, 0.0201, 0.0192, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 11:18:18,472 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6170, 2.7953, 2.3274, 3.8473, 2.9261, 3.9467, 1.3864, 2.9369], device='cuda:2'), covar=tensor([0.1324, 0.0645, 0.1235, 0.0152, 0.0286, 0.0392, 0.1565, 0.0743], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0160, 0.0184, 0.0146, 0.0202, 0.0211, 0.0184, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 11:19:02,056 INFO [train.py:904] (2/8) Epoch 11, batch 7250, loss[loss=0.2041, simple_loss=0.2795, pruned_loss=0.0644, over 16606.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3021, pruned_loss=0.06842, over 3047972.97 frames. ], batch size: 62, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:19:03,145 INFO [optim.py:368] (2/8) 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 11:19:55,675 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:20:00,955 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:20:16,099 INFO [train.py:904] (2/8) Epoch 11, batch 7300, loss[loss=0.2191, simple_loss=0.3061, pruned_loss=0.06602, over 16659.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3012, pruned_loss=0.06834, over 3049958.97 frames. ], batch size: 62, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:20:31,801 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4279, 2.5202, 1.9433, 2.2097, 2.7974, 2.3191, 3.1314, 3.0323], device='cuda:2'), covar=tensor([0.0060, 0.0267, 0.0429, 0.0363, 0.0182, 0.0321, 0.0141, 0.0161], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0198, 0.0195, 0.0195, 0.0199, 0.0199, 0.0203, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:21:01,036 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 11:21:34,169 INFO [train.py:904] (2/8) Epoch 11, batch 7350, loss[loss=0.2543, simple_loss=0.3195, pruned_loss=0.09451, over 11400.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.302, pruned_loss=0.06912, over 3035523.34 frames. ], batch size: 247, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:34,684 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:21:35,286 INFO [optim.py:368] (2/8) 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 11:22:54,323 INFO [train.py:904] (2/8) Epoch 11, batch 7400, loss[loss=0.1919, simple_loss=0.2818, pruned_loss=0.05104, over 16606.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3022, pruned_loss=0.06894, over 3040107.46 frames. ], batch size: 89, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:23:42,501 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:24:11,388 INFO [train.py:904] (2/8) Epoch 11, batch 7450, loss[loss=0.2227, simple_loss=0.3077, pruned_loss=0.06881, over 16599.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3042, pruned_loss=0.0706, over 3049622.79 frames. ], batch size: 62, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:24:13,636 INFO [optim.py:368] (2/8) 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 11:24:59,165 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:25:31,634 INFO [train.py:904] (2/8) Epoch 11, batch 7500, loss[loss=0.2068, simple_loss=0.2871, pruned_loss=0.06321, over 16565.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3043, pruned_loss=0.06961, over 3071104.80 frames. ], batch size: 75, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:16,354 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:26:27,942 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:26:51,093 INFO [train.py:904] (2/8) Epoch 11, batch 7550, loss[loss=0.2048, simple_loss=0.2978, pruned_loss=0.05589, over 16707.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3035, pruned_loss=0.06972, over 3064304.10 frames. ], batch size: 89, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:52,322 INFO [optim.py:368] (2/8) 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 11:27:46,153 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:28:02,005 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:28:06,704 INFO [train.py:904] (2/8) Epoch 11, batch 7600, loss[loss=0.221, simple_loss=0.3076, pruned_loss=0.0672, over 16456.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3018, pruned_loss=0.06878, over 3093783.66 frames. ], batch size: 75, lr: 6.15e-03, grad_scale: 8.0 2023-04-29 11:28:56,879 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:12,261 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:20,692 INFO [train.py:904] (2/8) Epoch 11, batch 7650, loss[loss=0.2188, simple_loss=0.3068, pruned_loss=0.06546, over 16340.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3026, pruned_loss=0.06977, over 3076336.78 frames. ], batch size: 146, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:29:23,629 INFO [optim.py:368] (2/8) 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 11:29:24,241 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3076, 3.3662, 1.8846, 3.7067, 2.4910, 3.6974, 2.0053, 2.6467], device='cuda:2'), covar=tensor([0.0236, 0.0380, 0.1587, 0.0141, 0.0839, 0.0550, 0.1499, 0.0688], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0162, 0.0186, 0.0124, 0.0166, 0.0202, 0.0194, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 11:30:25,831 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 11, batch 7700, loss[loss=0.2121, simple_loss=0.3045, pruned_loss=0.05987, over 16847.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3029, pruned_loss=0.07073, over 3076423.56 frames. ], batch size: 102, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:30:47,797 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 11:31:14,526 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 11:31:25,253 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:31:53,770 INFO [train.py:904] (2/8) Epoch 11, batch 7750, loss[loss=0.2443, simple_loss=0.3032, pruned_loss=0.09266, over 11341.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3034, pruned_loss=0.07132, over 3038930.06 frames. ], batch size: 247, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:31:56,716 INFO [optim.py:368] (2/8) 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 11:31:59,782 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:39,670 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:51,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3332, 3.2653, 3.3261, 3.4570, 3.4853, 3.2592, 3.4561, 3.5075], device='cuda:2'), covar=tensor([0.1268, 0.0941, 0.1209, 0.0711, 0.0789, 0.2104, 0.1025, 0.0869], device='cuda:2'), in_proj_covar=tensor([0.0517, 0.0637, 0.0773, 0.0651, 0.0501, 0.0499, 0.0513, 0.0580], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:32:58,057 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:33:10,716 INFO [train.py:904] (2/8) Epoch 11, batch 7800, loss[loss=0.2231, simple_loss=0.306, pruned_loss=0.07011, over 16679.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3048, pruned_loss=0.07201, over 3045188.35 frames. ], batch size: 76, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:33:35,168 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:34:25,569 INFO [train.py:904] (2/8) Epoch 11, batch 7850, loss[loss=0.2432, simple_loss=0.3185, pruned_loss=0.0839, over 15542.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3053, pruned_loss=0.07168, over 3051568.11 frames. ], batch size: 191, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:34:30,495 INFO [optim.py:368] (2/8) 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 11:34:30,932 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:35:07,064 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:35:27,275 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:35:41,206 INFO [train.py:904] (2/8) Epoch 11, batch 7900, loss[loss=0.2352, simple_loss=0.3169, pruned_loss=0.07675, over 16770.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3035, pruned_loss=0.07071, over 3067169.01 frames. ], batch size: 124, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:36:00,569 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7564, 5.1018, 5.2848, 5.1004, 5.1557, 5.6941, 5.1087, 4.9373], device='cuda:2'), covar=tensor([0.0991, 0.1703, 0.1831, 0.1682, 0.2347, 0.0918, 0.1421, 0.2298], device='cuda:2'), in_proj_covar=tensor([0.0352, 0.0489, 0.0530, 0.0420, 0.0558, 0.0557, 0.0420, 0.0574], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 11:36:37,534 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9427, 2.7928, 2.7685, 2.0404, 2.6611, 2.1544, 2.6691, 2.9454], device='cuda:2'), covar=tensor([0.0302, 0.0625, 0.0466, 0.1573, 0.0687, 0.0823, 0.0565, 0.0658], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0142, 0.0156, 0.0142, 0.0136, 0.0125, 0.0136, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 11:36:52,211 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:37:00,087 INFO [train.py:904] (2/8) Epoch 11, batch 7950, loss[loss=0.2928, simple_loss=0.3438, pruned_loss=0.1209, over 11530.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3041, pruned_loss=0.07078, over 3075907.76 frames. ], batch size: 248, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:37:04,710 INFO [optim.py:368] (2/8) 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 11:38:04,156 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:38:14,153 INFO [train.py:904] (2/8) Epoch 11, batch 8000, loss[loss=0.2215, simple_loss=0.3021, pruned_loss=0.07042, over 16725.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3052, pruned_loss=0.07168, over 3072759.87 frames. ], batch size: 134, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:39:24,916 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:39:27,478 INFO [train.py:904] (2/8) Epoch 11, batch 8050, loss[loss=0.2286, simple_loss=0.3174, pruned_loss=0.06988, over 16818.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3055, pruned_loss=0.07187, over 3069461.97 frames. ], batch size: 83, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:39:28,263 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 11:39:31,003 INFO [optim.py:368] (2/8) 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 11:40:06,492 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:40:40,895 INFO [train.py:904] (2/8) Epoch 11, batch 8100, loss[loss=0.182, simple_loss=0.272, pruned_loss=0.04599, over 17050.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3052, pruned_loss=0.07141, over 3059558.36 frames. ], batch size: 55, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:40:57,156 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 11:41:39,218 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:54,480 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:56,333 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 11:41:56,612 INFO [train.py:904] (2/8) Epoch 11, batch 8150, loss[loss=0.1978, simple_loss=0.2772, pruned_loss=0.0592, over 16615.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3025, pruned_loss=0.07048, over 3046100.50 frames. ], batch size: 134, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:42:01,359 INFO [optim.py:368] (2/8) 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 11:42:08,224 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:42:30,921 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:42:59,254 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:43:12,488 INFO [train.py:904] (2/8) Epoch 11, batch 8200, loss[loss=0.2603, simple_loss=0.319, pruned_loss=0.1007, over 11305.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2993, pruned_loss=0.06961, over 3068612.06 frames. ], batch size: 247, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:43:41,786 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:44:33,539 INFO [train.py:904] (2/8) Epoch 11, batch 8250, loss[loss=0.203, simple_loss=0.2932, pruned_loss=0.05639, over 16884.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.298, pruned_loss=0.06714, over 3048945.78 frames. ], batch size: 116, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:44:38,004 INFO [optim.py:368] (2/8) 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 11:45:46,562 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 11:45:52,506 INFO [train.py:904] (2/8) Epoch 11, batch 8300, loss[loss=0.1857, simple_loss=0.2815, pruned_loss=0.04498, over 16755.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2951, pruned_loss=0.06418, over 3031271.37 frames. ], batch size: 89, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:47:09,209 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:47:11,568 INFO [train.py:904] (2/8) Epoch 11, batch 8350, loss[loss=0.2469, simple_loss=0.3104, pruned_loss=0.09168, over 11910.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2947, pruned_loss=0.06213, over 3043640.36 frames. ], batch size: 248, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:47:16,940 INFO [optim.py:368] (2/8) 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 11:48:06,503 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0801, 3.9840, 4.4362, 4.3887, 4.3766, 4.1517, 4.1256, 4.0322], device='cuda:2'), covar=tensor([0.0271, 0.0589, 0.0375, 0.0392, 0.0448, 0.0319, 0.0970, 0.0464], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0329, 0.0331, 0.0310, 0.0378, 0.0345, 0.0448, 0.0283], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 11:48:25,293 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:48:30,268 INFO [train.py:904] (2/8) Epoch 11, batch 8400, loss[loss=0.226, simple_loss=0.2955, pruned_loss=0.07827, over 12124.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2922, pruned_loss=0.05986, over 3052293.92 frames. ], batch size: 247, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:21,775 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:45,360 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:47,548 INFO [train.py:904] (2/8) Epoch 11, batch 8450, loss[loss=0.2053, simple_loss=0.2772, pruned_loss=0.0667, over 12288.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2898, pruned_loss=0.05781, over 3040272.61 frames. ], batch size: 248, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:52,353 INFO [optim.py:368] (2/8) 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 11:50:23,588 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:50:49,984 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 11:51:00,424 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:09,670 INFO [train.py:904] (2/8) Epoch 11, batch 8500, loss[loss=0.168, simple_loss=0.2607, pruned_loss=0.03765, over 16489.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2855, pruned_loss=0.05495, over 3047393.45 frames. ], batch size: 62, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:51:18,039 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 11:51:31,231 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:33,926 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:41,405 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:52:16,987 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:52:31,121 INFO [train.py:904] (2/8) Epoch 11, batch 8550, loss[loss=0.1876, simple_loss=0.286, pruned_loss=0.04461, over 16872.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2829, pruned_loss=0.05366, over 3031264.76 frames. ], batch size: 96, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:52:37,011 INFO [optim.py:368] (2/8) 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 11:53:10,340 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6708, 1.5543, 2.0727, 2.6360, 2.4590, 2.9101, 1.8435, 2.9336], device='cuda:2'), covar=tensor([0.0146, 0.0432, 0.0278, 0.0198, 0.0234, 0.0148, 0.0391, 0.0112], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0167, 0.0149, 0.0153, 0.0164, 0.0120, 0.0167, 0.0111], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 11:53:19,579 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 11, batch 8600, loss[loss=0.1792, simple_loss=0.2738, pruned_loss=0.04225, over 16442.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2828, pruned_loss=0.05229, over 3051089.30 frames. ], batch size: 68, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:54:14,059 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:22,978 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:35,060 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:35,281 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-04-29 11:55:43,450 INFO [train.py:904] (2/8) Epoch 11, batch 8650, loss[loss=0.1867, simple_loss=0.2705, pruned_loss=0.05147, over 12405.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2812, pruned_loss=0.05099, over 3044087.12 frames. ], batch size: 250, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:55:53,865 INFO [optim.py:368] (2/8) 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 11:56:28,141 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:56:40,261 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:57:02,012 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4933, 3.6201, 3.3304, 3.1926, 3.1165, 3.4623, 3.2980, 3.2499], device='cuda:2'), covar=tensor([0.0657, 0.0660, 0.0326, 0.0290, 0.0684, 0.0589, 0.1252, 0.0643], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0301, 0.0274, 0.0255, 0.0290, 0.0293, 0.0191, 0.0318], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 11:57:30,807 INFO [train.py:904] (2/8) Epoch 11, batch 8700, loss[loss=0.1906, simple_loss=0.2871, pruned_loss=0.04708, over 15413.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.278, pruned_loss=0.0495, over 3034552.10 frames. ], batch size: 190, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:57:44,696 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8393, 3.7522, 3.9748, 3.8057, 3.8886, 4.3284, 3.9734, 3.7184], device='cuda:2'), covar=tensor([0.1807, 0.2103, 0.1775, 0.2347, 0.2801, 0.1289, 0.1397, 0.2706], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0459, 0.0500, 0.0393, 0.0523, 0.0527, 0.0398, 0.0539], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 11:57:52,175 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 11:58:12,291 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 11:58:34,004 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:59:05,959 INFO [train.py:904] (2/8) Epoch 11, batch 8750, loss[loss=0.1985, simple_loss=0.2986, pruned_loss=0.04923, over 16874.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2772, pruned_loss=0.0485, over 3029512.69 frames. ], batch size: 116, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:59:15,713 INFO [optim.py:368] (2/8) 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 11:59:36,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 12:00:22,507 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:00:22,680 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:00,283 INFO [train.py:904] (2/8) Epoch 11, batch 8800, loss[loss=0.1679, simple_loss=0.2585, pruned_loss=0.03865, over 16590.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2758, pruned_loss=0.04742, over 3028208.58 frames. ], batch size: 62, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 12:01:03,749 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:27,782 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:34,193 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:02:29,299 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:02:45,832 INFO [train.py:904] (2/8) Epoch 11, batch 8850, loss[loss=0.191, simple_loss=0.2744, pruned_loss=0.05378, over 12692.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2779, pruned_loss=0.04676, over 3020949.16 frames. ], batch size: 248, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:02:52,480 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:11,301 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:29,779 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:42,516 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:04:04,744 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1000, 2.6674, 2.6903, 1.7845, 2.8568, 2.9175, 2.4714, 2.4302], device='cuda:2'), covar=tensor([0.0623, 0.0220, 0.0174, 0.0979, 0.0074, 0.0152, 0.0422, 0.0414], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0097, 0.0084, 0.0136, 0.0067, 0.0098, 0.0118, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 12:04:28,003 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110400.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:04:30,860 INFO [train.py:904] (2/8) Epoch 11, batch 8900, loss[loss=0.1758, simple_loss=0.2686, pruned_loss=0.04149, over 15236.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2783, pruned_loss=0.04597, over 3038013.91 frames. ], batch size: 190, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:06:36,592 INFO [train.py:904] (2/8) Epoch 11, batch 8950, loss[loss=0.1785, simple_loss=0.2677, pruned_loss=0.0446, over 16996.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2783, pruned_loss=0.04636, over 3066927.57 frames. ], batch size: 109, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:06:45,507 INFO [optim.py:368] (2/8) 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 12:07:08,190 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:07:17,822 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:07:54,404 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4224, 3.5799, 1.9547, 3.9657, 2.5494, 3.8474, 2.0192, 2.8183], device='cuda:2'), covar=tensor([0.0219, 0.0300, 0.1644, 0.0109, 0.0802, 0.0431, 0.1539, 0.0684], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0156, 0.0183, 0.0120, 0.0161, 0.0194, 0.0190, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 12:08:26,547 INFO [train.py:904] (2/8) Epoch 11, batch 9000, loss[loss=0.1855, simple_loss=0.2679, pruned_loss=0.05151, over 12291.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2751, pruned_loss=0.04514, over 3055667.11 frames. ], batch size: 250, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:08:26,547 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 12:08:36,930 INFO [train.py:938] (2/8) Epoch 11, validation: loss=0.1545, simple_loss=0.2586, pruned_loss=0.02523, over 944034.00 frames. 2023-04-29 12:08:36,930 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 12:10:21,906 INFO [train.py:904] (2/8) Epoch 11, batch 9050, loss[loss=0.1748, simple_loss=0.2598, pruned_loss=0.04494, over 16172.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2771, pruned_loss=0.04602, over 3054945.12 frames. ], batch size: 165, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:10:28,904 INFO [optim.py:368] (2/8) 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 12:12:06,397 INFO [train.py:904] (2/8) Epoch 11, batch 9100, loss[loss=0.1885, simple_loss=0.2818, pruned_loss=0.04764, over 16711.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2761, pruned_loss=0.04608, over 3054543.76 frames. ], batch size: 62, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:12:25,129 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:13:36,696 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:06,379 INFO [train.py:904] (2/8) Epoch 11, batch 9150, loss[loss=0.1683, simple_loss=0.2673, pruned_loss=0.03464, over 16879.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2756, pruned_loss=0.04557, over 3036711.61 frames. ], batch size: 96, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:14:15,980 INFO [optim.py:368] (2/8) 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 12:14:23,372 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:52,271 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:53,800 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:56,536 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:15:32,737 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 12:15:46,115 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 11, batch 9200, loss[loss=0.1614, simple_loss=0.2406, pruned_loss=0.04112, over 12056.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2711, pruned_loss=0.04474, over 3020363.32 frames. ], batch size: 248, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:15:51,090 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 12:16:20,527 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9550, 4.2275, 4.0300, 4.0513, 3.6776, 3.7605, 3.8395, 4.1812], device='cuda:2'), covar=tensor([0.0997, 0.0864, 0.0901, 0.0634, 0.0762, 0.1764, 0.0968, 0.1033], device='cuda:2'), in_proj_covar=tensor([0.0514, 0.0634, 0.0525, 0.0442, 0.0398, 0.0419, 0.0534, 0.0488], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:16:24,197 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:16:52,954 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:04,252 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 12:17:18,047 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:24,346 INFO [train.py:904] (2/8) Epoch 11, batch 9250, loss[loss=0.1778, simple_loss=0.2693, pruned_loss=0.04312, over 16783.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2711, pruned_loss=0.04469, over 3010276.44 frames. ], batch size: 124, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:17:32,955 INFO [optim.py:368] (2/8) 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 12:17:53,020 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:55,827 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 12:18:05,074 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:06,529 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 12:19:16,916 INFO [train.py:904] (2/8) Epoch 11, batch 9300, loss[loss=0.167, simple_loss=0.2507, pruned_loss=0.04164, over 12076.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2694, pruned_loss=0.04402, over 3017560.16 frames. ], batch size: 246, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:19:45,571 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:50,899 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7457, 3.9108, 2.2082, 4.3508, 2.8913, 4.2336, 2.4466, 3.2198], device='cuda:2'), covar=tensor([0.0217, 0.0285, 0.1575, 0.0159, 0.0743, 0.0394, 0.1463, 0.0575], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0155, 0.0182, 0.0119, 0.0162, 0.0192, 0.0191, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 12:19:59,651 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:21:01,156 INFO [train.py:904] (2/8) Epoch 11, batch 9350, loss[loss=0.2079, simple_loss=0.2822, pruned_loss=0.06678, over 11819.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2695, pruned_loss=0.04397, over 3037393.65 frames. ], batch size: 250, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:21:10,062 INFO [optim.py:368] (2/8) 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 12:21:16,990 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-29 12:21:35,941 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:22:41,698 INFO [train.py:904] (2/8) Epoch 11, batch 9400, loss[loss=0.1804, simple_loss=0.2687, pruned_loss=0.04603, over 12437.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2697, pruned_loss=0.04359, over 3036619.83 frames. ], batch size: 246, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:22:52,445 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 12:23:36,514 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:23:57,202 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:21,277 INFO [train.py:904] (2/8) Epoch 11, batch 9450, loss[loss=0.1967, simple_loss=0.2735, pruned_loss=0.05998, over 12382.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2711, pruned_loss=0.04365, over 3033502.09 frames. ], batch size: 248, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:24:27,347 INFO [optim.py:368] (2/8) 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 12:24:33,040 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:34,596 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:50,262 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:06,699 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:17,173 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-29 12:25:33,498 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:00,931 INFO [train.py:904] (2/8) Epoch 11, batch 9500, loss[loss=0.1819, simple_loss=0.2763, pruned_loss=0.04379, over 15395.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.27, pruned_loss=0.04339, over 3024983.38 frames. ], batch size: 192, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:26:15,566 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:38,133 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:41,209 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4062, 4.4587, 4.6045, 4.4521, 4.4577, 4.9969, 4.6740, 4.3569], device='cuda:2'), covar=tensor([0.1194, 0.1780, 0.2020, 0.1886, 0.2606, 0.0957, 0.1329, 0.2290], device='cuda:2'), in_proj_covar=tensor([0.0325, 0.0457, 0.0499, 0.0398, 0.0520, 0.0533, 0.0397, 0.0532], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:26:43,502 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:27:46,818 INFO [train.py:904] (2/8) Epoch 11, batch 9550, loss[loss=0.1815, simple_loss=0.2656, pruned_loss=0.04868, over 12591.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2694, pruned_loss=0.04331, over 3019205.03 frames. ], batch size: 248, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:27:55,299 INFO [optim.py:368] (2/8) 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 12:28:33,609 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 12:29:10,466 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 12:29:26,769 INFO [train.py:904] (2/8) Epoch 11, batch 9600, loss[loss=0.1709, simple_loss=0.2524, pruned_loss=0.0447, over 12244.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.271, pruned_loss=0.04415, over 3007214.57 frames. ], batch size: 246, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:30:35,960 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0969, 1.4289, 1.6551, 2.0539, 2.1000, 2.1879, 1.6500, 2.1813], device='cuda:2'), covar=tensor([0.0210, 0.0388, 0.0225, 0.0240, 0.0236, 0.0162, 0.0337, 0.0092], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0166, 0.0148, 0.0151, 0.0161, 0.0118, 0.0165, 0.0108], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 12:31:14,963 INFO [train.py:904] (2/8) Epoch 11, batch 9650, loss[loss=0.1757, simple_loss=0.2725, pruned_loss=0.03944, over 16728.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2735, pruned_loss=0.04467, over 3025976.44 frames. ], batch size: 89, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:31:24,137 INFO [optim.py:368] (2/8) 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 12:33:03,902 INFO [train.py:904] (2/8) Epoch 11, batch 9700, loss[loss=0.162, simple_loss=0.2635, pruned_loss=0.03031, over 16848.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2725, pruned_loss=0.04432, over 3035506.40 frames. ], batch size: 102, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:33:49,082 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:33:52,459 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:34:46,293 INFO [train.py:904] (2/8) Epoch 11, batch 9750, loss[loss=0.1757, simple_loss=0.2699, pruned_loss=0.04073, over 16873.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2716, pruned_loss=0.04454, over 3045372.73 frames. ], batch size: 116, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:34:53,729 INFO [optim.py:368] (2/8) 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 12:35:17,273 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:35:58,508 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:04,590 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9258, 2.2264, 1.7763, 2.0592, 2.6118, 2.3069, 2.6909, 2.8180], device='cuda:2'), covar=tensor([0.0080, 0.0344, 0.0407, 0.0397, 0.0205, 0.0287, 0.0151, 0.0185], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0198, 0.0192, 0.0192, 0.0196, 0.0195, 0.0193, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:36:25,539 INFO [train.py:904] (2/8) Epoch 11, batch 9800, loss[loss=0.1577, simple_loss=0.2447, pruned_loss=0.03533, over 12458.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2716, pruned_loss=0.04358, over 3067115.81 frames. ], batch size: 250, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:36:40,681 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0734, 1.4219, 1.6591, 2.0805, 2.1287, 2.2407, 1.7233, 2.1978], device='cuda:2'), covar=tensor([0.0213, 0.0361, 0.0238, 0.0232, 0.0226, 0.0145, 0.0328, 0.0117], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0167, 0.0148, 0.0152, 0.0161, 0.0118, 0.0165, 0.0109], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 12:36:49,141 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:50,546 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:37:21,277 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 12:38:09,150 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4926, 4.5151, 4.9381, 4.9174, 4.9089, 4.6117, 4.5721, 4.3943], device='cuda:2'), covar=tensor([0.0280, 0.0610, 0.0370, 0.0388, 0.0447, 0.0372, 0.0831, 0.0384], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0308, 0.0314, 0.0293, 0.0353, 0.0330, 0.0417, 0.0266], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-29 12:38:09,308 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6519, 3.9595, 3.0095, 2.1903, 2.4963, 2.3940, 4.1767, 3.3653], device='cuda:2'), covar=tensor([0.2624, 0.0588, 0.1488, 0.2381, 0.2421, 0.1827, 0.0359, 0.1062], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0248, 0.0274, 0.0269, 0.0255, 0.0216, 0.0256, 0.0279], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:38:11,446 INFO [train.py:904] (2/8) Epoch 11, batch 9850, loss[loss=0.1611, simple_loss=0.2512, pruned_loss=0.0355, over 12573.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2729, pruned_loss=0.04342, over 3083393.08 frames. ], batch size: 248, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:38:20,190 INFO [optim.py:368] (2/8) 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 12:39:42,361 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 12:40:02,790 INFO [train.py:904] (2/8) Epoch 11, batch 9900, loss[loss=0.1656, simple_loss=0.2692, pruned_loss=0.03098, over 16802.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2729, pruned_loss=0.04317, over 3078408.50 frames. ], batch size: 83, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:41:24,420 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2083, 3.3659, 1.8584, 3.6127, 2.4776, 3.5445, 2.0191, 2.7385], device='cuda:2'), covar=tensor([0.0241, 0.0314, 0.1617, 0.0195, 0.0817, 0.0610, 0.1618, 0.0678], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0156, 0.0184, 0.0121, 0.0163, 0.0195, 0.0194, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 12:41:34,259 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:42:01,103 INFO [train.py:904] (2/8) Epoch 11, batch 9950, loss[loss=0.1722, simple_loss=0.2714, pruned_loss=0.03651, over 16642.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2753, pruned_loss=0.04357, over 3085732.47 frames. ], batch size: 134, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:42:11,464 INFO [optim.py:368] (2/8) 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 12:42:37,691 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8293, 3.7863, 3.9494, 3.7593, 3.8680, 4.2832, 3.9413, 3.6426], device='cuda:2'), covar=tensor([0.1872, 0.2205, 0.1955, 0.2385, 0.2851, 0.1546, 0.1507, 0.2890], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0455, 0.0498, 0.0392, 0.0519, 0.0532, 0.0396, 0.0531], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:42:40,555 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:43:45,926 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8660, 5.1878, 4.9993, 4.9987, 4.6701, 4.6657, 4.6326, 5.2521], device='cuda:2'), covar=tensor([0.0923, 0.0774, 0.0868, 0.0612, 0.0700, 0.0827, 0.0921, 0.0762], device='cuda:2'), in_proj_covar=tensor([0.0505, 0.0629, 0.0510, 0.0434, 0.0393, 0.0411, 0.0521, 0.0478], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:44:02,030 INFO [train.py:904] (2/8) Epoch 11, batch 10000, loss[loss=0.1586, simple_loss=0.248, pruned_loss=0.03459, over 17007.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2736, pruned_loss=0.04297, over 3094988.42 frames. ], batch size: 55, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:44:45,116 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:49,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4045, 4.3690, 4.7823, 4.7341, 4.7418, 4.4989, 4.4407, 4.3334], device='cuda:2'), covar=tensor([0.0226, 0.0464, 0.0324, 0.0402, 0.0393, 0.0302, 0.0740, 0.0335], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0309, 0.0313, 0.0294, 0.0353, 0.0331, 0.0419, 0.0267], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:2') 2023-04-29 12:44:51,245 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5561, 3.6225, 3.3552, 3.1019, 3.2049, 3.5255, 3.3316, 3.2938], device='cuda:2'), covar=tensor([0.0574, 0.0456, 0.0275, 0.0266, 0.0582, 0.0397, 0.1136, 0.0465], device='cuda:2'), in_proj_covar=tensor([0.0228, 0.0293, 0.0268, 0.0249, 0.0285, 0.0288, 0.0187, 0.0309], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:44:55,124 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:45:42,135 INFO [train.py:904] (2/8) Epoch 11, batch 10050, loss[loss=0.1923, simple_loss=0.2867, pruned_loss=0.04898, over 16382.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2741, pruned_loss=0.04309, over 3093367.62 frames. ], batch size: 146, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:45:50,236 INFO [optim.py:368] (2/8) 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 12:45:53,375 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 12:46:22,297 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:46:22,513 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0736, 2.2293, 1.8092, 1.9894, 2.6325, 2.3553, 2.8052, 2.8315], device='cuda:2'), covar=tensor([0.0098, 0.0381, 0.0465, 0.0460, 0.0237, 0.0318, 0.0202, 0.0232], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0202, 0.0195, 0.0196, 0.0199, 0.0197, 0.0195, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:46:38,798 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:47:14,832 INFO [train.py:904] (2/8) Epoch 11, batch 10100, loss[loss=0.1888, simple_loss=0.2722, pruned_loss=0.0527, over 16948.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2747, pruned_loss=0.04342, over 3084211.68 frames. ], batch size: 109, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:47:15,634 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6470, 2.6864, 2.3728, 2.2727, 3.0423, 2.6743, 3.3922, 3.1887], device='cuda:2'), covar=tensor([0.0071, 0.0297, 0.0329, 0.0376, 0.0191, 0.0296, 0.0155, 0.0186], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0201, 0.0193, 0.0194, 0.0197, 0.0196, 0.0193, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 12:47:35,953 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:48:22,943 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-29 12:48:58,023 INFO [train.py:904] (2/8) Epoch 12, batch 0, loss[loss=0.2214, simple_loss=0.2987, pruned_loss=0.07208, over 17057.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2987, pruned_loss=0.07208, over 17057.00 frames. ], batch size: 53, lr: 5.82e-03, grad_scale: 8.0 2023-04-29 12:48:58,023 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 12:49:05,307 INFO [train.py:938] (2/8) Epoch 12, validation: loss=0.1543, simple_loss=0.2577, pruned_loss=0.0254, over 944034.00 frames. 2023-04-29 12:49:05,307 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 12:49:12,535 INFO [optim.py:368] (2/8) 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 12:49:21,581 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:50:15,979 INFO [train.py:904] (2/8) Epoch 12, batch 50, loss[loss=0.2194, simple_loss=0.2921, pruned_loss=0.07333, over 16776.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2869, pruned_loss=0.06198, over 751230.57 frames. ], batch size: 124, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:50:43,336 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7969, 4.0752, 4.3298, 3.0453, 3.6644, 4.2361, 3.8522, 2.4834], device='cuda:2'), covar=tensor([0.0373, 0.0050, 0.0027, 0.0266, 0.0089, 0.0055, 0.0056, 0.0346], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0067, 0.0068, 0.0124, 0.0077, 0.0086, 0.0075, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 12:51:25,697 INFO [train.py:904] (2/8) Epoch 12, batch 100, loss[loss=0.2057, simple_loss=0.2877, pruned_loss=0.0619, over 16516.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2816, pruned_loss=0.05873, over 1321376.24 frames. ], batch size: 68, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:51:34,341 INFO [optim.py:368] (2/8) 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 12:52:07,162 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9759, 1.8955, 2.3505, 2.9964, 2.6230, 3.2428, 2.3115, 3.3638], device='cuda:2'), covar=tensor([0.0172, 0.0391, 0.0248, 0.0201, 0.0241, 0.0167, 0.0328, 0.0109], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0171, 0.0152, 0.0156, 0.0166, 0.0122, 0.0170, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 12:52:18,939 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 12:52:31,943 INFO [train.py:904] (2/8) Epoch 12, batch 150, loss[loss=0.205, simple_loss=0.2809, pruned_loss=0.06453, over 16362.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2802, pruned_loss=0.05842, over 1768605.84 frames. ], batch size: 145, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:02,368 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:53:31,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0937, 1.9353, 2.4476, 3.0864, 2.7429, 3.3481, 2.4829, 3.3924], device='cuda:2'), covar=tensor([0.0164, 0.0397, 0.0255, 0.0206, 0.0237, 0.0162, 0.0321, 0.0118], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0170, 0.0151, 0.0157, 0.0166, 0.0122, 0.0169, 0.0112], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 12:53:40,723 INFO [train.py:904] (2/8) Epoch 12, batch 200, loss[loss=0.2056, simple_loss=0.2994, pruned_loss=0.05594, over 17049.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2792, pruned_loss=0.0578, over 2115161.53 frames. ], batch size: 53, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:41,250 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 12:53:50,233 INFO [optim.py:368] (2/8) 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 12:54:21,398 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:54:49,915 INFO [train.py:904] (2/8) Epoch 12, batch 250, loss[loss=0.1888, simple_loss=0.2592, pruned_loss=0.05917, over 16456.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2758, pruned_loss=0.05603, over 2388325.95 frames. ], batch size: 75, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:55:26,515 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:55:32,029 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7709, 1.6820, 2.1563, 2.6096, 2.6272, 2.5504, 1.8330, 2.8207], device='cuda:2'), covar=tensor([0.0134, 0.0355, 0.0239, 0.0195, 0.0192, 0.0195, 0.0334, 0.0088], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0171, 0.0153, 0.0159, 0.0168, 0.0123, 0.0170, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 12:55:57,752 INFO [train.py:904] (2/8) Epoch 12, batch 300, loss[loss=0.1716, simple_loss=0.2671, pruned_loss=0.03809, over 17137.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2724, pruned_loss=0.05373, over 2602595.07 frames. ], batch size: 47, lr: 5.82e-03, grad_scale: 1.0 2023-04-29 12:56:09,473 INFO [optim.py:368] (2/8) 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 12:57:10,667 INFO [train.py:904] (2/8) Epoch 12, batch 350, loss[loss=0.1691, simple_loss=0.2535, pruned_loss=0.04233, over 16386.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.269, pruned_loss=0.05176, over 2766916.36 frames. ], batch size: 68, lr: 5.81e-03, grad_scale: 1.0 2023-04-29 12:58:17,749 INFO [train.py:904] (2/8) Epoch 12, batch 400, loss[loss=0.1856, simple_loss=0.2596, pruned_loss=0.05577, over 16803.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.268, pruned_loss=0.05246, over 2895917.25 frames. ], batch size: 102, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:58:27,674 INFO [optim.py:368] (2/8) 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 12:58:36,974 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:58:53,645 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-29 12:59:26,047 INFO [train.py:904] (2/8) Epoch 12, batch 450, loss[loss=0.1586, simple_loss=0.2467, pruned_loss=0.0352, over 17238.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2658, pruned_loss=0.05117, over 2996152.78 frames. ], batch size: 45, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:59:46,980 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3451, 4.3512, 4.5040, 4.3999, 4.3189, 4.9498, 4.4694, 4.1817], device='cuda:2'), covar=tensor([0.1456, 0.2064, 0.2149, 0.2163, 0.3350, 0.1319, 0.1703, 0.2793], device='cuda:2'), in_proj_covar=tensor([0.0351, 0.0501, 0.0550, 0.0435, 0.0581, 0.0579, 0.0436, 0.0583], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 12:59:55,920 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:00,707 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:27,067 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:00:33,812 INFO [train.py:904] (2/8) Epoch 12, batch 500, loss[loss=0.1747, simple_loss=0.2662, pruned_loss=0.04165, over 17134.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2642, pruned_loss=0.05013, over 3058140.25 frames. ], batch size: 48, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:00:45,216 INFO [optim.py:368] (2/8) 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 13:01:02,759 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:01:19,590 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 13:01:29,232 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 13:01:32,810 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9344, 4.9818, 5.5388, 5.4993, 5.4717, 5.1414, 5.0639, 4.8679], device='cuda:2'), covar=tensor([0.0319, 0.0469, 0.0298, 0.0346, 0.0370, 0.0317, 0.0880, 0.0393], device='cuda:2'), in_proj_covar=tensor([0.0333, 0.0346, 0.0348, 0.0328, 0.0389, 0.0367, 0.0468, 0.0298], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 13:01:44,735 INFO [train.py:904] (2/8) Epoch 12, batch 550, loss[loss=0.1863, simple_loss=0.2739, pruned_loss=0.0493, over 16572.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2634, pruned_loss=0.0497, over 3100183.78 frames. ], batch size: 62, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:02:55,015 INFO [train.py:904] (2/8) Epoch 12, batch 600, loss[loss=0.1764, simple_loss=0.2673, pruned_loss=0.04278, over 17115.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2632, pruned_loss=0.04991, over 3149979.71 frames. ], batch size: 49, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:03:06,851 INFO [optim.py:368] (2/8) 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 13:04:02,133 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 13:04:04,841 INFO [train.py:904] (2/8) Epoch 12, batch 650, loss[loss=0.1697, simple_loss=0.2633, pruned_loss=0.03807, over 17064.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2631, pruned_loss=0.05011, over 3185677.36 frames. ], batch size: 55, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:05:14,225 INFO [train.py:904] (2/8) Epoch 12, batch 700, loss[loss=0.2144, simple_loss=0.2794, pruned_loss=0.07471, over 16883.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2625, pruned_loss=0.04991, over 3223738.71 frames. ], batch size: 116, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:05:14,657 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2103, 4.9160, 5.1143, 5.4227, 5.5445, 4.8358, 5.4769, 5.4891], device='cuda:2'), covar=tensor([0.1562, 0.1321, 0.2245, 0.0821, 0.0658, 0.0740, 0.0680, 0.0685], device='cuda:2'), in_proj_covar=tensor([0.0543, 0.0676, 0.0820, 0.0683, 0.0521, 0.0527, 0.0542, 0.0618], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:05:26,018 INFO [optim.py:368] (2/8) 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 13:06:07,009 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:06:24,418 INFO [train.py:904] (2/8) Epoch 12, batch 750, loss[loss=0.1773, simple_loss=0.2665, pruned_loss=0.04405, over 17133.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2631, pruned_loss=0.05006, over 3242304.30 frames. ], batch size: 48, lr: 5.80e-03, grad_scale: 2.0 2023-04-29 13:06:51,814 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:19,611 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8570, 2.1944, 2.4033, 4.7559, 2.2712, 2.7493, 2.4111, 2.4202], device='cuda:2'), covar=tensor([0.0883, 0.3547, 0.2294, 0.0285, 0.3727, 0.2222, 0.2936, 0.3428], device='cuda:2'), in_proj_covar=tensor([0.0366, 0.0394, 0.0333, 0.0322, 0.0415, 0.0449, 0.0357, 0.0461], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:07:29,161 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:07:32,736 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:34,635 INFO [train.py:904] (2/8) Epoch 12, batch 800, loss[loss=0.1446, simple_loss=0.2261, pruned_loss=0.03156, over 16792.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2622, pruned_loss=0.04913, over 3254667.12 frames. ], batch size: 39, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:07:45,049 INFO [optim.py:368] (2/8) 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 13:08:33,906 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:08:42,803 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-29 13:08:42,876 INFO [train.py:904] (2/8) Epoch 12, batch 850, loss[loss=0.1613, simple_loss=0.2461, pruned_loss=0.03823, over 16849.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2616, pruned_loss=0.04861, over 3269513.97 frames. ], batch size: 42, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:08:58,625 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0832, 4.8235, 5.0969, 5.3623, 5.5033, 4.8104, 5.4868, 5.4352], device='cuda:2'), covar=tensor([0.1506, 0.1197, 0.1569, 0.0583, 0.0435, 0.0761, 0.0390, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0560, 0.0693, 0.0841, 0.0703, 0.0534, 0.0540, 0.0553, 0.0635], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:09:26,246 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-04-29 13:09:52,005 INFO [train.py:904] (2/8) Epoch 12, batch 900, loss[loss=0.1498, simple_loss=0.2291, pruned_loss=0.03519, over 16764.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2603, pruned_loss=0.04815, over 3278218.31 frames. ], batch size: 39, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:10:02,362 INFO [optim.py:368] (2/8) 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 13:10:05,165 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:26,064 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:35,785 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9989, 4.6163, 3.4300, 2.4543, 2.8786, 2.6372, 4.7377, 3.9185], device='cuda:2'), covar=tensor([0.2471, 0.0489, 0.1421, 0.2248, 0.2576, 0.1725, 0.0356, 0.1088], device='cuda:2'), in_proj_covar=tensor([0.0303, 0.0259, 0.0284, 0.0277, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:10:59,466 INFO [train.py:904] (2/8) Epoch 12, batch 950, loss[loss=0.1763, simple_loss=0.2552, pruned_loss=0.04869, over 16244.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2607, pruned_loss=0.04876, over 3282520.53 frames. ], batch size: 165, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:11:21,284 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:28,349 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:50,137 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:12:07,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6228, 2.6199, 1.7793, 2.7239, 2.1894, 2.7802, 2.0359, 2.3907], device='cuda:2'), covar=tensor([0.0286, 0.0361, 0.1450, 0.0240, 0.0685, 0.0471, 0.1232, 0.0629], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0166, 0.0189, 0.0134, 0.0167, 0.0208, 0.0198, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 13:12:07,851 INFO [train.py:904] (2/8) Epoch 12, batch 1000, loss[loss=0.1809, simple_loss=0.2539, pruned_loss=0.05392, over 12348.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2598, pruned_loss=0.0487, over 3283059.67 frames. ], batch size: 246, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:12:18,367 INFO [optim.py:368] (2/8) 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 13:12:43,263 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:12:49,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5630, 3.4661, 2.7435, 2.1839, 2.2876, 2.1633, 3.5280, 3.1518], device='cuda:2'), covar=tensor([0.2465, 0.0632, 0.1552, 0.2358, 0.2373, 0.1870, 0.0506, 0.1272], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0258, 0.0283, 0.0276, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:13:15,637 INFO [train.py:904] (2/8) Epoch 12, batch 1050, loss[loss=0.1627, simple_loss=0.2606, pruned_loss=0.03243, over 17122.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2591, pruned_loss=0.04878, over 3277999.83 frames. ], batch size: 49, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:13:32,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8234, 4.5624, 4.5269, 3.3463, 3.8000, 4.4674, 4.0033, 2.6959], device='cuda:2'), covar=tensor([0.0397, 0.0032, 0.0024, 0.0263, 0.0074, 0.0058, 0.0063, 0.0347], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0070, 0.0069, 0.0125, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 13:13:35,391 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 13:13:36,038 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7713, 3.7851, 4.0295, 2.0613, 4.1899, 4.1992, 3.2280, 3.3172], device='cuda:2'), covar=tensor([0.0686, 0.0184, 0.0199, 0.1105, 0.0062, 0.0136, 0.0341, 0.0318], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0141, 0.0070, 0.0106, 0.0121, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 13:13:42,857 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:14,518 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:23,335 INFO [train.py:904] (2/8) Epoch 12, batch 1100, loss[loss=0.1485, simple_loss=0.2244, pruned_loss=0.03637, over 16800.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2575, pruned_loss=0.04777, over 3283307.43 frames. ], batch size: 39, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:14:34,069 INFO [optim.py:368] (2/8) 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 13:14:48,134 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:15:33,616 INFO [train.py:904] (2/8) Epoch 12, batch 1150, loss[loss=0.163, simple_loss=0.2547, pruned_loss=0.03565, over 17315.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2578, pruned_loss=0.04721, over 3293714.50 frames. ], batch size: 52, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:15:33,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3059, 3.2466, 3.4920, 2.5937, 3.2253, 3.5569, 3.2373, 2.2021], device='cuda:2'), covar=tensor([0.0360, 0.0103, 0.0041, 0.0248, 0.0077, 0.0070, 0.0074, 0.0322], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0071, 0.0070, 0.0127, 0.0080, 0.0089, 0.0078, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 13:15:50,956 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 13:15:57,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8917, 4.1480, 2.5548, 4.6702, 3.0856, 4.6048, 2.5387, 3.3356], device='cuda:2'), covar=tensor([0.0219, 0.0292, 0.1364, 0.0209, 0.0693, 0.0388, 0.1437, 0.0589], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0168, 0.0191, 0.0136, 0.0168, 0.0210, 0.0200, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 13:16:23,375 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7750, 4.8001, 5.3212, 5.3225, 5.2743, 4.9290, 4.8819, 4.7100], device='cuda:2'), covar=tensor([0.0323, 0.0536, 0.0384, 0.0357, 0.0466, 0.0354, 0.0845, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0347, 0.0360, 0.0362, 0.0340, 0.0406, 0.0382, 0.0483, 0.0310], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 13:16:31,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1214, 2.9981, 3.1754, 2.2637, 3.0019, 3.2081, 3.0342, 1.9498], device='cuda:2'), covar=tensor([0.0398, 0.0119, 0.0055, 0.0319, 0.0085, 0.0105, 0.0091, 0.0379], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0071, 0.0071, 0.0128, 0.0080, 0.0090, 0.0079, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 13:16:42,798 INFO [train.py:904] (2/8) Epoch 12, batch 1200, loss[loss=0.2056, simple_loss=0.2943, pruned_loss=0.05841, over 16611.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2576, pruned_loss=0.04723, over 3290720.30 frames. ], batch size: 62, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:16:52,675 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 12, batch 1250, loss[loss=0.1585, simple_loss=0.24, pruned_loss=0.03844, over 15920.00 frames. ], tot_loss[loss=0.177, simple_loss=0.258, pruned_loss=0.04799, over 3296256.47 frames. ], batch size: 35, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:18:12,304 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:16,589 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:32,351 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:18:41,576 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 13:18:57,923 INFO [train.py:904] (2/8) Epoch 12, batch 1300, loss[loss=0.1981, simple_loss=0.2697, pruned_loss=0.06327, over 16716.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2578, pruned_loss=0.04781, over 3285058.88 frames. ], batch size: 124, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:19:09,589 INFO [optim.py:368] (2/8) 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 13:19:11,415 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1419, 2.0726, 2.2533, 3.5677, 2.1190, 2.3373, 2.2267, 2.1860], device='cuda:2'), covar=tensor([0.1114, 0.3278, 0.2435, 0.0557, 0.3651, 0.2338, 0.3195, 0.3266], device='cuda:2'), in_proj_covar=tensor([0.0369, 0.0397, 0.0336, 0.0324, 0.0416, 0.0455, 0.0361, 0.0465], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:19:27,112 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:19:41,011 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:20:08,552 INFO [train.py:904] (2/8) Epoch 12, batch 1350, loss[loss=0.1489, simple_loss=0.2391, pruned_loss=0.02932, over 17221.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2576, pruned_loss=0.04769, over 3302275.83 frames. ], batch size: 44, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:20:44,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3737, 2.2544, 1.8344, 2.0920, 2.5926, 2.3855, 2.5941, 2.7106], device='cuda:2'), covar=tensor([0.0185, 0.0264, 0.0361, 0.0329, 0.0166, 0.0243, 0.0176, 0.0185], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0213, 0.0205, 0.0204, 0.0212, 0.0212, 0.0217, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:21:07,387 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:21:16,369 INFO [train.py:904] (2/8) Epoch 12, batch 1400, loss[loss=0.1643, simple_loss=0.2373, pruned_loss=0.04564, over 16969.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2593, pruned_loss=0.04837, over 3310938.08 frames. ], batch size: 41, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:21:26,436 INFO [optim.py:368] (2/8) 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 13:21:40,587 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9715, 1.7977, 2.3466, 2.8181, 2.7822, 2.9377, 1.9206, 3.0391], device='cuda:2'), covar=tensor([0.0156, 0.0392, 0.0248, 0.0218, 0.0211, 0.0178, 0.0376, 0.0114], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0178, 0.0159, 0.0163, 0.0173, 0.0130, 0.0176, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 13:22:11,881 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 13:22:12,584 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:22:24,188 INFO [train.py:904] (2/8) Epoch 12, batch 1450, loss[loss=0.1789, simple_loss=0.2738, pruned_loss=0.04198, over 17026.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2582, pruned_loss=0.048, over 3308332.36 frames. ], batch size: 55, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:23:05,059 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8438, 4.9134, 5.4651, 5.3892, 5.4201, 5.0549, 5.0096, 4.7945], device='cuda:2'), covar=tensor([0.0298, 0.0478, 0.0350, 0.0463, 0.0382, 0.0318, 0.0747, 0.0362], device='cuda:2'), in_proj_covar=tensor([0.0346, 0.0360, 0.0362, 0.0340, 0.0404, 0.0382, 0.0481, 0.0309], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 13:23:35,102 INFO [train.py:904] (2/8) Epoch 12, batch 1500, loss[loss=0.182, simple_loss=0.2509, pruned_loss=0.05654, over 16424.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2576, pruned_loss=0.04747, over 3317004.34 frames. ], batch size: 146, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:23:45,775 INFO [optim.py:368] (2/8) 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 13:24:18,393 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 13:24:30,791 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 13:24:43,433 INFO [train.py:904] (2/8) Epoch 12, batch 1550, loss[loss=0.2212, simple_loss=0.2843, pruned_loss=0.07907, over 16824.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2583, pruned_loss=0.04798, over 3322706.43 frames. ], batch size: 90, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:25:06,541 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:27,366 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:28,548 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:36,391 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 13:25:54,140 INFO [train.py:904] (2/8) Epoch 12, batch 1600, loss[loss=0.2159, simple_loss=0.2977, pruned_loss=0.06709, over 15597.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2608, pruned_loss=0.04909, over 3308374.27 frames. ], batch size: 191, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:26:04,701 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:14,724 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8271, 2.8395, 2.4755, 4.5168, 3.5181, 4.2822, 1.6542, 3.0297], device='cuda:2'), covar=tensor([0.1358, 0.0705, 0.1242, 0.0174, 0.0255, 0.0377, 0.1449, 0.0815], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0159, 0.0180, 0.0148, 0.0194, 0.0211, 0.0181, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 13:26:22,738 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:26:27,894 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:33,951 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:53,571 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:59,811 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4361, 3.6115, 3.4856, 2.1349, 3.0456, 2.6581, 3.8753, 3.8007], device='cuda:2'), covar=tensor([0.0205, 0.0678, 0.0562, 0.1686, 0.0742, 0.0855, 0.0433, 0.0764], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0146, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 13:27:02,727 INFO [train.py:904] (2/8) Epoch 12, batch 1650, loss[loss=0.1999, simple_loss=0.2785, pruned_loss=0.06061, over 16541.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2625, pruned_loss=0.05012, over 3309698.43 frames. ], batch size: 75, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:27:29,755 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:28:11,022 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6435, 3.6993, 2.9008, 2.1932, 2.4657, 2.2011, 3.7024, 3.3101], device='cuda:2'), covar=tensor([0.2363, 0.0635, 0.1347, 0.2326, 0.2499, 0.1899, 0.0474, 0.1234], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0257, 0.0281, 0.0277, 0.0277, 0.0223, 0.0265, 0.0298], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:28:12,395 INFO [train.py:904] (2/8) Epoch 12, batch 1700, loss[loss=0.2067, simple_loss=0.2827, pruned_loss=0.06538, over 16720.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2649, pruned_loss=0.05154, over 3301755.57 frames. ], batch size: 134, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:28:23,616 INFO [optim.py:368] (2/8) 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 13:29:22,273 INFO [train.py:904] (2/8) Epoch 12, batch 1750, loss[loss=0.1785, simple_loss=0.2689, pruned_loss=0.04406, over 16829.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2666, pruned_loss=0.05214, over 3306312.38 frames. ], batch size: 42, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:29:37,726 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 13:30:15,835 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8268, 2.6965, 2.1921, 2.4899, 3.0131, 2.8651, 3.6043, 3.3016], device='cuda:2'), covar=tensor([0.0083, 0.0309, 0.0397, 0.0325, 0.0212, 0.0268, 0.0173, 0.0193], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0214, 0.0206, 0.0206, 0.0214, 0.0213, 0.0220, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:30:32,314 INFO [train.py:904] (2/8) Epoch 12, batch 1800, loss[loss=0.2001, simple_loss=0.2883, pruned_loss=0.05597, over 17132.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2685, pruned_loss=0.05205, over 3313166.20 frames. ], batch size: 47, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:30:43,442 INFO [optim.py:368] (2/8) 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 13:31:42,916 INFO [train.py:904] (2/8) Epoch 12, batch 1850, loss[loss=0.2046, simple_loss=0.281, pruned_loss=0.06405, over 16333.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2688, pruned_loss=0.05165, over 3314549.87 frames. ], batch size: 165, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:32:25,489 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7995, 4.2232, 4.3319, 1.8992, 4.5276, 4.7699, 3.2070, 3.4583], device='cuda:2'), covar=tensor([0.0895, 0.0147, 0.0197, 0.1302, 0.0085, 0.0086, 0.0417, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0141, 0.0071, 0.0108, 0.0122, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 13:32:53,584 INFO [train.py:904] (2/8) Epoch 12, batch 1900, loss[loss=0.1907, simple_loss=0.2815, pruned_loss=0.04995, over 16694.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2677, pruned_loss=0.05088, over 3320861.28 frames. ], batch size: 57, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:33:04,794 INFO [optim.py:368] (2/8) 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 13:33:29,754 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:33:49,006 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:05,626 INFO [train.py:904] (2/8) Epoch 12, batch 1950, loss[loss=0.2007, simple_loss=0.2722, pruned_loss=0.06461, over 16778.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2685, pruned_loss=0.05098, over 3306887.84 frames. ], batch size: 124, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:34:27,367 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:38,487 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:08,103 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:16,104 INFO [train.py:904] (2/8) Epoch 12, batch 2000, loss[loss=0.1843, simple_loss=0.2704, pruned_loss=0.04911, over 17045.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2669, pruned_loss=0.04994, over 3310627.05 frames. ], batch size: 53, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:35:27,899 INFO [optim.py:368] (2/8) 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 13:35:28,254 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5685, 4.9362, 4.6924, 4.6747, 4.4231, 4.4453, 4.4364, 4.9732], device='cuda:2'), covar=tensor([0.1191, 0.0825, 0.1044, 0.0683, 0.0910, 0.1076, 0.1058, 0.0854], device='cuda:2'), in_proj_covar=tensor([0.0577, 0.0716, 0.0584, 0.0503, 0.0449, 0.0462, 0.0598, 0.0548], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:35:52,978 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:36:04,256 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8274, 3.7434, 3.9041, 4.0093, 4.0496, 3.6272, 3.8622, 4.0914], device='cuda:2'), covar=tensor([0.1133, 0.0932, 0.1001, 0.0532, 0.0541, 0.2000, 0.1829, 0.0609], device='cuda:2'), in_proj_covar=tensor([0.0571, 0.0714, 0.0858, 0.0727, 0.0547, 0.0560, 0.0566, 0.0654], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:36:25,541 INFO [train.py:904] (2/8) Epoch 12, batch 2050, loss[loss=0.1862, simple_loss=0.2581, pruned_loss=0.05712, over 16727.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2672, pruned_loss=0.05061, over 3317299.21 frames. ], batch size: 89, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:36:31,269 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0889, 4.0424, 4.3552, 2.0960, 4.6074, 4.5912, 3.0562, 3.6358], device='cuda:2'), covar=tensor([0.0651, 0.0189, 0.0226, 0.1084, 0.0057, 0.0139, 0.0437, 0.0342], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0099, 0.0088, 0.0138, 0.0070, 0.0107, 0.0120, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 13:36:32,468 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:37:33,720 INFO [train.py:904] (2/8) Epoch 12, batch 2100, loss[loss=0.2149, simple_loss=0.303, pruned_loss=0.06335, over 16710.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2682, pruned_loss=0.05106, over 3328232.78 frames. ], batch size: 62, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:37:45,424 INFO [optim.py:368] (2/8) 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 13:38:44,791 INFO [train.py:904] (2/8) Epoch 12, batch 2150, loss[loss=0.2022, simple_loss=0.2707, pruned_loss=0.06681, over 16883.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2693, pruned_loss=0.05163, over 3318565.66 frames. ], batch size: 109, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:39:10,874 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3889, 3.9876, 3.4767, 1.7856, 2.8034, 2.2933, 3.9209, 3.9341], device='cuda:2'), covar=tensor([0.0322, 0.0694, 0.0743, 0.2346, 0.1108, 0.1207, 0.0634, 0.0987], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0148, 0.0160, 0.0146, 0.0137, 0.0125, 0.0139, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 13:39:54,111 INFO [train.py:904] (2/8) Epoch 12, batch 2200, loss[loss=0.2173, simple_loss=0.2806, pruned_loss=0.07704, over 16766.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.269, pruned_loss=0.05128, over 3330506.89 frames. ], batch size: 134, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:40:05,118 INFO [optim.py:368] (2/8) 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 13:40:48,888 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:41:03,694 INFO [train.py:904] (2/8) Epoch 12, batch 2250, loss[loss=0.2063, simple_loss=0.2784, pruned_loss=0.06707, over 16861.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2695, pruned_loss=0.05167, over 3326273.04 frames. ], batch size: 116, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:41:54,644 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:42:14,190 INFO [train.py:904] (2/8) Epoch 12, batch 2300, loss[loss=0.1946, simple_loss=0.2649, pruned_loss=0.06217, over 16876.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2702, pruned_loss=0.05208, over 3312250.81 frames. ], batch size: 116, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:42:24,215 INFO [optim.py:368] (2/8) 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 13:42:32,918 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8536, 2.0648, 2.2835, 3.0915, 2.1447, 2.2330, 2.2363, 2.1807], device='cuda:2'), covar=tensor([0.1065, 0.2773, 0.1877, 0.0581, 0.3283, 0.2102, 0.2622, 0.2917], device='cuda:2'), in_proj_covar=tensor([0.0371, 0.0398, 0.0336, 0.0324, 0.0414, 0.0459, 0.0363, 0.0467], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:42:43,123 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:25,573 INFO [train.py:904] (2/8) Epoch 12, batch 2350, loss[loss=0.2206, simple_loss=0.2947, pruned_loss=0.07328, over 16519.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2695, pruned_loss=0.05199, over 3307818.23 frames. ], batch size: 75, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:43:25,833 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:28,729 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 13:44:35,995 INFO [train.py:904] (2/8) Epoch 12, batch 2400, loss[loss=0.1769, simple_loss=0.2667, pruned_loss=0.04358, over 16876.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2701, pruned_loss=0.05209, over 3315863.56 frames. ], batch size: 96, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:44:48,485 INFO [optim.py:368] (2/8) 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 13:45:49,044 INFO [train.py:904] (2/8) Epoch 12, batch 2450, loss[loss=0.2088, simple_loss=0.2875, pruned_loss=0.06505, over 16253.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2714, pruned_loss=0.05239, over 3306510.94 frames. ], batch size: 165, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:46:23,308 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:46:29,782 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9493, 4.8367, 4.8587, 4.5883, 4.4909, 4.8716, 4.7476, 4.5593], device='cuda:2'), covar=tensor([0.0585, 0.0575, 0.0263, 0.0260, 0.0896, 0.0412, 0.0400, 0.0657], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0349, 0.0316, 0.0294, 0.0336, 0.0341, 0.0215, 0.0368], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 13:46:41,394 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6751, 3.8362, 2.1067, 4.3289, 2.8490, 4.3133, 2.3881, 2.9889], device='cuda:2'), covar=tensor([0.0242, 0.0341, 0.1548, 0.0236, 0.0763, 0.0503, 0.1495, 0.0710], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0169, 0.0190, 0.0140, 0.0170, 0.0213, 0.0199, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 13:46:57,384 INFO [train.py:904] (2/8) Epoch 12, batch 2500, loss[loss=0.1836, simple_loss=0.2598, pruned_loss=0.05369, over 16493.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2717, pruned_loss=0.05236, over 3308559.15 frames. ], batch size: 75, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:47:02,090 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7822, 2.9669, 2.7690, 4.8027, 3.6977, 4.4345, 1.6616, 3.0169], device='cuda:2'), covar=tensor([0.1404, 0.0725, 0.1075, 0.0201, 0.0334, 0.0393, 0.1485, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0160, 0.0181, 0.0151, 0.0197, 0.0212, 0.0181, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 13:47:09,683 INFO [optim.py:368] (2/8) 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 13:47:48,643 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:48:06,982 INFO [train.py:904] (2/8) Epoch 12, batch 2550, loss[loss=0.1812, simple_loss=0.2567, pruned_loss=0.0529, over 16747.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2713, pruned_loss=0.05197, over 3306361.97 frames. ], batch size: 89, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:48:19,530 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 13:49:15,389 INFO [train.py:904] (2/8) Epoch 12, batch 2600, loss[loss=0.1747, simple_loss=0.2713, pruned_loss=0.03898, over 17089.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2715, pruned_loss=0.05201, over 3316546.84 frames. ], batch size: 49, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:49:25,933 INFO [optim.py:368] (2/8) 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 13:49:45,739 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:24,314 INFO [train.py:904] (2/8) Epoch 12, batch 2650, loss[loss=0.1797, simple_loss=0.2706, pruned_loss=0.04442, over 17228.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2716, pruned_loss=0.0513, over 3324131.47 frames. ], batch size: 44, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:50:24,685 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:51,415 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:51:04,928 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-29 13:51:12,542 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 13:51:32,414 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 12, batch 2700, loss[loss=0.197, simple_loss=0.2723, pruned_loss=0.06087, over 12458.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.272, pruned_loss=0.05092, over 3322089.86 frames. ], batch size: 246, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:51:45,484 INFO [optim.py:368] (2/8) 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 13:52:01,033 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 13:52:41,338 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6948, 4.2765, 3.1067, 2.2175, 2.8145, 2.3135, 4.7139, 3.7965], device='cuda:2'), covar=tensor([0.2741, 0.0555, 0.1516, 0.2307, 0.2426, 0.1858, 0.0324, 0.1023], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0259, 0.0284, 0.0281, 0.0282, 0.0224, 0.0270, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:52:44,855 INFO [train.py:904] (2/8) Epoch 12, batch 2750, loss[loss=0.1795, simple_loss=0.2618, pruned_loss=0.04857, over 15994.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2709, pruned_loss=0.05003, over 3321576.06 frames. ], batch size: 35, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:53:05,714 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5351, 4.4538, 4.4272, 4.2130, 4.1310, 4.4865, 4.2706, 4.1845], device='cuda:2'), covar=tensor([0.0536, 0.0522, 0.0263, 0.0253, 0.0790, 0.0402, 0.0478, 0.0658], device='cuda:2'), in_proj_covar=tensor([0.0268, 0.0349, 0.0318, 0.0294, 0.0336, 0.0341, 0.0216, 0.0370], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 13:53:54,633 INFO [train.py:904] (2/8) Epoch 12, batch 2800, loss[loss=0.1814, simple_loss=0.2731, pruned_loss=0.04485, over 16669.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2706, pruned_loss=0.05004, over 3320776.65 frames. ], batch size: 57, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:54:06,072 INFO [optim.py:368] (2/8) 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 13:54:38,611 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:55:04,052 INFO [train.py:904] (2/8) Epoch 12, batch 2850, loss[loss=0.1733, simple_loss=0.2609, pruned_loss=0.04281, over 17030.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2704, pruned_loss=0.0499, over 3319601.36 frames. ], batch size: 55, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:55:38,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9466, 2.4943, 1.9675, 2.3450, 2.8978, 2.7281, 3.0402, 3.0403], device='cuda:2'), covar=tensor([0.0139, 0.0284, 0.0401, 0.0335, 0.0182, 0.0257, 0.0210, 0.0212], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0214, 0.0203, 0.0206, 0.0213, 0.0212, 0.0222, 0.0204], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:56:13,202 INFO [train.py:904] (2/8) Epoch 12, batch 2900, loss[loss=0.1693, simple_loss=0.2577, pruned_loss=0.04046, over 17099.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2692, pruned_loss=0.05019, over 3326887.19 frames. ], batch size: 53, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:56:24,536 INFO [optim.py:368] (2/8) 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 13:56:44,504 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2966, 3.7692, 3.9394, 2.1971, 3.2160, 2.6139, 3.8746, 3.8289], device='cuda:2'), covar=tensor([0.0270, 0.0695, 0.0441, 0.1654, 0.0688, 0.0844, 0.0584, 0.0919], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0148, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 13:57:11,079 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 13:57:20,462 INFO [train.py:904] (2/8) Epoch 12, batch 2950, loss[loss=0.1515, simple_loss=0.2405, pruned_loss=0.03123, over 16751.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2687, pruned_loss=0.05052, over 3330730.99 frames. ], batch size: 39, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:23,572 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1188, 5.5504, 5.7274, 5.4228, 5.5491, 6.0599, 5.5423, 5.3126], device='cuda:2'), covar=tensor([0.0739, 0.1804, 0.1791, 0.2017, 0.2378, 0.0830, 0.1373, 0.2294], device='cuda:2'), in_proj_covar=tensor([0.0360, 0.0517, 0.0560, 0.0440, 0.0596, 0.0587, 0.0446, 0.0596], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 13:58:25,983 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 13:58:26,800 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5216, 2.3480, 1.7756, 2.1792, 2.7221, 2.4703, 2.8026, 2.8665], device='cuda:2'), covar=tensor([0.0140, 0.0256, 0.0362, 0.0321, 0.0173, 0.0243, 0.0175, 0.0179], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0213, 0.0203, 0.0205, 0.0213, 0.0212, 0.0221, 0.0204], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 13:58:28,573 INFO [train.py:904] (2/8) Epoch 12, batch 3000, loss[loss=0.1798, simple_loss=0.2782, pruned_loss=0.04073, over 17057.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2683, pruned_loss=0.05021, over 3332969.97 frames. ], batch size: 50, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:28,573 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 13:58:38,465 INFO [train.py:938] (2/8) Epoch 12, validation: loss=0.14, simple_loss=0.2459, pruned_loss=0.01708, over 944034.00 frames. 2023-04-29 13:58:38,466 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 13:58:50,228 INFO [optim.py:368] (2/8) 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 13:59:48,674 INFO [train.py:904] (2/8) Epoch 12, batch 3050, loss[loss=0.1862, simple_loss=0.2646, pruned_loss=0.05392, over 16875.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.268, pruned_loss=0.05027, over 3329094.87 frames. ], batch size: 96, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 14:00:08,913 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:00:56,624 INFO [train.py:904] (2/8) Epoch 12, batch 3100, loss[loss=0.189, simple_loss=0.2587, pruned_loss=0.05967, over 16807.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2669, pruned_loss=0.05005, over 3329921.74 frames. ], batch size: 102, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:01:10,357 INFO [optim.py:368] (2/8) 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 14:01:31,895 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:01:40,468 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:02:05,275 INFO [train.py:904] (2/8) Epoch 12, batch 3150, loss[loss=0.1546, simple_loss=0.2413, pruned_loss=0.03398, over 16820.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2659, pruned_loss=0.04975, over 3331907.49 frames. ], batch size: 42, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:02:17,873 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3512, 2.4538, 2.0431, 2.2713, 2.8139, 2.5534, 3.1205, 3.0279], device='cuda:2'), covar=tensor([0.0144, 0.0326, 0.0395, 0.0334, 0.0200, 0.0292, 0.0217, 0.0186], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0211, 0.0203, 0.0205, 0.0213, 0.0213, 0.0221, 0.0204], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:02:30,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1138, 5.6464, 5.8685, 5.5145, 5.6030, 6.1611, 5.7023, 5.4089], device='cuda:2'), covar=tensor([0.0815, 0.1839, 0.1931, 0.1988, 0.2574, 0.0937, 0.1326, 0.2250], device='cuda:2'), in_proj_covar=tensor([0.0365, 0.0522, 0.0565, 0.0445, 0.0604, 0.0588, 0.0449, 0.0603], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:02:45,190 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:02:47,935 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 14:03:14,011 INFO [train.py:904] (2/8) Epoch 12, batch 3200, loss[loss=0.1991, simple_loss=0.2788, pruned_loss=0.05972, over 16298.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2665, pruned_loss=0.05003, over 3323417.88 frames. ], batch size: 165, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:03:26,055 INFO [optim.py:368] (2/8) 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 14:03:40,758 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:03:59,946 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0583, 5.2223, 4.9156, 4.6666, 4.0732, 5.1042, 5.1619, 4.5643], device='cuda:2'), covar=tensor([0.0777, 0.0495, 0.0487, 0.0371, 0.2131, 0.0490, 0.0365, 0.0884], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0353, 0.0324, 0.0299, 0.0342, 0.0346, 0.0218, 0.0376], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:04:22,406 INFO [train.py:904] (2/8) Epoch 12, batch 3250, loss[loss=0.1698, simple_loss=0.2626, pruned_loss=0.03847, over 16983.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.267, pruned_loss=0.05029, over 3316185.21 frames. ], batch size: 41, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:05,224 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:05:32,344 INFO [train.py:904] (2/8) Epoch 12, batch 3300, loss[loss=0.1454, simple_loss=0.2248, pruned_loss=0.03304, over 17270.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2679, pruned_loss=0.0509, over 3318422.94 frames. ], batch size: 45, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:45,364 INFO [optim.py:368] (2/8) 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 14:06:42,021 INFO [train.py:904] (2/8) Epoch 12, batch 3350, loss[loss=0.2309, simple_loss=0.3088, pruned_loss=0.07654, over 12127.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2681, pruned_loss=0.05103, over 3315077.85 frames. ], batch size: 246, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:07:13,242 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 14:07:50,786 INFO [train.py:904] (2/8) Epoch 12, batch 3400, loss[loss=0.2171, simple_loss=0.2843, pruned_loss=0.07491, over 16269.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2677, pruned_loss=0.05039, over 3325377.00 frames. ], batch size: 165, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:08:04,044 INFO [optim.py:368] (2/8) 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 14:08:12,783 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1224, 4.0203, 4.5285, 2.1102, 4.7476, 4.6925, 3.3762, 3.6178], device='cuda:2'), covar=tensor([0.0674, 0.0208, 0.0179, 0.1133, 0.0058, 0.0115, 0.0353, 0.0350], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0102, 0.0090, 0.0139, 0.0071, 0.0110, 0.0123, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 14:08:18,295 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:08:37,762 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:09:00,023 INFO [train.py:904] (2/8) Epoch 12, batch 3450, loss[loss=0.1603, simple_loss=0.2544, pruned_loss=0.03314, over 17188.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2674, pruned_loss=0.05035, over 3326571.64 frames. ], batch size: 44, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:09:15,900 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.73 vs. limit=5.0 2023-04-29 14:09:29,955 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7608, 3.8149, 2.9902, 2.2601, 2.5827, 2.2712, 3.7882, 3.4196], device='cuda:2'), covar=tensor([0.2298, 0.0589, 0.1283, 0.2403, 0.2331, 0.1798, 0.0458, 0.1246], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0259, 0.0285, 0.0281, 0.0284, 0.0224, 0.0269, 0.0304], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:10:01,662 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:10:08,329 INFO [train.py:904] (2/8) Epoch 12, batch 3500, loss[loss=0.1743, simple_loss=0.2624, pruned_loss=0.04313, over 16561.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2663, pruned_loss=0.04955, over 3314696.95 frames. ], batch size: 75, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:10:23,312 INFO [optim.py:368] (2/8) 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 14:11:19,452 INFO [train.py:904] (2/8) Epoch 12, batch 3550, loss[loss=0.1621, simple_loss=0.2522, pruned_loss=0.03607, over 17106.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2638, pruned_loss=0.04836, over 3318917.14 frames. ], batch size: 47, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:11:53,680 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:12:28,659 INFO [train.py:904] (2/8) Epoch 12, batch 3600, loss[loss=0.1623, simple_loss=0.2412, pruned_loss=0.04175, over 16784.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2618, pruned_loss=0.04797, over 3316394.14 frames. ], batch size: 83, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:12:43,127 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4650, 4.4196, 4.4192, 3.3868, 4.3840, 1.7247, 4.1244, 3.9502], device='cuda:2'), covar=tensor([0.0178, 0.0143, 0.0190, 0.0575, 0.0126, 0.2859, 0.0183, 0.0302], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0129, 0.0179, 0.0168, 0.0150, 0.0189, 0.0168, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:12:43,848 INFO [optim.py:368] (2/8) 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 14:13:04,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0458, 5.5368, 5.7276, 5.4183, 5.4396, 6.1016, 5.6718, 5.4388], device='cuda:2'), covar=tensor([0.0874, 0.1835, 0.1824, 0.1962, 0.2700, 0.0921, 0.1212, 0.2010], device='cuda:2'), in_proj_covar=tensor([0.0361, 0.0515, 0.0558, 0.0437, 0.0594, 0.0579, 0.0441, 0.0593], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:13:36,425 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6044, 3.6856, 3.9616, 1.9472, 4.1376, 4.0102, 3.1852, 3.0996], device='cuda:2'), covar=tensor([0.0693, 0.0177, 0.0167, 0.1095, 0.0050, 0.0135, 0.0328, 0.0345], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0101, 0.0089, 0.0138, 0.0070, 0.0110, 0.0122, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 14:13:40,320 INFO [train.py:904] (2/8) Epoch 12, batch 3650, loss[loss=0.1677, simple_loss=0.243, pruned_loss=0.04621, over 16792.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2611, pruned_loss=0.04794, over 3312753.52 frames. ], batch size: 124, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:14:55,143 INFO [train.py:904] (2/8) Epoch 12, batch 3700, loss[loss=0.1734, simple_loss=0.2602, pruned_loss=0.04331, over 17017.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2609, pruned_loss=0.04966, over 3295139.59 frames. ], batch size: 41, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:15:09,328 INFO [optim.py:368] (2/8) 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 14:15:22,901 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:24,733 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:25,797 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:10,075 INFO [train.py:904] (2/8) Epoch 12, batch 3750, loss[loss=0.1955, simple_loss=0.2683, pruned_loss=0.06133, over 16690.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2619, pruned_loss=0.05144, over 3281320.68 frames. ], batch size: 134, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:16:36,699 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:37,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5903, 2.3588, 2.2500, 3.3727, 2.5443, 3.5885, 1.4478, 2.6097], device='cuda:2'), covar=tensor([0.1477, 0.0785, 0.1190, 0.0176, 0.0205, 0.0397, 0.1652, 0.0903], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0162, 0.0182, 0.0155, 0.0202, 0.0215, 0.0184, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 14:16:47,370 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0494, 4.0726, 3.9168, 3.7271, 3.7260, 4.0160, 3.6759, 3.8096], device='cuda:2'), covar=tensor([0.0558, 0.0472, 0.0280, 0.0246, 0.0697, 0.0427, 0.0920, 0.0507], device='cuda:2'), in_proj_covar=tensor([0.0264, 0.0345, 0.0314, 0.0291, 0.0334, 0.0338, 0.0212, 0.0365], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:16:54,243 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:55,325 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:09,433 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:23,459 INFO [train.py:904] (2/8) Epoch 12, batch 3800, loss[loss=0.1937, simple_loss=0.2631, pruned_loss=0.06215, over 16461.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2641, pruned_loss=0.05323, over 3276768.96 frames. ], batch size: 68, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:17:38,965 INFO [optim.py:368] (2/8) 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 14:17:45,333 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1748, 3.2337, 3.4750, 2.4109, 3.1420, 3.5597, 3.3362, 1.9669], device='cuda:2'), covar=tensor([0.0402, 0.0083, 0.0043, 0.0298, 0.0087, 0.0073, 0.0064, 0.0380], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0071, 0.0070, 0.0124, 0.0080, 0.0090, 0.0080, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:18:34,729 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 14:18:37,586 INFO [train.py:904] (2/8) Epoch 12, batch 3850, loss[loss=0.1846, simple_loss=0.2581, pruned_loss=0.05553, over 16836.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2642, pruned_loss=0.05422, over 3267376.15 frames. ], batch size: 90, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:18:46,292 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5087, 4.5446, 4.6910, 4.5648, 4.5672, 5.1241, 4.6933, 4.4195], device='cuda:2'), covar=tensor([0.1418, 0.1682, 0.1595, 0.1956, 0.2760, 0.0998, 0.1387, 0.2345], device='cuda:2'), in_proj_covar=tensor([0.0362, 0.0517, 0.0561, 0.0438, 0.0592, 0.0579, 0.0444, 0.0595], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:19:04,468 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 14:19:16,991 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:19:26,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3662, 3.3391, 3.4235, 3.5351, 3.5789, 3.3030, 3.4756, 3.6375], device='cuda:2'), covar=tensor([0.1218, 0.0876, 0.1121, 0.0611, 0.0630, 0.2222, 0.1187, 0.0704], device='cuda:2'), in_proj_covar=tensor([0.0562, 0.0700, 0.0845, 0.0714, 0.0533, 0.0555, 0.0554, 0.0645], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:19:52,698 INFO [train.py:904] (2/8) Epoch 12, batch 3900, loss[loss=0.2038, simple_loss=0.2715, pruned_loss=0.06804, over 16787.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2636, pruned_loss=0.05481, over 3273877.01 frames. ], batch size: 134, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:19:57,188 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-29 14:19:59,767 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4237, 3.6817, 3.9354, 2.7149, 3.5709, 3.9470, 3.7154, 2.2013], device='cuda:2'), covar=tensor([0.0415, 0.0102, 0.0045, 0.0310, 0.0069, 0.0078, 0.0064, 0.0378], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0071, 0.0071, 0.0125, 0.0080, 0.0090, 0.0080, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:20:07,961 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:20:43,159 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-04-29 14:20:46,582 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4332, 4.3675, 4.2960, 4.1069, 4.0703, 4.3855, 4.0699, 4.1321], device='cuda:2'), covar=tensor([0.0554, 0.0568, 0.0287, 0.0260, 0.0764, 0.0441, 0.0653, 0.0593], device='cuda:2'), in_proj_covar=tensor([0.0261, 0.0340, 0.0310, 0.0288, 0.0330, 0.0333, 0.0209, 0.0361], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:21:03,112 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 14:21:08,905 INFO [train.py:904] (2/8) Epoch 12, batch 3950, loss[loss=0.2021, simple_loss=0.2701, pruned_loss=0.06707, over 16948.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2631, pruned_loss=0.05557, over 3272365.75 frames. ], batch size: 109, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:21:21,131 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2219, 4.1019, 4.2504, 4.3927, 4.4669, 4.0138, 4.2809, 4.4816], device='cuda:2'), covar=tensor([0.1178, 0.0972, 0.1206, 0.0568, 0.0590, 0.1333, 0.1554, 0.0620], device='cuda:2'), in_proj_covar=tensor([0.0560, 0.0701, 0.0843, 0.0712, 0.0534, 0.0557, 0.0553, 0.0645], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:21:50,897 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5550, 2.8996, 2.9781, 1.9196, 2.7001, 2.1743, 3.0676, 3.2068], device='cuda:2'), covar=tensor([0.0253, 0.0722, 0.0523, 0.1630, 0.0747, 0.0875, 0.0550, 0.0670], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 14:22:21,470 INFO [train.py:904] (2/8) Epoch 12, batch 4000, loss[loss=0.1883, simple_loss=0.2688, pruned_loss=0.05389, over 16308.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.263, pruned_loss=0.05591, over 3276996.91 frames. ], batch size: 165, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:22:34,737 INFO [optim.py:368] (2/8) 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 14:23:35,775 INFO [train.py:904] (2/8) Epoch 12, batch 4050, loss[loss=0.1931, simple_loss=0.2755, pruned_loss=0.05534, over 16430.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.263, pruned_loss=0.05461, over 3278006.96 frames. ], batch size: 146, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:24:12,118 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:13,876 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:35,759 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:50,501 INFO [train.py:904] (2/8) Epoch 12, batch 4100, loss[loss=0.2118, simple_loss=0.295, pruned_loss=0.06428, over 15409.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2645, pruned_loss=0.05381, over 3272406.09 frames. ], batch size: 191, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:25:05,531 INFO [optim.py:368] (2/8) 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 14:25:48,281 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:26:05,810 INFO [train.py:904] (2/8) Epoch 12, batch 4150, loss[loss=0.2127, simple_loss=0.3019, pruned_loss=0.06178, over 16371.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.272, pruned_loss=0.05655, over 3248959.77 frames. ], batch size: 146, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:21,077 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 14:27:23,036 INFO [train.py:904] (2/8) Epoch 12, batch 4200, loss[loss=0.2185, simple_loss=0.3112, pruned_loss=0.06292, over 16810.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2791, pruned_loss=0.0585, over 3211009.76 frames. ], batch size: 83, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:37,143 INFO [optim.py:368] (2/8) 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 14:28:22,419 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:28:36,619 INFO [train.py:904] (2/8) Epoch 12, batch 4250, loss[loss=0.1765, simple_loss=0.2756, pruned_loss=0.03873, over 16877.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2816, pruned_loss=0.05794, over 3199235.67 frames. ], batch size: 96, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:28:51,752 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 14:29:10,830 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1453, 2.0895, 2.1318, 3.7532, 2.0176, 2.4902, 2.1964, 2.2403], device='cuda:2'), covar=tensor([0.1047, 0.3278, 0.2356, 0.0449, 0.3701, 0.2248, 0.2942, 0.3143], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0402, 0.0336, 0.0324, 0.0415, 0.0463, 0.0365, 0.0470], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:29:13,840 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 14:29:49,145 INFO [train.py:904] (2/8) Epoch 12, batch 4300, loss[loss=0.2107, simple_loss=0.2977, pruned_loss=0.06185, over 16692.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2832, pruned_loss=0.05727, over 3191982.76 frames. ], batch size: 57, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:29:50,865 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:30:00,184 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5061, 4.4744, 4.4383, 2.7484, 3.9965, 4.4312, 3.9434, 2.1898], device='cuda:2'), covar=tensor([0.0454, 0.0018, 0.0030, 0.0355, 0.0055, 0.0058, 0.0055, 0.0407], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0071, 0.0072, 0.0127, 0.0081, 0.0090, 0.0081, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:30:03,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5220, 3.5127, 3.4730, 2.9023, 3.3908, 2.0948, 3.2022, 2.7113], device='cuda:2'), covar=tensor([0.0122, 0.0107, 0.0137, 0.0192, 0.0093, 0.2094, 0.0111, 0.0191], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0122, 0.0169, 0.0160, 0.0142, 0.0182, 0.0159, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:30:04,741 INFO [optim.py:368] (2/8) 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 14:31:07,414 INFO [train.py:904] (2/8) Epoch 12, batch 4350, loss[loss=0.2119, simple_loss=0.2919, pruned_loss=0.06594, over 16404.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.286, pruned_loss=0.05787, over 3202083.06 frames. ], batch size: 68, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:31:10,215 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3767, 5.3944, 5.2572, 5.0328, 4.8665, 5.2788, 5.1904, 4.9533], device='cuda:2'), covar=tensor([0.0512, 0.0193, 0.0211, 0.0189, 0.0868, 0.0277, 0.0211, 0.0468], device='cuda:2'), in_proj_covar=tensor([0.0250, 0.0326, 0.0299, 0.0276, 0.0317, 0.0318, 0.0202, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:31:27,383 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:43,175 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5717, 4.4140, 4.6615, 4.8184, 4.9566, 4.4526, 4.9188, 4.9624], device='cuda:2'), covar=tensor([0.1454, 0.1117, 0.1415, 0.0575, 0.0449, 0.0836, 0.0517, 0.0494], device='cuda:2'), in_proj_covar=tensor([0.0535, 0.0669, 0.0796, 0.0679, 0.0508, 0.0528, 0.0528, 0.0613], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:31:46,028 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:46,146 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4270, 4.1719, 4.2348, 2.6257, 3.7316, 4.2112, 3.6834, 2.4406], device='cuda:2'), covar=tensor([0.0407, 0.0020, 0.0023, 0.0325, 0.0058, 0.0057, 0.0059, 0.0303], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0071, 0.0072, 0.0127, 0.0081, 0.0090, 0.0081, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:31:47,980 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:22,077 INFO [train.py:904] (2/8) Epoch 12, batch 4400, loss[loss=0.2419, simple_loss=0.3215, pruned_loss=0.08109, over 17065.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.288, pruned_loss=0.05886, over 3199174.45 frames. ], batch size: 53, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:32:37,554 INFO [optim.py:368] (2/8) 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 14:32:54,755 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:57,360 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:57,520 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:33:35,426 INFO [train.py:904] (2/8) Epoch 12, batch 4450, loss[loss=0.2148, simple_loss=0.302, pruned_loss=0.06375, over 16517.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2919, pruned_loss=0.06038, over 3198856.70 frames. ], batch size: 68, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:33:40,121 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:33:57,032 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:34:49,327 INFO [train.py:904] (2/8) Epoch 12, batch 4500, loss[loss=0.2281, simple_loss=0.3108, pruned_loss=0.0727, over 16311.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2924, pruned_loss=0.0609, over 3204256.54 frames. ], batch size: 165, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:35:03,473 INFO [optim.py:368] (2/8) 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 14:35:10,231 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:35:25,613 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:35:34,740 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-29 14:36:02,108 INFO [train.py:904] (2/8) Epoch 12, batch 4550, loss[loss=0.2102, simple_loss=0.2915, pruned_loss=0.06449, over 16283.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2922, pruned_loss=0.06137, over 3197051.37 frames. ], batch size: 165, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:36:26,517 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3916, 2.9622, 2.9892, 1.8577, 2.6171, 2.1964, 2.9907, 3.1653], device='cuda:2'), covar=tensor([0.0295, 0.0669, 0.0591, 0.1873, 0.0850, 0.0969, 0.0641, 0.0715], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 14:37:08,828 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:37:11,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0403, 3.7645, 3.6875, 2.3122, 3.3969, 3.7563, 3.4993, 2.1071], device='cuda:2'), covar=tensor([0.0487, 0.0028, 0.0030, 0.0384, 0.0063, 0.0065, 0.0053, 0.0368], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0069, 0.0070, 0.0125, 0.0080, 0.0089, 0.0079, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:37:14,080 INFO [train.py:904] (2/8) Epoch 12, batch 4600, loss[loss=0.2068, simple_loss=0.2956, pruned_loss=0.05902, over 16338.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2933, pruned_loss=0.06186, over 3200013.52 frames. ], batch size: 165, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:37:28,833 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7048, 1.7290, 1.5499, 1.4775, 1.8569, 1.5572, 1.6489, 1.9157], device='cuda:2'), covar=tensor([0.0120, 0.0182, 0.0310, 0.0252, 0.0138, 0.0204, 0.0126, 0.0141], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0205, 0.0202, 0.0200, 0.0206, 0.0206, 0.0212, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:37:29,429 INFO [optim.py:368] (2/8) 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 14:38:26,067 INFO [train.py:904] (2/8) Epoch 12, batch 4650, loss[loss=0.1973, simple_loss=0.2795, pruned_loss=0.05755, over 17051.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.293, pruned_loss=0.06185, over 3208946.99 frames. ], batch size: 55, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:38:55,848 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:38:58,792 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-04-29 14:39:38,311 INFO [train.py:904] (2/8) Epoch 12, batch 4700, loss[loss=0.2178, simple_loss=0.297, pruned_loss=0.06929, over 12031.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2901, pruned_loss=0.06049, over 3217256.15 frames. ], batch size: 248, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:39:53,945 INFO [optim.py:368] (2/8) 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 14:40:05,844 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:40:27,456 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:40:36,708 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5681, 2.6240, 1.7426, 2.7632, 2.1776, 2.8066, 1.9912, 2.3522], device='cuda:2'), covar=tensor([0.0255, 0.0377, 0.1327, 0.0159, 0.0666, 0.0476, 0.1264, 0.0631], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0165, 0.0189, 0.0133, 0.0165, 0.0207, 0.0196, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 14:40:53,366 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4878, 4.3030, 4.5201, 4.7237, 4.8647, 4.4054, 4.8130, 4.8744], device='cuda:2'), covar=tensor([0.1323, 0.1051, 0.1464, 0.0628, 0.0452, 0.0871, 0.0537, 0.0448], device='cuda:2'), in_proj_covar=tensor([0.0527, 0.0660, 0.0790, 0.0676, 0.0502, 0.0521, 0.0523, 0.0607], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:40:54,109 INFO [train.py:904] (2/8) Epoch 12, batch 4750, loss[loss=0.1685, simple_loss=0.2577, pruned_loss=0.03971, over 16716.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2855, pruned_loss=0.05815, over 3216670.76 frames. ], batch size: 83, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:41:29,741 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 14:41:53,534 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-29 14:41:58,936 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:05,727 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 14:42:07,121 INFO [train.py:904] (2/8) Epoch 12, batch 4800, loss[loss=0.185, simple_loss=0.2861, pruned_loss=0.04194, over 16874.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2828, pruned_loss=0.05647, over 3203877.63 frames. ], batch size: 96, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:42:22,124 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:23,017 INFO [optim.py:368] (2/8) 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 14:42:39,548 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:48,694 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-29 14:43:23,406 INFO [train.py:904] (2/8) Epoch 12, batch 4850, loss[loss=0.2461, simple_loss=0.3125, pruned_loss=0.08987, over 12115.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2833, pruned_loss=0.05589, over 3179581.46 frames. ], batch size: 247, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:43:28,245 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:43:30,805 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:43:49,109 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8959, 5.2113, 4.9881, 5.0030, 4.7361, 4.7129, 4.6342, 5.2824], device='cuda:2'), covar=tensor([0.0959, 0.0772, 0.0876, 0.0610, 0.0695, 0.0764, 0.0933, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0538, 0.0669, 0.0554, 0.0472, 0.0426, 0.0434, 0.0557, 0.0524], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:44:31,786 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:44:38,142 INFO [train.py:904] (2/8) Epoch 12, batch 4900, loss[loss=0.1833, simple_loss=0.2727, pruned_loss=0.0469, over 16665.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2825, pruned_loss=0.05469, over 3163306.45 frames. ], batch size: 134, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:44:52,629 INFO [optim.py:368] (2/8) 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 14:44:58,497 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:45:42,831 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:45:52,019 INFO [train.py:904] (2/8) Epoch 12, batch 4950, loss[loss=0.2005, simple_loss=0.2814, pruned_loss=0.05979, over 17061.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2831, pruned_loss=0.05451, over 3149458.30 frames. ], batch size: 55, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:46:16,314 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:47:04,268 INFO [train.py:904] (2/8) Epoch 12, batch 5000, loss[loss=0.2286, simple_loss=0.3104, pruned_loss=0.07336, over 12412.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.284, pruned_loss=0.05429, over 3169294.02 frames. ], batch size: 247, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:47:17,032 INFO [optim.py:368] (2/8) 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 14:47:30,320 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:47:43,239 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:47:43,385 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:15,687 INFO [train.py:904] (2/8) Epoch 12, batch 5050, loss[loss=0.192, simple_loss=0.2808, pruned_loss=0.05156, over 15341.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2845, pruned_loss=0.0541, over 3178563.56 frames. ], batch size: 190, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:48:18,345 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:37,163 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6057, 4.3762, 4.2958, 2.9123, 3.6624, 4.2617, 3.7734, 2.4565], device='cuda:2'), covar=tensor([0.0389, 0.0019, 0.0026, 0.0288, 0.0072, 0.0069, 0.0070, 0.0330], device='cuda:2'), in_proj_covar=tensor([0.0125, 0.0068, 0.0069, 0.0124, 0.0079, 0.0088, 0.0078, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 14:48:38,165 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:17,705 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-29 14:49:24,608 INFO [train.py:904] (2/8) Epoch 12, batch 5100, loss[loss=0.1731, simple_loss=0.2604, pruned_loss=0.04295, over 16533.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2824, pruned_loss=0.05317, over 3188935.51 frames. ], batch size: 68, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:49:37,395 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 14:49:43,703 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:52,535 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:35,774 INFO [train.py:904] (2/8) Epoch 12, batch 5150, loss[loss=0.1956, simple_loss=0.2922, pruned_loss=0.04944, over 16808.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.282, pruned_loss=0.05256, over 3184751.12 frames. ], batch size: 83, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:50:36,702 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:50:36,836 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:47,784 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:03,103 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:47,912 INFO [train.py:904] (2/8) Epoch 12, batch 5200, loss[loss=0.1779, simple_loss=0.267, pruned_loss=0.04438, over 16872.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2808, pruned_loss=0.05206, over 3177413.27 frames. ], batch size: 96, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:52:00,670 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:01,658 INFO [optim.py:368] (2/8) 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 14:52:03,391 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:58,580 INFO [train.py:904] (2/8) Epoch 12, batch 5250, loss[loss=0.2033, simple_loss=0.2881, pruned_loss=0.05927, over 15399.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2784, pruned_loss=0.05169, over 3185714.13 frames. ], batch size: 191, lr: 5.69e-03, grad_scale: 16.0 2023-04-29 14:53:35,260 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 14:54:11,425 INFO [train.py:904] (2/8) Epoch 12, batch 5300, loss[loss=0.2088, simple_loss=0.2893, pruned_loss=0.06415, over 12339.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2751, pruned_loss=0.05021, over 3190523.18 frames. ], batch size: 248, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:54:27,261 INFO [optim.py:368] (2/8) 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 14:54:41,976 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:54:45,079 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7539, 4.7067, 4.5558, 3.2963, 4.5735, 1.4860, 4.2627, 4.2875], device='cuda:2'), covar=tensor([0.0141, 0.0103, 0.0186, 0.0801, 0.0145, 0.3337, 0.0180, 0.0304], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0120, 0.0164, 0.0157, 0.0137, 0.0179, 0.0154, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:54:49,746 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:55:21,985 INFO [train.py:904] (2/8) Epoch 12, batch 5350, loss[loss=0.1983, simple_loss=0.276, pruned_loss=0.06033, over 12033.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2733, pruned_loss=0.0496, over 3195814.94 frames. ], batch size: 246, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:55:23,842 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3118, 1.5002, 1.9248, 2.2736, 2.3517, 2.5802, 1.7028, 2.4307], device='cuda:2'), covar=tensor([0.0163, 0.0412, 0.0226, 0.0233, 0.0226, 0.0131, 0.0355, 0.0098], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0172, 0.0154, 0.0160, 0.0169, 0.0125, 0.0172, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 14:55:58,443 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:31,809 INFO [train.py:904] (2/8) Epoch 12, batch 5400, loss[loss=0.1812, simple_loss=0.2775, pruned_loss=0.04244, over 16751.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2767, pruned_loss=0.05058, over 3195565.78 frames. ], batch size: 83, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:56:43,927 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:49,067 INFO [optim.py:368] (2/8) 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 14:57:46,200 INFO [train.py:904] (2/8) Epoch 12, batch 5450, loss[loss=0.211, simple_loss=0.293, pruned_loss=0.06446, over 16609.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2801, pruned_loss=0.05244, over 3168658.46 frames. ], batch size: 57, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:57:46,726 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:58:21,718 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1530, 1.8720, 2.0252, 3.4953, 1.7420, 2.2295, 2.0345, 1.9840], device='cuda:2'), covar=tensor([0.1174, 0.3919, 0.2562, 0.0636, 0.4812, 0.2614, 0.3343, 0.3886], device='cuda:2'), in_proj_covar=tensor([0.0364, 0.0393, 0.0330, 0.0316, 0.0410, 0.0453, 0.0358, 0.0458], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 14:58:57,078 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:00,351 INFO [train.py:904] (2/8) Epoch 12, batch 5500, loss[loss=0.2519, simple_loss=0.3394, pruned_loss=0.08221, over 16335.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2877, pruned_loss=0.05757, over 3112707.24 frames. ], batch size: 165, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:59:09,375 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:13,923 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 15:00:18,011 INFO [train.py:904] (2/8) Epoch 12, batch 5550, loss[loss=0.2932, simple_loss=0.3466, pruned_loss=0.1199, over 11407.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2952, pruned_loss=0.06278, over 3118312.01 frames. ], batch size: 246, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:00:22,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1964, 4.0201, 4.2308, 4.4152, 4.4966, 4.0546, 4.4646, 4.4999], device='cuda:2'), covar=tensor([0.1497, 0.1143, 0.1436, 0.0561, 0.0563, 0.1244, 0.0615, 0.0559], device='cuda:2'), in_proj_covar=tensor([0.0527, 0.0663, 0.0795, 0.0671, 0.0504, 0.0520, 0.0525, 0.0609], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:00:24,330 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 15:00:30,330 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:49,815 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:55,170 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-29 15:01:00,883 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 15:01:39,143 INFO [train.py:904] (2/8) Epoch 12, batch 5600, loss[loss=0.259, simple_loss=0.3367, pruned_loss=0.09059, over 16915.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.3004, pruned_loss=0.06743, over 3096492.82 frames. ], batch size: 109, lr: 5.68e-03, grad_scale: 8.0 2023-04-29 15:01:58,900 INFO [optim.py:368] (2/8) 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 15:02:16,788 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:02:30,336 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:02,066 INFO [train.py:904] (2/8) Epoch 12, batch 5650, loss[loss=0.2117, simple_loss=0.302, pruned_loss=0.06065, over 16875.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3057, pruned_loss=0.07203, over 3065197.85 frames. ], batch size: 96, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:03:33,933 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:45,053 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:50,689 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7494, 3.3093, 3.3020, 1.9313, 2.8632, 2.0783, 3.3505, 3.4561], device='cuda:2'), covar=tensor([0.0239, 0.0591, 0.0539, 0.1874, 0.0772, 0.1014, 0.0573, 0.0803], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0146, 0.0159, 0.0142, 0.0136, 0.0123, 0.0137, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 15:04:18,926 INFO [train.py:904] (2/8) Epoch 12, batch 5700, loss[loss=0.2914, simple_loss=0.3449, pruned_loss=0.119, over 11375.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3082, pruned_loss=0.07418, over 3057611.69 frames. ], batch size: 246, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:04:32,818 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:04:41,588 INFO [optim.py:368] (2/8) 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 15:05:10,004 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6634, 3.2599, 3.2352, 1.8228, 2.7660, 2.0340, 3.1659, 3.3765], device='cuda:2'), covar=tensor([0.0297, 0.0653, 0.0533, 0.2013, 0.0801, 0.1034, 0.0735, 0.0977], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0147, 0.0159, 0.0143, 0.0137, 0.0124, 0.0138, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 15:05:21,407 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:05:39,176 INFO [train.py:904] (2/8) Epoch 12, batch 5750, loss[loss=0.2282, simple_loss=0.3096, pruned_loss=0.07333, over 16352.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3117, pruned_loss=0.07653, over 3033603.25 frames. ], batch size: 165, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:05:48,658 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:05:57,336 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-04-29 15:07:00,149 INFO [train.py:904] (2/8) Epoch 12, batch 5800, loss[loss=0.2015, simple_loss=0.2954, pruned_loss=0.05381, over 15538.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3111, pruned_loss=0.07514, over 3034242.30 frames. ], batch size: 191, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:07:02,047 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3043, 1.5601, 1.8867, 2.1692, 2.3083, 2.4568, 1.6097, 2.3424], device='cuda:2'), covar=tensor([0.0165, 0.0392, 0.0219, 0.0243, 0.0207, 0.0132, 0.0376, 0.0104], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0169, 0.0152, 0.0156, 0.0166, 0.0124, 0.0170, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 15:07:04,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8209, 1.3874, 1.6642, 1.6310, 1.8124, 1.8639, 1.5889, 1.7281], device='cuda:2'), covar=tensor([0.0169, 0.0262, 0.0141, 0.0197, 0.0183, 0.0130, 0.0276, 0.0084], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0169, 0.0152, 0.0156, 0.0166, 0.0124, 0.0170, 0.0116], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 15:07:09,837 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:07:21,330 INFO [optim.py:368] (2/8) 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 15:08:16,561 INFO [train.py:904] (2/8) Epoch 12, batch 5850, loss[loss=0.193, simple_loss=0.2854, pruned_loss=0.05027, over 16852.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.308, pruned_loss=0.07282, over 3034707.07 frames. ], batch size: 83, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:08:24,194 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:09:27,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1061, 2.0445, 2.1704, 3.6945, 2.0183, 2.4475, 2.1567, 2.2040], device='cuda:2'), covar=tensor([0.1061, 0.3175, 0.2212, 0.0445, 0.3748, 0.2056, 0.2930, 0.2895], device='cuda:2'), in_proj_covar=tensor([0.0364, 0.0393, 0.0329, 0.0315, 0.0410, 0.0451, 0.0358, 0.0458], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:09:37,155 INFO [train.py:904] (2/8) Epoch 12, batch 5900, loss[loss=0.2352, simple_loss=0.321, pruned_loss=0.07473, over 17160.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3069, pruned_loss=0.07236, over 3045673.32 frames. ], batch size: 46, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:10:01,560 INFO [optim.py:368] (2/8) 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 15:10:18,435 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:10:55,831 INFO [train.py:904] (2/8) Epoch 12, batch 5950, loss[loss=0.224, simple_loss=0.3036, pruned_loss=0.07222, over 16716.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3079, pruned_loss=0.07073, over 3066083.61 frames. ], batch size: 124, lr: 5.67e-03, grad_scale: 2.0 2023-04-29 15:11:40,335 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1076, 1.8738, 2.1609, 3.6309, 1.8620, 2.1942, 2.0067, 2.0132], device='cuda:2'), covar=tensor([0.1205, 0.4007, 0.2484, 0.0590, 0.4710, 0.2727, 0.3572, 0.3808], device='cuda:2'), in_proj_covar=tensor([0.0366, 0.0395, 0.0331, 0.0317, 0.0413, 0.0455, 0.0360, 0.0460], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:12:14,104 INFO [train.py:904] (2/8) Epoch 12, batch 6000, loss[loss=0.2072, simple_loss=0.2947, pruned_loss=0.05992, over 16578.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3068, pruned_loss=0.07018, over 3077937.96 frames. ], batch size: 75, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:12:14,104 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 15:12:25,313 INFO [train.py:938] (2/8) Epoch 12, validation: loss=0.161, simple_loss=0.2739, pruned_loss=0.02405, over 944034.00 frames. 2023-04-29 15:12:25,314 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 15:12:26,041 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 15:12:46,506 INFO [optim.py:368] (2/8) 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 15:13:18,134 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:13:46,105 INFO [train.py:904] (2/8) Epoch 12, batch 6050, loss[loss=0.2188, simple_loss=0.3067, pruned_loss=0.06543, over 16910.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3047, pruned_loss=0.06923, over 3092382.29 frames. ], batch size: 109, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:15:02,175 INFO [train.py:904] (2/8) Epoch 12, batch 6100, loss[loss=0.2107, simple_loss=0.2941, pruned_loss=0.06371, over 16284.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3036, pruned_loss=0.06805, over 3102492.22 frames. ], batch size: 165, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:15:24,819 INFO [optim.py:368] (2/8) 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 15:15:56,694 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8092, 4.9371, 5.3712, 5.3242, 5.3353, 5.0076, 4.9718, 4.6431], device='cuda:2'), covar=tensor([0.0305, 0.0427, 0.0276, 0.0360, 0.0398, 0.0278, 0.0783, 0.0395], device='cuda:2'), in_proj_covar=tensor([0.0345, 0.0356, 0.0359, 0.0340, 0.0409, 0.0382, 0.0483, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 15:16:04,998 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-29 15:16:19,683 INFO [train.py:904] (2/8) Epoch 12, batch 6150, loss[loss=0.1982, simple_loss=0.2897, pruned_loss=0.05335, over 16877.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3011, pruned_loss=0.06693, over 3110416.85 frames. ], batch size: 90, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:17:19,819 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0717, 5.6908, 5.8303, 5.6119, 5.6563, 6.1510, 5.6466, 5.4092], device='cuda:2'), covar=tensor([0.0870, 0.1542, 0.1705, 0.1730, 0.2176, 0.0925, 0.1484, 0.2499], device='cuda:2'), in_proj_covar=tensor([0.0355, 0.0492, 0.0542, 0.0422, 0.0580, 0.0565, 0.0429, 0.0581], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 15:17:38,950 INFO [train.py:904] (2/8) Epoch 12, batch 6200, loss[loss=0.238, simple_loss=0.3143, pruned_loss=0.08081, over 11680.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2994, pruned_loss=0.06671, over 3101106.64 frames. ], batch size: 248, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:18:00,667 INFO [optim.py:368] (2/8) 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 15:18:17,965 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:18:52,875 INFO [train.py:904] (2/8) Epoch 12, batch 6250, loss[loss=0.1911, simple_loss=0.2778, pruned_loss=0.05214, over 16794.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2996, pruned_loss=0.06685, over 3098668.37 frames. ], batch size: 83, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:19:27,990 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:20:06,464 INFO [train.py:904] (2/8) Epoch 12, batch 6300, loss[loss=0.1866, simple_loss=0.2738, pruned_loss=0.0497, over 16222.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3002, pruned_loss=0.06695, over 3087723.59 frames. ], batch size: 165, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:20:28,834 INFO [optim.py:368] (2/8) 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 15:20:57,583 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:21:25,214 INFO [train.py:904] (2/8) Epoch 12, batch 6350, loss[loss=0.1916, simple_loss=0.2841, pruned_loss=0.04951, over 16885.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3006, pruned_loss=0.06759, over 3102615.86 frames. ], batch size: 96, lr: 5.66e-03, grad_scale: 4.0 2023-04-29 15:22:11,806 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:21,808 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:37,545 INFO [train.py:904] (2/8) Epoch 12, batch 6400, loss[loss=0.2256, simple_loss=0.3105, pruned_loss=0.07035, over 15182.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3009, pruned_loss=0.06836, over 3107968.99 frames. ], batch size: 190, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:22:57,666 INFO [optim.py:368] (2/8) 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 15:23:49,709 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:23:50,964 INFO [train.py:904] (2/8) Epoch 12, batch 6450, loss[loss=0.1994, simple_loss=0.2825, pruned_loss=0.05814, over 17063.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3009, pruned_loss=0.0676, over 3106480.90 frames. ], batch size: 53, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:02,598 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6937, 2.6949, 2.2070, 2.5206, 3.0825, 2.7654, 3.4666, 3.3523], device='cuda:2'), covar=tensor([0.0058, 0.0275, 0.0383, 0.0322, 0.0186, 0.0275, 0.0147, 0.0156], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0201, 0.0198, 0.0197, 0.0201, 0.0200, 0.0208, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:25:08,003 INFO [train.py:904] (2/8) Epoch 12, batch 6500, loss[loss=0.1984, simple_loss=0.2836, pruned_loss=0.05658, over 16742.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2987, pruned_loss=0.06668, over 3101977.21 frames. ], batch size: 89, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:29,348 INFO [optim.py:368] (2/8) 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 15:25:32,634 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:26:28,542 INFO [train.py:904] (2/8) Epoch 12, batch 6550, loss[loss=0.2354, simple_loss=0.3355, pruned_loss=0.06764, over 16546.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3019, pruned_loss=0.06773, over 3110308.25 frames. ], batch size: 68, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:27:10,831 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 15:27:44,402 INFO [train.py:904] (2/8) Epoch 12, batch 6600, loss[loss=0.2212, simple_loss=0.3053, pruned_loss=0.06851, over 16486.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3038, pruned_loss=0.06837, over 3105769.87 frames. ], batch size: 68, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:28:05,473 INFO [optim.py:368] (2/8) 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 15:28:10,153 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:28:41,336 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:01,033 INFO [train.py:904] (2/8) Epoch 12, batch 6650, loss[loss=0.2588, simple_loss=0.3241, pruned_loss=0.09673, over 11542.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3048, pruned_loss=0.07011, over 3074505.59 frames. ], batch size: 247, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:29:43,612 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:45,565 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:16,322 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:17,398 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2045, 4.2778, 4.4346, 4.3140, 4.3023, 4.8216, 4.4018, 4.1484], device='cuda:2'), covar=tensor([0.1589, 0.1720, 0.1827, 0.1923, 0.2611, 0.1056, 0.1516, 0.2607], device='cuda:2'), in_proj_covar=tensor([0.0359, 0.0498, 0.0548, 0.0432, 0.0586, 0.0572, 0.0434, 0.0588], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 15:30:18,844 INFO [train.py:904] (2/8) Epoch 12, batch 6700, loss[loss=0.219, simple_loss=0.3, pruned_loss=0.06899, over 15514.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3036, pruned_loss=0.07, over 3077315.22 frames. ], batch size: 191, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:30:39,918 INFO [optim.py:368] (2/8) 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 15:31:14,320 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3965, 5.7741, 5.4125, 5.4405, 5.0479, 5.1330, 5.1289, 5.8461], device='cuda:2'), covar=tensor([0.1105, 0.0752, 0.0999, 0.0737, 0.0852, 0.0671, 0.1054, 0.0883], device='cuda:2'), in_proj_covar=tensor([0.0557, 0.0689, 0.0568, 0.0487, 0.0438, 0.0452, 0.0573, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:31:17,440 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:27,189 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:35,205 INFO [train.py:904] (2/8) Epoch 12, batch 6750, loss[loss=0.2121, simple_loss=0.291, pruned_loss=0.06661, over 17116.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3007, pruned_loss=0.06878, over 3106409.22 frames. ], batch size: 47, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:31:56,870 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2203, 2.1299, 2.1821, 4.0498, 2.0152, 2.4856, 2.1640, 2.3076], device='cuda:2'), covar=tensor([0.1116, 0.3177, 0.2471, 0.0403, 0.3897, 0.2331, 0.3117, 0.3000], device='cuda:2'), in_proj_covar=tensor([0.0365, 0.0394, 0.0332, 0.0316, 0.0414, 0.0455, 0.0362, 0.0461], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:32:49,915 INFO [train.py:904] (2/8) Epoch 12, batch 6800, loss[loss=0.225, simple_loss=0.3066, pruned_loss=0.07167, over 15463.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3004, pruned_loss=0.0685, over 3107160.46 frames. ], batch size: 190, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:33:11,648 INFO [optim.py:368] (2/8) 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 15:34:04,797 INFO [train.py:904] (2/8) Epoch 12, batch 6850, loss[loss=0.2077, simple_loss=0.3188, pruned_loss=0.04827, over 16785.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3014, pruned_loss=0.06891, over 3099675.41 frames. ], batch size: 102, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:34:36,242 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 15:35:15,879 INFO [train.py:904] (2/8) Epoch 12, batch 6900, loss[loss=0.2109, simple_loss=0.3035, pruned_loss=0.0592, over 17100.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3039, pruned_loss=0.06854, over 3114046.62 frames. ], batch size: 49, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:35:36,843 INFO [optim.py:368] (2/8) 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 15:36:30,545 INFO [train.py:904] (2/8) Epoch 12, batch 6950, loss[loss=0.1972, simple_loss=0.2883, pruned_loss=0.05301, over 16731.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3061, pruned_loss=0.0704, over 3107996.52 frames. ], batch size: 83, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:37:04,602 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:14,698 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:24,245 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5552, 2.3934, 2.2548, 4.1285, 2.6676, 3.9560, 1.4991, 2.6206], device='cuda:2'), covar=tensor([0.1432, 0.0844, 0.1383, 0.0151, 0.0303, 0.0456, 0.1591, 0.0937], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0161, 0.0181, 0.0150, 0.0198, 0.0207, 0.0182, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 15:37:35,482 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:44,791 INFO [train.py:904] (2/8) Epoch 12, batch 7000, loss[loss=0.2208, simple_loss=0.3033, pruned_loss=0.06915, over 15429.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.306, pruned_loss=0.06955, over 3102200.77 frames. ], batch size: 191, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:38:05,443 INFO [optim.py:368] (2/8) 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 15:38:33,803 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:44,053 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:50,366 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:59,538 INFO [train.py:904] (2/8) Epoch 12, batch 7050, loss[loss=0.238, simple_loss=0.3136, pruned_loss=0.08113, over 16269.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.307, pruned_loss=0.07057, over 3057923.64 frames. ], batch size: 165, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:39:35,542 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9163, 5.2203, 4.9615, 4.9540, 4.6870, 4.6763, 4.6252, 5.2951], device='cuda:2'), covar=tensor([0.0995, 0.0778, 0.0995, 0.0734, 0.0790, 0.0854, 0.1016, 0.0839], device='cuda:2'), in_proj_covar=tensor([0.0560, 0.0696, 0.0575, 0.0492, 0.0443, 0.0455, 0.0578, 0.0537], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:40:01,768 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:40:14,607 INFO [train.py:904] (2/8) Epoch 12, batch 7100, loss[loss=0.2523, simple_loss=0.3083, pruned_loss=0.0981, over 11052.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3058, pruned_loss=0.0704, over 3047128.38 frames. ], batch size: 248, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:40:31,972 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2796, 3.4373, 3.7986, 1.8015, 3.9710, 4.0603, 2.9198, 2.7953], device='cuda:2'), covar=tensor([0.0844, 0.0216, 0.0176, 0.1178, 0.0056, 0.0104, 0.0405, 0.0488], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0102, 0.0089, 0.0139, 0.0070, 0.0106, 0.0123, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 15:40:36,859 INFO [optim.py:368] (2/8) 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 15:41:29,295 INFO [train.py:904] (2/8) Epoch 12, batch 7150, loss[loss=0.2321, simple_loss=0.3013, pruned_loss=0.08147, over 11661.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3048, pruned_loss=0.0703, over 3052336.52 frames. ], batch size: 247, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:42:01,924 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:42:41,526 INFO [train.py:904] (2/8) Epoch 12, batch 7200, loss[loss=0.1983, simple_loss=0.2925, pruned_loss=0.05203, over 16736.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3027, pruned_loss=0.06781, over 3072240.46 frames. ], batch size: 76, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:43:00,621 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 15:43:03,912 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:43:28,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2535, 4.0463, 4.0081, 2.5508, 3.6250, 4.0068, 3.5840, 1.8501], device='cuda:2'), covar=tensor([0.0512, 0.0036, 0.0047, 0.0397, 0.0092, 0.0126, 0.0092, 0.0529], device='cuda:2'), in_proj_covar=tensor([0.0126, 0.0067, 0.0070, 0.0124, 0.0079, 0.0090, 0.0079, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 15:43:36,875 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7331, 1.6110, 2.1560, 2.5711, 2.5133, 2.9525, 1.5267, 2.8269], device='cuda:2'), covar=tensor([0.0124, 0.0414, 0.0231, 0.0194, 0.0200, 0.0110, 0.0479, 0.0078], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0171, 0.0154, 0.0158, 0.0171, 0.0124, 0.0171, 0.0117], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 15:43:40,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2652, 3.6862, 3.5787, 1.9973, 2.9400, 2.2846, 3.6275, 3.7476], device='cuda:2'), covar=tensor([0.0277, 0.0638, 0.0579, 0.1924, 0.0858, 0.0957, 0.0648, 0.0945], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0147, 0.0161, 0.0146, 0.0139, 0.0126, 0.0139, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 15:44:00,081 INFO [train.py:904] (2/8) Epoch 12, batch 7250, loss[loss=0.1828, simple_loss=0.2716, pruned_loss=0.04702, over 16822.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2994, pruned_loss=0.06583, over 3095557.14 frames. ], batch size: 102, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:44:35,495 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:04,507 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:15,112 INFO [train.py:904] (2/8) Epoch 12, batch 7300, loss[loss=0.2179, simple_loss=0.3031, pruned_loss=0.06635, over 15420.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2988, pruned_loss=0.06575, over 3083650.34 frames. ], batch size: 191, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:45:36,388 INFO [optim.py:368] (2/8) 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 15:45:45,781 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:03,498 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:06,457 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:28,512 INFO [train.py:904] (2/8) Epoch 12, batch 7350, loss[loss=0.2169, simple_loss=0.3016, pruned_loss=0.06609, over 17265.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2996, pruned_loss=0.06685, over 3076584.88 frames. ], batch size: 52, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:46:56,191 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6940, 3.9408, 3.0587, 2.2779, 2.9589, 2.5155, 4.2896, 3.6059], device='cuda:2'), covar=tensor([0.2820, 0.0683, 0.1744, 0.2349, 0.2288, 0.1754, 0.0414, 0.1042], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0259, 0.0291, 0.0285, 0.0283, 0.0225, 0.0270, 0.0303], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:47:14,407 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:47:42,935 INFO [train.py:904] (2/8) Epoch 12, batch 7400, loss[loss=0.2078, simple_loss=0.2997, pruned_loss=0.05794, over 16554.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3007, pruned_loss=0.06794, over 3060770.90 frames. ], batch size: 68, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:48:06,306 INFO [optim.py:368] (2/8) 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 15:48:19,138 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:48:47,019 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:49:01,209 INFO [train.py:904] (2/8) Epoch 12, batch 7450, loss[loss=0.1993, simple_loss=0.2911, pruned_loss=0.05378, over 16592.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3018, pruned_loss=0.06829, over 3082961.67 frames. ], batch size: 57, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:49:55,932 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:50:20,478 INFO [train.py:904] (2/8) Epoch 12, batch 7500, loss[loss=0.2072, simple_loss=0.2956, pruned_loss=0.05937, over 16823.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3024, pruned_loss=0.06802, over 3069769.80 frames. ], batch size: 116, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:50:23,472 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 15:50:42,266 INFO [optim.py:368] (2/8) 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 15:50:47,876 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 15:51:35,616 INFO [train.py:904] (2/8) Epoch 12, batch 7550, loss[loss=0.1979, simple_loss=0.2791, pruned_loss=0.05829, over 16741.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3022, pruned_loss=0.06873, over 3057730.27 frames. ], batch size: 124, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:51:48,079 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-29 15:52:50,121 INFO [train.py:904] (2/8) Epoch 12, batch 7600, loss[loss=0.1915, simple_loss=0.27, pruned_loss=0.05653, over 17025.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3014, pruned_loss=0.06883, over 3055897.70 frames. ], batch size: 53, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:53:12,404 INFO [optim.py:368] (2/8) 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 15:53:20,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1637, 3.2997, 1.8489, 3.6180, 2.3666, 3.6055, 2.0196, 2.5523], device='cuda:2'), covar=tensor([0.0248, 0.0396, 0.1680, 0.0146, 0.0888, 0.0547, 0.1572, 0.0784], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0164, 0.0189, 0.0129, 0.0166, 0.0203, 0.0195, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 15:53:30,764 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3208, 2.0332, 2.0453, 3.9707, 2.0302, 2.5948, 2.1313, 2.2603], device='cuda:2'), covar=tensor([0.0965, 0.3294, 0.2496, 0.0424, 0.3839, 0.2128, 0.3152, 0.2942], device='cuda:2'), in_proj_covar=tensor([0.0364, 0.0395, 0.0330, 0.0317, 0.0414, 0.0453, 0.0360, 0.0459], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:53:39,883 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3171, 3.4429, 1.9292, 3.8667, 2.5272, 3.8336, 2.0803, 2.5938], device='cuda:2'), covar=tensor([0.0262, 0.0429, 0.1706, 0.0158, 0.0856, 0.0481, 0.1613, 0.0784], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0164, 0.0189, 0.0130, 0.0167, 0.0204, 0.0195, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 15:53:43,227 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:54:05,863 INFO [train.py:904] (2/8) Epoch 12, batch 7650, loss[loss=0.2862, simple_loss=0.3439, pruned_loss=0.1143, over 11325.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3019, pruned_loss=0.06912, over 3061417.13 frames. ], batch size: 246, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:54:55,612 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:55:20,165 INFO [train.py:904] (2/8) Epoch 12, batch 7700, loss[loss=0.2507, simple_loss=0.3153, pruned_loss=0.09303, over 11516.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3018, pruned_loss=0.06986, over 3060889.46 frames. ], batch size: 248, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:55:42,613 INFO [optim.py:368] (2/8) 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 15:56:36,113 INFO [train.py:904] (2/8) Epoch 12, batch 7750, loss[loss=0.2213, simple_loss=0.307, pruned_loss=0.06775, over 16716.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.302, pruned_loss=0.07012, over 3035638.56 frames. ], batch size: 134, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:56:54,123 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5982, 4.4260, 4.6382, 4.8151, 4.9513, 4.4539, 4.9734, 4.9106], device='cuda:2'), covar=tensor([0.1697, 0.1157, 0.1606, 0.0606, 0.0570, 0.0909, 0.0583, 0.0594], device='cuda:2'), in_proj_covar=tensor([0.0518, 0.0652, 0.0782, 0.0667, 0.0503, 0.0512, 0.0533, 0.0601], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:57:18,792 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:57:43,634 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4127, 5.8036, 5.5083, 5.5192, 5.1910, 5.0585, 5.1972, 5.8840], device='cuda:2'), covar=tensor([0.1117, 0.0784, 0.1002, 0.0742, 0.0810, 0.0693, 0.1096, 0.0857], device='cuda:2'), in_proj_covar=tensor([0.0545, 0.0676, 0.0562, 0.0479, 0.0431, 0.0447, 0.0566, 0.0528], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:57:45,523 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 15:57:48,713 INFO [train.py:904] (2/8) Epoch 12, batch 7800, loss[loss=0.2843, simple_loss=0.3366, pruned_loss=0.116, over 11381.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3033, pruned_loss=0.07127, over 3025422.21 frames. ], batch size: 246, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:58:10,539 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7981, 3.9404, 2.1978, 4.3982, 2.9010, 4.3369, 2.4148, 2.9337], device='cuda:2'), covar=tensor([0.0213, 0.0302, 0.1569, 0.0169, 0.0706, 0.0522, 0.1429, 0.0746], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0163, 0.0190, 0.0130, 0.0168, 0.0205, 0.0197, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 15:58:11,184 INFO [optim.py:368] (2/8) 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 15:59:04,089 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4836, 3.4714, 3.4254, 2.8290, 3.2959, 2.1734, 3.0988, 2.7271], device='cuda:2'), covar=tensor([0.0123, 0.0102, 0.0149, 0.0207, 0.0085, 0.1862, 0.0115, 0.0183], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0118, 0.0163, 0.0155, 0.0135, 0.0179, 0.0151, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 15:59:04,882 INFO [train.py:904] (2/8) Epoch 12, batch 7850, loss[loss=0.248, simple_loss=0.3166, pruned_loss=0.08965, over 11261.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3043, pruned_loss=0.0711, over 3026570.93 frames. ], batch size: 248, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 15:59:11,081 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 15:59:23,788 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:00:21,535 INFO [train.py:904] (2/8) Epoch 12, batch 7900, loss[loss=0.2128, simple_loss=0.302, pruned_loss=0.06179, over 16890.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3025, pruned_loss=0.0698, over 3041530.78 frames. ], batch size: 116, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:00:45,732 INFO [optim.py:368] (2/8) 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 16:00:56,302 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:01:23,851 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4376, 4.4443, 4.8548, 4.8405, 4.8731, 4.5313, 4.5369, 4.3646], device='cuda:2'), covar=tensor([0.0310, 0.0475, 0.0421, 0.0421, 0.0470, 0.0377, 0.0892, 0.0442], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0360, 0.0359, 0.0340, 0.0409, 0.0381, 0.0478, 0.0305], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 16:01:37,090 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 16:01:38,575 INFO [train.py:904] (2/8) Epoch 12, batch 7950, loss[loss=0.1947, simple_loss=0.2824, pruned_loss=0.05347, over 16718.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3027, pruned_loss=0.0704, over 3028490.49 frames. ], batch size: 89, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:02:53,353 INFO [train.py:904] (2/8) Epoch 12, batch 8000, loss[loss=0.2207, simple_loss=0.3132, pruned_loss=0.06412, over 16804.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3033, pruned_loss=0.07096, over 3027336.86 frames. ], batch size: 102, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 16:03:06,815 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1756, 3.3640, 3.5574, 3.5359, 3.5308, 3.3416, 3.3640, 3.3917], device='cuda:2'), covar=tensor([0.0396, 0.0606, 0.0395, 0.0446, 0.0514, 0.0543, 0.0823, 0.0574], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0359, 0.0357, 0.0340, 0.0409, 0.0380, 0.0475, 0.0305], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 16:03:12,809 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:03:15,320 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5615, 2.8475, 2.8147, 4.9716, 3.8603, 4.3951, 1.4969, 3.5161], device='cuda:2'), covar=tensor([0.1577, 0.0850, 0.1255, 0.0192, 0.0571, 0.0411, 0.1825, 0.0848], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0163, 0.0183, 0.0151, 0.0200, 0.0208, 0.0183, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 16:03:17,120 INFO [optim.py:368] (2/8) 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 16:04:07,956 INFO [train.py:904] (2/8) Epoch 12, batch 8050, loss[loss=0.2608, simple_loss=0.3206, pruned_loss=0.1005, over 11669.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3021, pruned_loss=0.06939, over 3051752.80 frames. ], batch size: 248, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:04:30,043 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 16:04:42,928 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:04:49,748 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:05:17,337 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 16:05:21,186 INFO [train.py:904] (2/8) Epoch 12, batch 8100, loss[loss=0.187, simple_loss=0.2796, pruned_loss=0.04725, over 16891.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3021, pruned_loss=0.06915, over 3055526.04 frames. ], batch size: 96, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:05:47,745 INFO [optim.py:368] (2/8) 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 16:06:00,175 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:28,057 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:34,782 INFO [train.py:904] (2/8) Epoch 12, batch 8150, loss[loss=0.1812, simple_loss=0.2589, pruned_loss=0.05174, over 16319.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2995, pruned_loss=0.06797, over 3063809.21 frames. ], batch size: 35, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:07:50,608 INFO [train.py:904] (2/8) Epoch 12, batch 8200, loss[loss=0.1935, simple_loss=0.2862, pruned_loss=0.05037, over 16840.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2966, pruned_loss=0.06659, over 3085975.32 frames. ], batch size: 96, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:07:59,093 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 16:08:18,242 INFO [optim.py:368] (2/8) 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 16:08:19,310 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:09:09,078 INFO [train.py:904] (2/8) Epoch 12, batch 8250, loss[loss=0.182, simple_loss=0.2836, pruned_loss=0.04017, over 16879.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2961, pruned_loss=0.06434, over 3084022.08 frames. ], batch size: 96, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:09:13,247 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-29 16:09:26,733 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0576, 4.0509, 4.4244, 4.3940, 4.3865, 4.1351, 4.1036, 4.0591], device='cuda:2'), covar=tensor([0.0297, 0.0563, 0.0347, 0.0401, 0.0437, 0.0348, 0.0898, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0339, 0.0357, 0.0354, 0.0338, 0.0407, 0.0379, 0.0476, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 16:10:28,023 INFO [train.py:904] (2/8) Epoch 12, batch 8300, loss[loss=0.1944, simple_loss=0.2857, pruned_loss=0.05153, over 16885.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2935, pruned_loss=0.0615, over 3076574.96 frames. ], batch size: 109, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:10:57,564 INFO [optim.py:368] (2/8) 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 16:11:27,126 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2295, 1.5558, 1.9222, 2.1538, 2.2840, 2.4225, 1.6523, 2.4315], device='cuda:2'), covar=tensor([0.0147, 0.0366, 0.0196, 0.0209, 0.0218, 0.0142, 0.0340, 0.0093], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0170, 0.0154, 0.0157, 0.0169, 0.0123, 0.0169, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 16:11:52,968 INFO [train.py:904] (2/8) Epoch 12, batch 8350, loss[loss=0.206, simple_loss=0.2945, pruned_loss=0.05871, over 16934.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2924, pruned_loss=0.05907, over 3083247.84 frames. ], batch size: 109, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:12:24,541 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:12:24,676 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5677, 3.5229, 3.4860, 2.8957, 3.3810, 2.0497, 3.2713, 2.8731], device='cuda:2'), covar=tensor([0.0123, 0.0108, 0.0140, 0.0223, 0.0096, 0.2092, 0.0112, 0.0218], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0117, 0.0162, 0.0153, 0.0134, 0.0179, 0.0150, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:12:31,202 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3039, 4.7568, 3.6412, 2.5846, 3.3830, 2.9017, 5.1043, 4.2625], device='cuda:2'), covar=tensor([0.2150, 0.0404, 0.1301, 0.2271, 0.2183, 0.1533, 0.0255, 0.0812], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0250, 0.0279, 0.0275, 0.0273, 0.0218, 0.0262, 0.0289], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:13:14,450 INFO [train.py:904] (2/8) Epoch 12, batch 8400, loss[loss=0.1898, simple_loss=0.2844, pruned_loss=0.04764, over 16146.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2895, pruned_loss=0.05695, over 3074778.15 frames. ], batch size: 165, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:13:38,514 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0585, 2.7901, 2.8586, 2.0145, 2.6926, 2.2097, 2.7522, 2.9317], device='cuda:2'), covar=tensor([0.0316, 0.0753, 0.0510, 0.1805, 0.0762, 0.0915, 0.0644, 0.0751], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0142, 0.0156, 0.0141, 0.0134, 0.0123, 0.0134, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 16:13:42,956 INFO [optim.py:368] (2/8) 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 16:13:53,534 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 16:13:54,458 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9042, 3.5586, 3.4406, 1.8885, 2.9591, 2.5298, 3.2710, 3.5356], device='cuda:2'), covar=tensor([0.0297, 0.0632, 0.0546, 0.1972, 0.0802, 0.0911, 0.0780, 0.0884], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0141, 0.0155, 0.0141, 0.0134, 0.0123, 0.0134, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 16:14:31,500 INFO [train.py:904] (2/8) Epoch 12, batch 8450, loss[loss=0.1773, simple_loss=0.2617, pruned_loss=0.04649, over 12460.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2872, pruned_loss=0.05509, over 3069151.91 frames. ], batch size: 247, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:14:47,541 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5379, 3.6431, 3.4060, 3.1093, 3.1888, 3.5255, 3.3176, 3.3084], device='cuda:2'), covar=tensor([0.0574, 0.0492, 0.0265, 0.0249, 0.0559, 0.0342, 0.1108, 0.0516], device='cuda:2'), in_proj_covar=tensor([0.0238, 0.0321, 0.0283, 0.0261, 0.0300, 0.0303, 0.0198, 0.0331], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:15:03,143 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-29 16:15:07,156 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4358, 2.0429, 1.7471, 1.6993, 2.2814, 1.9263, 2.1776, 2.4099], device='cuda:2'), covar=tensor([0.0132, 0.0258, 0.0345, 0.0349, 0.0180, 0.0246, 0.0151, 0.0166], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0200, 0.0195, 0.0195, 0.0201, 0.0198, 0.0201, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:15:19,907 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 16:15:22,923 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 16:15:42,507 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-29 16:15:47,448 INFO [train.py:904] (2/8) Epoch 12, batch 8500, loss[loss=0.1754, simple_loss=0.2787, pruned_loss=0.03606, over 16794.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2832, pruned_loss=0.05256, over 3062810.03 frames. ], batch size: 102, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:16:15,430 INFO [optim.py:368] (2/8) 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 16:16:15,902 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:17:07,733 INFO [train.py:904] (2/8) Epoch 12, batch 8550, loss[loss=0.2022, simple_loss=0.2957, pruned_loss=0.0544, over 16890.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2809, pruned_loss=0.05161, over 3042285.46 frames. ], batch size: 109, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:17:37,326 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:18:41,889 INFO [train.py:904] (2/8) Epoch 12, batch 8600, loss[loss=0.198, simple_loss=0.2777, pruned_loss=0.05915, over 12428.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2815, pruned_loss=0.05069, over 3045982.10 frames. ], batch size: 250, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:19:19,442 INFO [optim.py:368] (2/8) 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 16:20:19,209 INFO [train.py:904] (2/8) Epoch 12, batch 8650, loss[loss=0.167, simple_loss=0.2694, pruned_loss=0.03225, over 15217.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2789, pruned_loss=0.0489, over 3052157.57 frames. ], batch size: 190, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:21:01,218 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:22:02,153 INFO [train.py:904] (2/8) Epoch 12, batch 8700, loss[loss=0.1664, simple_loss=0.2658, pruned_loss=0.03349, over 16483.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2757, pruned_loss=0.04731, over 3045337.67 frames. ], batch size: 68, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:22:33,637 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120369.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:22:34,423 INFO [optim.py:368] (2/8) 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 16:23:36,530 INFO [train.py:904] (2/8) Epoch 12, batch 8750, loss[loss=0.193, simple_loss=0.2861, pruned_loss=0.04996, over 15353.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2751, pruned_loss=0.04697, over 3045014.78 frames. ], batch size: 191, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:25:30,723 INFO [train.py:904] (2/8) Epoch 12, batch 8800, loss[loss=0.2007, simple_loss=0.2871, pruned_loss=0.05714, over 16713.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2735, pruned_loss=0.0459, over 3046151.22 frames. ], batch size: 134, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:26:08,429 INFO [optim.py:368] (2/8) 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 16:26:16,125 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0414, 2.7608, 2.8188, 1.9239, 2.5870, 2.1438, 2.6802, 2.9193], device='cuda:2'), covar=tensor([0.0301, 0.0801, 0.0532, 0.1877, 0.0781, 0.0979, 0.0695, 0.0744], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0140, 0.0155, 0.0141, 0.0134, 0.0123, 0.0134, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 16:26:17,833 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-04-29 16:26:24,595 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 16:27:17,123 INFO [train.py:904] (2/8) Epoch 12, batch 8850, loss[loss=0.1727, simple_loss=0.2594, pruned_loss=0.04297, over 12542.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2753, pruned_loss=0.04517, over 3023823.45 frames. ], batch size: 250, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:29:04,467 INFO [train.py:904] (2/8) Epoch 12, batch 8900, loss[loss=0.2067, simple_loss=0.2963, pruned_loss=0.05855, over 16892.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2762, pruned_loss=0.04474, over 3023087.87 frames. ], batch size: 116, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:29:37,100 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3589, 3.2610, 3.2807, 3.5001, 3.5246, 3.2376, 3.5253, 3.5658], device='cuda:2'), covar=tensor([0.1194, 0.1041, 0.1415, 0.0780, 0.0792, 0.2735, 0.1035, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0501, 0.0632, 0.0751, 0.0642, 0.0485, 0.0496, 0.0510, 0.0584], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:29:39,305 INFO [optim.py:368] (2/8) 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 16:29:49,223 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5279, 4.5262, 4.9290, 4.9135, 4.9152, 4.6021, 4.6041, 4.4662], device='cuda:2'), covar=tensor([0.0263, 0.0521, 0.0390, 0.0423, 0.0408, 0.0331, 0.0874, 0.0373], device='cuda:2'), in_proj_covar=tensor([0.0321, 0.0336, 0.0334, 0.0318, 0.0382, 0.0358, 0.0445, 0.0288], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 16:31:10,991 INFO [train.py:904] (2/8) Epoch 12, batch 8950, loss[loss=0.1817, simple_loss=0.2649, pruned_loss=0.04923, over 12286.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2759, pruned_loss=0.04544, over 3027538.77 frames. ], batch size: 248, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:32:42,182 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8327, 3.8019, 3.9780, 3.7586, 3.8313, 4.2867, 3.9510, 3.6693], device='cuda:2'), covar=tensor([0.1941, 0.2105, 0.2051, 0.2442, 0.2861, 0.1444, 0.1500, 0.2789], device='cuda:2'), in_proj_covar=tensor([0.0337, 0.0469, 0.0513, 0.0403, 0.0546, 0.0542, 0.0410, 0.0546], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 16:33:00,365 INFO [train.py:904] (2/8) Epoch 12, batch 9000, loss[loss=0.1726, simple_loss=0.2677, pruned_loss=0.03871, over 15316.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2724, pruned_loss=0.044, over 3031878.36 frames. ], batch size: 191, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:33:00,365 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 16:33:10,342 INFO [train.py:938] (2/8) Epoch 12, validation: loss=0.1532, simple_loss=0.2571, pruned_loss=0.02465, over 944034.00 frames. 2023-04-29 16:33:10,344 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 16:33:42,499 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 16:33:49,106 INFO [optim.py:368] (2/8) 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 16:34:01,172 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:34:16,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2930, 1.5491, 1.8589, 2.2635, 2.2885, 2.4071, 1.6528, 2.4179], device='cuda:2'), covar=tensor([0.0180, 0.0392, 0.0271, 0.0247, 0.0234, 0.0158, 0.0402, 0.0128], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0171, 0.0156, 0.0156, 0.0168, 0.0122, 0.0170, 0.0115], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 16:34:54,270 INFO [train.py:904] (2/8) Epoch 12, batch 9050, loss[loss=0.1835, simple_loss=0.274, pruned_loss=0.04652, over 15446.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2738, pruned_loss=0.04452, over 3048530.04 frames. ], batch size: 190, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:35:25,439 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5409, 3.6139, 2.9444, 2.2029, 2.2311, 2.2608, 3.8245, 3.3007], device='cuda:2'), covar=tensor([0.2527, 0.0583, 0.1430, 0.2485, 0.2501, 0.1806, 0.0341, 0.0971], device='cuda:2'), in_proj_covar=tensor([0.0295, 0.0246, 0.0278, 0.0272, 0.0263, 0.0216, 0.0259, 0.0284], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:36:04,418 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:36:39,406 INFO [train.py:904] (2/8) Epoch 12, batch 9100, loss[loss=0.1788, simple_loss=0.2638, pruned_loss=0.0469, over 12420.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2734, pruned_loss=0.04501, over 3037092.81 frames. ], batch size: 247, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:37:15,445 INFO [optim.py:368] (2/8) 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 16:38:04,561 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5351, 3.5432, 3.4987, 2.9718, 3.3974, 2.0016, 3.2438, 2.9620], device='cuda:2'), covar=tensor([0.0115, 0.0098, 0.0138, 0.0184, 0.0087, 0.2060, 0.0105, 0.0197], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0116, 0.0158, 0.0147, 0.0132, 0.0178, 0.0147, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:38:36,941 INFO [train.py:904] (2/8) Epoch 12, batch 9150, loss[loss=0.1811, simple_loss=0.2737, pruned_loss=0.04427, over 15361.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2738, pruned_loss=0.04449, over 3047562.26 frames. ], batch size: 191, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:38:55,809 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0169, 2.6873, 2.7410, 1.9588, 2.5497, 2.1097, 2.6817, 2.8277], device='cuda:2'), covar=tensor([0.0286, 0.0771, 0.0530, 0.1792, 0.0787, 0.0941, 0.0662, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0139, 0.0155, 0.0141, 0.0133, 0.0123, 0.0133, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 16:40:21,514 INFO [train.py:904] (2/8) Epoch 12, batch 9200, loss[loss=0.1558, simple_loss=0.2407, pruned_loss=0.03547, over 12009.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2694, pruned_loss=0.04319, over 3050819.07 frames. ], batch size: 247, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:55,536 INFO [optim.py:368] (2/8) 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 16:42:00,506 INFO [train.py:904] (2/8) Epoch 12, batch 9250, loss[loss=0.1683, simple_loss=0.26, pruned_loss=0.03828, over 16277.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2686, pruned_loss=0.04315, over 3039380.16 frames. ], batch size: 35, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:43:49,113 INFO [train.py:904] (2/8) Epoch 12, batch 9300, loss[loss=0.1676, simple_loss=0.2551, pruned_loss=0.04003, over 16404.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2678, pruned_loss=0.04275, over 3054913.82 frames. ], batch size: 146, lr: 5.60e-03, grad_scale: 4.0 2023-04-29 16:44:31,798 INFO [optim.py:368] (2/8) 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 16:45:24,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6842, 3.6123, 3.5643, 3.8695, 3.9527, 3.5815, 3.9330, 3.9817], device='cuda:2'), covar=tensor([0.1424, 0.1107, 0.2016, 0.0962, 0.0841, 0.2200, 0.0979, 0.0963], device='cuda:2'), in_proj_covar=tensor([0.0490, 0.0618, 0.0736, 0.0629, 0.0474, 0.0487, 0.0499, 0.0571], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:45:32,147 INFO [train.py:904] (2/8) Epoch 12, batch 9350, loss[loss=0.1905, simple_loss=0.282, pruned_loss=0.04952, over 16176.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2678, pruned_loss=0.04273, over 3064470.93 frames. ], batch size: 165, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:45:35,038 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:46:33,081 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:13,259 INFO [train.py:904] (2/8) Epoch 12, batch 9400, loss[loss=0.171, simple_loss=0.2583, pruned_loss=0.04184, over 12780.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2679, pruned_loss=0.04279, over 3058241.98 frames. ], batch size: 247, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:47:39,131 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:50,521 INFO [optim.py:368] (2/8) 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 16:48:34,541 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 16:48:55,033 INFO [train.py:904] (2/8) Epoch 12, batch 9450, loss[loss=0.1895, simple_loss=0.2723, pruned_loss=0.05335, over 12355.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2693, pruned_loss=0.04267, over 3054689.82 frames. ], batch size: 246, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:49:13,461 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-29 16:50:34,751 INFO [train.py:904] (2/8) Epoch 12, batch 9500, loss[loss=0.1733, simple_loss=0.2629, pruned_loss=0.0419, over 12689.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2693, pruned_loss=0.04251, over 3056202.96 frames. ], batch size: 246, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:51:01,757 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:51:13,616 INFO [optim.py:368] (2/8) 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 16:51:15,777 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0947, 2.8401, 2.8598, 2.0775, 2.6782, 2.2754, 2.6164, 2.9407], device='cuda:2'), covar=tensor([0.0351, 0.0776, 0.0489, 0.1558, 0.0767, 0.0822, 0.0843, 0.0749], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0138, 0.0155, 0.0140, 0.0133, 0.0122, 0.0132, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 16:51:32,006 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-04-29 16:51:38,936 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 16:52:19,999 INFO [train.py:904] (2/8) Epoch 12, batch 9550, loss[loss=0.195, simple_loss=0.2914, pruned_loss=0.04923, over 16197.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.269, pruned_loss=0.04269, over 3058208.15 frames. ], batch size: 165, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:53:08,741 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:53:32,389 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9209, 4.2323, 3.2535, 2.3582, 2.8788, 2.5784, 4.4984, 3.8381], device='cuda:2'), covar=tensor([0.2363, 0.0481, 0.1372, 0.2221, 0.2198, 0.1661, 0.0333, 0.0835], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0247, 0.0277, 0.0271, 0.0258, 0.0215, 0.0258, 0.0283], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 16:53:33,704 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0827, 2.5224, 2.6609, 1.8442, 2.8312, 2.8836, 2.4901, 2.4245], device='cuda:2'), covar=tensor([0.0647, 0.0194, 0.0166, 0.0952, 0.0083, 0.0170, 0.0433, 0.0407], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0097, 0.0084, 0.0135, 0.0067, 0.0101, 0.0118, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 16:53:59,925 INFO [train.py:904] (2/8) Epoch 12, batch 9600, loss[loss=0.2091, simple_loss=0.3026, pruned_loss=0.05782, over 16264.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2705, pruned_loss=0.04352, over 3053942.14 frames. ], batch size: 165, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:54:12,296 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 16:54:35,186 INFO [optim.py:368] (2/8) 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 16:55:45,648 INFO [train.py:904] (2/8) Epoch 12, batch 9650, loss[loss=0.1774, simple_loss=0.268, pruned_loss=0.04337, over 16839.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2723, pruned_loss=0.0438, over 3050451.76 frames. ], batch size: 124, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:55:56,690 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9088, 2.7541, 2.7748, 2.0398, 2.7256, 2.2008, 2.6450, 2.8640], device='cuda:2'), covar=tensor([0.0250, 0.0717, 0.0462, 0.1598, 0.0687, 0.0844, 0.0562, 0.0675], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0137, 0.0154, 0.0140, 0.0133, 0.0122, 0.0132, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 16:56:09,922 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 16:56:52,029 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:57:30,749 INFO [train.py:904] (2/8) Epoch 12, batch 9700, loss[loss=0.1754, simple_loss=0.2693, pruned_loss=0.04078, over 16376.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2714, pruned_loss=0.04331, over 3063960.47 frames. ], batch size: 146, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:57:44,380 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:57:50,849 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5512, 2.5683, 2.2202, 3.8812, 2.0866, 3.7980, 1.4108, 2.6911], device='cuda:2'), covar=tensor([0.1654, 0.0880, 0.1384, 0.0187, 0.0137, 0.0373, 0.1907, 0.0904], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0143, 0.0185, 0.0202, 0.0181, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 16:58:05,427 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7857, 1.3174, 1.6754, 1.6800, 1.8938, 1.8741, 1.5934, 1.8508], device='cuda:2'), covar=tensor([0.0211, 0.0287, 0.0159, 0.0204, 0.0197, 0.0166, 0.0298, 0.0092], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0168, 0.0154, 0.0156, 0.0165, 0.0120, 0.0168, 0.0113], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:2') 2023-04-29 16:58:07,288 INFO [optim.py:368] (2/8) 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 16:58:10,418 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 16:58:31,091 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:59:14,333 INFO [train.py:904] (2/8) Epoch 12, batch 9750, loss[loss=0.1699, simple_loss=0.2547, pruned_loss=0.04258, over 12278.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2701, pruned_loss=0.04345, over 3060352.73 frames. ], batch size: 250, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:59:23,478 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:00:53,746 INFO [train.py:904] (2/8) Epoch 12, batch 9800, loss[loss=0.1588, simple_loss=0.2638, pruned_loss=0.02685, over 16838.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2693, pruned_loss=0.04233, over 3061749.69 frames. ], batch size: 83, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:01:01,231 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6788, 2.1997, 2.2655, 4.4110, 2.1658, 2.6566, 2.2956, 2.4837], device='cuda:2'), covar=tensor([0.0783, 0.3208, 0.2402, 0.0295, 0.3797, 0.2207, 0.3233, 0.2687], device='cuda:2'), in_proj_covar=tensor([0.0354, 0.0382, 0.0325, 0.0306, 0.0404, 0.0434, 0.0350, 0.0445], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:01:24,673 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:01:29,885 INFO [optim.py:368] (2/8) 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 17:02:38,494 INFO [train.py:904] (2/8) Epoch 12, batch 9850, loss[loss=0.1817, simple_loss=0.2613, pruned_loss=0.05108, over 12535.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2706, pruned_loss=0.04245, over 3073814.62 frames. ], batch size: 250, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:03:17,662 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:03:44,620 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8518, 3.8232, 3.9825, 3.7731, 3.9472, 4.3163, 3.9437, 3.6291], device='cuda:2'), covar=tensor([0.1663, 0.1933, 0.1779, 0.2257, 0.2304, 0.1277, 0.1361, 0.2597], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0463, 0.0512, 0.0400, 0.0534, 0.0533, 0.0402, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:04:30,196 INFO [train.py:904] (2/8) Epoch 12, batch 9900, loss[loss=0.1774, simple_loss=0.2856, pruned_loss=0.03464, over 16888.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2716, pruned_loss=0.04256, over 3082357.06 frames. ], batch size: 102, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:04:37,655 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 17:05:12,985 INFO [optim.py:368] (2/8) 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 17:06:26,643 INFO [train.py:904] (2/8) Epoch 12, batch 9950, loss[loss=0.1885, simple_loss=0.2862, pruned_loss=0.0454, over 16769.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2743, pruned_loss=0.04303, over 3094432.11 frames. ], batch size: 124, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:08:26,809 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8475, 4.0139, 4.3565, 4.3831, 4.3489, 4.0718, 3.7878, 3.9686], device='cuda:2'), covar=tensor([0.0526, 0.0707, 0.0650, 0.0614, 0.0669, 0.0568, 0.1300, 0.0518], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0332, 0.0331, 0.0315, 0.0376, 0.0355, 0.0438, 0.0285], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:2') 2023-04-29 17:08:27,932 INFO [train.py:904] (2/8) Epoch 12, batch 10000, loss[loss=0.1787, simple_loss=0.2774, pruned_loss=0.03994, over 16636.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2723, pruned_loss=0.04197, over 3118677.65 frames. ], batch size: 134, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:08:44,224 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:08:59,914 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:09:06,349 INFO [optim.py:368] (2/8) 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 17:09:07,077 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3500, 3.4212, 2.0673, 3.7300, 2.5488, 3.6805, 2.1703, 2.6433], device='cuda:2'), covar=tensor([0.0235, 0.0335, 0.1536, 0.0182, 0.0779, 0.0524, 0.1454, 0.0752], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0155, 0.0180, 0.0123, 0.0162, 0.0191, 0.0191, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 17:10:10,826 INFO [train.py:904] (2/8) Epoch 12, batch 10050, loss[loss=0.18, simple_loss=0.2778, pruned_loss=0.04107, over 15379.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2723, pruned_loss=0.04178, over 3114272.70 frames. ], batch size: 190, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:10:21,422 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:11:13,697 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 17:11:21,635 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-29 17:11:22,878 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 17:11:46,088 INFO [train.py:904] (2/8) Epoch 12, batch 10100, loss[loss=0.1754, simple_loss=0.2684, pruned_loss=0.04121, over 16222.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2727, pruned_loss=0.04233, over 3101594.94 frames. ], batch size: 165, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:12:05,575 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:12:23,724 INFO [optim.py:368] (2/8) 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 17:13:30,553 INFO [train.py:904] (2/8) Epoch 13, batch 0, loss[loss=0.1928, simple_loss=0.2861, pruned_loss=0.04977, over 17131.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2861, pruned_loss=0.04977, over 17131.00 frames. ], batch size: 48, lr: 5.36e-03, grad_scale: 8.0 2023-04-29 17:13:30,553 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 17:13:38,104 INFO [train.py:938] (2/8) Epoch 13, validation: loss=0.1523, simple_loss=0.2559, pruned_loss=0.02431, over 944034.00 frames. 2023-04-29 17:13:38,105 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 17:14:04,098 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:14:23,066 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 17:14:49,499 INFO [train.py:904] (2/8) Epoch 13, batch 50, loss[loss=0.1845, simple_loss=0.2723, pruned_loss=0.04833, over 17104.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2851, pruned_loss=0.05943, over 740143.54 frames. ], batch size: 47, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:15:11,431 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:15:18,816 INFO [optim.py:368] (2/8) 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 17:15:58,269 INFO [train.py:904] (2/8) Epoch 13, batch 100, loss[loss=0.1904, simple_loss=0.283, pruned_loss=0.04888, over 17272.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2789, pruned_loss=0.05674, over 1317799.06 frames. ], batch size: 52, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:17:04,477 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:17:07,072 INFO [train.py:904] (2/8) Epoch 13, batch 150, loss[loss=0.173, simple_loss=0.2652, pruned_loss=0.04038, over 17113.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2763, pruned_loss=0.05631, over 1760276.50 frames. ], batch size: 47, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:17:27,927 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:17:35,630 INFO [optim.py:368] (2/8) 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 17:18:06,992 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-29 17:18:18,195 INFO [train.py:904] (2/8) Epoch 13, batch 200, loss[loss=0.1956, simple_loss=0.2806, pruned_loss=0.05533, over 16743.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2747, pruned_loss=0.0549, over 2112596.57 frames. ], batch size: 62, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:18:30,512 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:18:35,256 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:19:26,236 INFO [train.py:904] (2/8) Epoch 13, batch 250, loss[loss=0.1725, simple_loss=0.2693, pruned_loss=0.03785, over 17121.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2725, pruned_loss=0.05385, over 2380791.60 frames. ], batch size: 49, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:19:41,354 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:19:54,205 INFO [optim.py:368] (2/8) 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 17:20:14,390 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 17:20:34,740 INFO [train.py:904] (2/8) Epoch 13, batch 300, loss[loss=0.224, simple_loss=0.2878, pruned_loss=0.08007, over 12299.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.27, pruned_loss=0.05279, over 2588358.09 frames. ], batch size: 246, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:20:47,419 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:20:55,779 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:21:42,719 INFO [train.py:904] (2/8) Epoch 13, batch 350, loss[loss=0.1698, simple_loss=0.2593, pruned_loss=0.04017, over 17106.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2677, pruned_loss=0.05191, over 2749395.71 frames. ], batch size: 48, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:22:13,955 INFO [optim.py:368] (2/8) 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 17:22:20,607 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:22:43,772 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:22:53,136 INFO [train.py:904] (2/8) Epoch 13, batch 400, loss[loss=0.1693, simple_loss=0.2618, pruned_loss=0.03846, over 17127.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2664, pruned_loss=0.0516, over 2870396.35 frames. ], batch size: 49, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:23:37,065 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3638, 4.6349, 4.4492, 4.4591, 4.1424, 4.1394, 4.2025, 4.6872], device='cuda:2'), covar=tensor([0.1029, 0.0911, 0.0971, 0.0732, 0.0828, 0.1427, 0.0973, 0.0956], device='cuda:2'), in_proj_covar=tensor([0.0559, 0.0705, 0.0574, 0.0498, 0.0450, 0.0460, 0.0588, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:24:03,761 INFO [train.py:904] (2/8) Epoch 13, batch 450, loss[loss=0.1742, simple_loss=0.2622, pruned_loss=0.04308, over 17227.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2643, pruned_loss=0.05007, over 2973088.86 frames. ], batch size: 45, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:06,469 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6121, 5.9819, 5.7635, 5.7736, 5.3414, 5.2625, 5.3687, 6.1481], device='cuda:2'), covar=tensor([0.1239, 0.0910, 0.1081, 0.0788, 0.0867, 0.0695, 0.1099, 0.0879], device='cuda:2'), in_proj_covar=tensor([0.0563, 0.0710, 0.0578, 0.0501, 0.0453, 0.0462, 0.0592, 0.0538], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:24:06,710 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7151, 2.1350, 2.3919, 4.6071, 2.2311, 2.6247, 2.3349, 2.3969], device='cuda:2'), covar=tensor([0.0940, 0.3666, 0.2564, 0.0375, 0.3970, 0.2591, 0.3181, 0.3648], device='cuda:2'), in_proj_covar=tensor([0.0367, 0.0397, 0.0337, 0.0321, 0.0415, 0.0454, 0.0363, 0.0463], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:24:09,665 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:24:32,115 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1812, 5.7499, 5.9553, 5.6960, 5.7792, 6.3194, 5.9049, 5.6076], device='cuda:2'), covar=tensor([0.0781, 0.1742, 0.2128, 0.2247, 0.2500, 0.0949, 0.1362, 0.2242], device='cuda:2'), in_proj_covar=tensor([0.0364, 0.0513, 0.0567, 0.0444, 0.0592, 0.0588, 0.0444, 0.0590], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 17:24:32,944 INFO [optim.py:368] (2/8) 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 17:24:54,247 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4601, 4.2859, 4.4774, 4.6874, 4.7920, 4.2862, 4.7051, 4.7768], device='cuda:2'), covar=tensor([0.1449, 0.1150, 0.1477, 0.0707, 0.0569, 0.1111, 0.1135, 0.0687], device='cuda:2'), in_proj_covar=tensor([0.0542, 0.0678, 0.0818, 0.0688, 0.0518, 0.0531, 0.0549, 0.0623], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:25:13,886 INFO [train.py:904] (2/8) Epoch 13, batch 500, loss[loss=0.1986, simple_loss=0.2739, pruned_loss=0.06171, over 12320.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2632, pruned_loss=0.04938, over 3037335.95 frames. ], batch size: 247, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:25:19,449 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:25:36,838 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:26:21,438 INFO [train.py:904] (2/8) Epoch 13, batch 550, loss[loss=0.1648, simple_loss=0.2507, pruned_loss=0.0394, over 17171.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2618, pruned_loss=0.04866, over 3091375.24 frames. ], batch size: 46, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:26:49,557 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1235, 1.9997, 2.6552, 3.0649, 2.8295, 3.4675, 2.2389, 3.4209], device='cuda:2'), covar=tensor([0.0159, 0.0373, 0.0220, 0.0215, 0.0217, 0.0128, 0.0336, 0.0117], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0160, 0.0163, 0.0174, 0.0127, 0.0175, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 17:26:50,249 INFO [optim.py:368] (2/8) 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 17:27:00,778 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:27:30,037 INFO [train.py:904] (2/8) Epoch 13, batch 600, loss[loss=0.1787, simple_loss=0.2546, pruned_loss=0.0514, over 16492.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2623, pruned_loss=0.04998, over 3137138.65 frames. ], batch size: 75, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:28:37,794 INFO [train.py:904] (2/8) Epoch 13, batch 650, loss[loss=0.1648, simple_loss=0.2598, pruned_loss=0.03492, over 17030.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2612, pruned_loss=0.04968, over 3180819.82 frames. ], batch size: 50, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:29:01,164 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:29:07,724 INFO [optim.py:368] (2/8) 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 17:29:08,003 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:29:47,560 INFO [train.py:904] (2/8) Epoch 13, batch 700, loss[loss=0.173, simple_loss=0.2551, pruned_loss=0.04539, over 16938.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2612, pruned_loss=0.04939, over 3206475.01 frames. ], batch size: 41, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:30:25,439 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:56,258 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:57,590 INFO [train.py:904] (2/8) Epoch 13, batch 750, loss[loss=0.1797, simple_loss=0.2526, pruned_loss=0.05335, over 16711.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2611, pruned_loss=0.04936, over 3233763.89 frames. ], batch size: 89, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:31:06,671 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9628, 3.2610, 3.0389, 2.0809, 2.7661, 2.3958, 3.4315, 3.4215], device='cuda:2'), covar=tensor([0.0272, 0.0749, 0.0691, 0.1723, 0.0867, 0.0860, 0.0580, 0.0769], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0145, 0.0159, 0.0144, 0.0138, 0.0125, 0.0136, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 17:31:27,840 INFO [optim.py:368] (2/8) 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 17:32:09,553 INFO [train.py:904] (2/8) Epoch 13, batch 800, loss[loss=0.1647, simple_loss=0.2606, pruned_loss=0.03441, over 17132.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2609, pruned_loss=0.0491, over 3249854.85 frames. ], batch size: 49, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:32:15,243 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:15,726 INFO [train.py:904] (2/8) Epoch 13, batch 850, loss[loss=0.1643, simple_loss=0.2349, pruned_loss=0.04684, over 16694.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2605, pruned_loss=0.04896, over 3261849.19 frames. ], batch size: 89, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:33:18,234 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:36,314 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8472, 3.0593, 3.1405, 2.0972, 2.7924, 2.2419, 3.3511, 3.3066], device='cuda:2'), covar=tensor([0.0215, 0.0847, 0.0559, 0.1679, 0.0808, 0.0922, 0.0517, 0.0790], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0146, 0.0159, 0.0145, 0.0139, 0.0125, 0.0136, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 17:33:44,225 INFO [optim.py:368] (2/8) 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 17:33:46,988 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:34:05,724 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 17:34:21,254 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-04-29 17:34:23,501 INFO [train.py:904] (2/8) Epoch 13, batch 900, loss[loss=0.1716, simple_loss=0.2572, pruned_loss=0.04304, over 16667.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2597, pruned_loss=0.04805, over 3277211.18 frames. ], batch size: 62, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:35:33,155 INFO [train.py:904] (2/8) Epoch 13, batch 950, loss[loss=0.1825, simple_loss=0.261, pruned_loss=0.05199, over 16698.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2599, pruned_loss=0.04787, over 3285374.32 frames. ], batch size: 89, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:36:02,667 INFO [optim.py:368] (2/8) 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 17:36:03,748 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 17:36:43,313 INFO [train.py:904] (2/8) Epoch 13, batch 1000, loss[loss=0.1578, simple_loss=0.2368, pruned_loss=0.03941, over 16857.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2586, pruned_loss=0.04788, over 3280588.70 frames. ], batch size: 102, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:37:09,706 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:14,621 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:52,043 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:52,802 INFO [train.py:904] (2/8) Epoch 13, batch 1050, loss[loss=0.1859, simple_loss=0.2778, pruned_loss=0.047, over 17104.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.258, pruned_loss=0.04724, over 3294490.59 frames. ], batch size: 53, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:38:20,475 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:38:21,179 INFO [optim.py:368] (2/8) 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 17:38:53,031 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8592, 3.0900, 3.1337, 2.1201, 2.7796, 2.1808, 3.3791, 3.3384], device='cuda:2'), covar=tensor([0.0252, 0.0811, 0.0552, 0.1653, 0.0793, 0.0961, 0.0524, 0.0799], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0146, 0.0159, 0.0145, 0.0139, 0.0125, 0.0137, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 17:38:58,267 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:39:01,457 INFO [train.py:904] (2/8) Epoch 13, batch 1100, loss[loss=0.1712, simple_loss=0.2683, pruned_loss=0.03704, over 17283.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.257, pruned_loss=0.04676, over 3295647.64 frames. ], batch size: 52, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:39:23,976 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5112, 4.9251, 5.3089, 5.3262, 5.2988, 4.8663, 4.5274, 4.6588], device='cuda:2'), covar=tensor([0.0675, 0.0599, 0.0570, 0.0632, 0.0604, 0.0683, 0.1370, 0.0525], device='cuda:2'), in_proj_covar=tensor([0.0353, 0.0372, 0.0369, 0.0352, 0.0418, 0.0396, 0.0491, 0.0318], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 17:39:39,857 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2878, 2.5776, 2.1188, 2.3401, 2.9045, 2.7006, 3.1613, 3.0968], device='cuda:2'), covar=tensor([0.0177, 0.0306, 0.0406, 0.0331, 0.0189, 0.0273, 0.0232, 0.0191], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0219, 0.0211, 0.0210, 0.0217, 0.0215, 0.0222, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:39:45,807 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:40:11,764 INFO [train.py:904] (2/8) Epoch 13, batch 1150, loss[loss=0.177, simple_loss=0.2495, pruned_loss=0.05231, over 16840.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2567, pruned_loss=0.04628, over 3305995.28 frames. ], batch size: 83, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:40:39,512 INFO [optim.py:368] (2/8) 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 17:40:43,377 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:41:20,393 INFO [train.py:904] (2/8) Epoch 13, batch 1200, loss[loss=0.1553, simple_loss=0.2432, pruned_loss=0.03373, over 17233.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2569, pruned_loss=0.04588, over 3319017.07 frames. ], batch size: 44, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:41:47,884 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1748, 5.1817, 4.9621, 4.3867, 4.9575, 1.7969, 4.7573, 4.9462], device='cuda:2'), covar=tensor([0.0095, 0.0071, 0.0172, 0.0423, 0.0100, 0.2506, 0.0141, 0.0195], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0127, 0.0172, 0.0161, 0.0145, 0.0188, 0.0161, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:41:50,192 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:42:17,204 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7868, 3.1428, 2.6501, 4.5844, 3.7301, 4.2286, 1.6699, 3.0361], device='cuda:2'), covar=tensor([0.1481, 0.0666, 0.1195, 0.0218, 0.0291, 0.0423, 0.1659, 0.0789], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0160, 0.0183, 0.0152, 0.0194, 0.0210, 0.0183, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 17:42:30,139 INFO [train.py:904] (2/8) Epoch 13, batch 1250, loss[loss=0.181, simple_loss=0.2434, pruned_loss=0.05934, over 16791.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2568, pruned_loss=0.04742, over 3314565.07 frames. ], batch size: 83, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:42:34,854 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9946, 2.5532, 1.9606, 2.2738, 2.9374, 2.7335, 3.0981, 3.0635], device='cuda:2'), covar=tensor([0.0154, 0.0294, 0.0414, 0.0352, 0.0162, 0.0223, 0.0215, 0.0190], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0219, 0.0211, 0.0210, 0.0218, 0.0215, 0.0222, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:42:59,815 INFO [optim.py:368] (2/8) 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 17:43:02,399 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 17:43:40,470 INFO [train.py:904] (2/8) Epoch 13, batch 1300, loss[loss=0.1458, simple_loss=0.233, pruned_loss=0.02933, over 16866.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2568, pruned_loss=0.04667, over 3323600.98 frames. ], batch size: 42, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:44:12,381 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1475, 5.2277, 4.9983, 4.6658, 4.3852, 5.0984, 5.1654, 4.6511], device='cuda:2'), covar=tensor([0.0845, 0.0437, 0.0426, 0.0415, 0.1397, 0.0546, 0.0300, 0.0819], device='cuda:2'), in_proj_covar=tensor([0.0263, 0.0355, 0.0316, 0.0295, 0.0336, 0.0337, 0.0212, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:44:12,391 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:44:25,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7786, 4.6216, 4.6206, 4.2794, 4.2463, 4.6686, 4.6197, 4.4096], device='cuda:2'), covar=tensor([0.0667, 0.0773, 0.0330, 0.0307, 0.1044, 0.0513, 0.0454, 0.0725], device='cuda:2'), in_proj_covar=tensor([0.0263, 0.0355, 0.0316, 0.0295, 0.0336, 0.0337, 0.0212, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:44:43,626 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9146, 3.8564, 4.3132, 2.1929, 4.5238, 4.5172, 3.3104, 3.4685], device='cuda:2'), covar=tensor([0.0667, 0.0211, 0.0185, 0.1031, 0.0060, 0.0124, 0.0340, 0.0364], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0102, 0.0091, 0.0139, 0.0069, 0.0110, 0.0122, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 17:44:49,682 INFO [train.py:904] (2/8) Epoch 13, batch 1350, loss[loss=0.1639, simple_loss=0.2476, pruned_loss=0.0401, over 16562.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2578, pruned_loss=0.04697, over 3328933.52 frames. ], batch size: 68, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:45:17,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:45:17,396 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.27 vs. limit=5.0 2023-04-29 17:45:17,976 INFO [optim.py:368] (2/8) 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 17:45:45,406 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0591, 5.5343, 5.7298, 5.5135, 5.5206, 6.1198, 5.5550, 5.3151], device='cuda:2'), covar=tensor([0.0815, 0.1792, 0.2026, 0.2024, 0.2869, 0.0857, 0.1487, 0.2388], device='cuda:2'), in_proj_covar=tensor([0.0369, 0.0521, 0.0571, 0.0454, 0.0611, 0.0594, 0.0451, 0.0604], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 17:45:58,577 INFO [train.py:904] (2/8) Epoch 13, batch 1400, loss[loss=0.1837, simple_loss=0.2532, pruned_loss=0.05714, over 16437.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2578, pruned_loss=0.04735, over 3319706.69 frames. ], batch size: 146, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:46:26,200 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-29 17:46:35,857 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:47:09,271 INFO [train.py:904] (2/8) Epoch 13, batch 1450, loss[loss=0.1714, simple_loss=0.264, pruned_loss=0.03936, over 17132.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2573, pruned_loss=0.04718, over 3328242.39 frames. ], batch size: 48, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:47:38,907 INFO [optim.py:368] (2/8) 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 17:47:59,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1833, 4.0320, 4.2265, 4.3668, 4.4609, 4.0315, 4.1899, 4.4384], device='cuda:2'), covar=tensor([0.1241, 0.0934, 0.1211, 0.0585, 0.0509, 0.1273, 0.2257, 0.0614], device='cuda:2'), in_proj_covar=tensor([0.0571, 0.0716, 0.0865, 0.0723, 0.0546, 0.0567, 0.0578, 0.0666], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:48:02,063 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3685, 4.6616, 4.4595, 4.4936, 4.2139, 4.1574, 4.2375, 4.6860], device='cuda:2'), covar=tensor([0.1013, 0.0869, 0.0863, 0.0662, 0.0662, 0.1363, 0.0933, 0.0894], device='cuda:2'), in_proj_covar=tensor([0.0576, 0.0731, 0.0589, 0.0517, 0.0461, 0.0471, 0.0608, 0.0560], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:48:19,760 INFO [train.py:904] (2/8) Epoch 13, batch 1500, loss[loss=0.1912, simple_loss=0.2795, pruned_loss=0.05148, over 17083.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2577, pruned_loss=0.04771, over 3329394.99 frames. ], batch size: 55, lr: 5.33e-03, grad_scale: 4.0 2023-04-29 17:48:37,146 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:48:37,238 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8851, 2.9769, 2.5907, 2.7284, 3.1731, 3.0307, 3.6146, 3.4007], device='cuda:2'), covar=tensor([0.0077, 0.0285, 0.0345, 0.0302, 0.0220, 0.0264, 0.0196, 0.0192], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0218, 0.0209, 0.0209, 0.0218, 0.0216, 0.0223, 0.0206], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:49:30,720 INFO [train.py:904] (2/8) Epoch 13, batch 1550, loss[loss=0.2019, simple_loss=0.2645, pruned_loss=0.06966, over 16761.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2593, pruned_loss=0.04859, over 3322852.76 frames. ], batch size: 124, lr: 5.32e-03, grad_scale: 4.0 2023-04-29 17:49:41,543 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7787, 3.7764, 4.1499, 2.1450, 4.2572, 4.2307, 3.2530, 3.3098], device='cuda:2'), covar=tensor([0.0708, 0.0191, 0.0160, 0.1094, 0.0065, 0.0152, 0.0378, 0.0376], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0102, 0.0091, 0.0139, 0.0070, 0.0111, 0.0122, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 17:50:00,254 INFO [optim.py:368] (2/8) 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 17:50:01,762 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:50:31,279 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3439, 2.1979, 1.5491, 1.9048, 2.4868, 2.3084, 2.6368, 2.5835], device='cuda:2'), covar=tensor([0.0180, 0.0380, 0.0546, 0.0451, 0.0250, 0.0318, 0.0185, 0.0259], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0218, 0.0210, 0.0209, 0.0218, 0.0216, 0.0223, 0.0206], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:50:39,399 INFO [train.py:904] (2/8) Epoch 13, batch 1600, loss[loss=0.2016, simple_loss=0.2809, pruned_loss=0.06114, over 16478.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.261, pruned_loss=0.04926, over 3319571.52 frames. ], batch size: 68, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:51:14,490 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:51:43,117 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7140, 3.8154, 3.0845, 2.2518, 2.5491, 2.2960, 3.8780, 3.4838], device='cuda:2'), covar=tensor([0.2397, 0.0591, 0.1283, 0.2517, 0.2435, 0.1828, 0.0505, 0.1157], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0260, 0.0287, 0.0281, 0.0278, 0.0226, 0.0271, 0.0301], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:51:47,303 INFO [train.py:904] (2/8) Epoch 13, batch 1650, loss[loss=0.1993, simple_loss=0.2699, pruned_loss=0.06432, over 16765.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2612, pruned_loss=0.04908, over 3326843.09 frames. ], batch size: 124, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:52:18,026 INFO [optim.py:368] (2/8) 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 17:52:38,171 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:52:55,812 INFO [train.py:904] (2/8) Epoch 13, batch 1700, loss[loss=0.1755, simple_loss=0.2796, pruned_loss=0.03572, over 17031.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2625, pruned_loss=0.04926, over 3321036.55 frames. ], batch size: 50, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:53:31,975 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:54:04,789 INFO [train.py:904] (2/8) Epoch 13, batch 1750, loss[loss=0.179, simple_loss=0.259, pruned_loss=0.04949, over 15987.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2635, pruned_loss=0.04944, over 3315215.05 frames. ], batch size: 35, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:54:34,129 INFO [optim.py:368] (2/8) 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 17:54:37,987 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:54:40,345 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-29 17:55:14,401 INFO [train.py:904] (2/8) Epoch 13, batch 1800, loss[loss=0.2016, simple_loss=0.2728, pruned_loss=0.06523, over 16928.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2652, pruned_loss=0.04978, over 3317913.63 frames. ], batch size: 116, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:55:53,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7130, 2.5075, 2.0739, 2.3798, 2.8526, 2.6941, 3.4283, 3.0619], device='cuda:2'), covar=tensor([0.0091, 0.0341, 0.0418, 0.0361, 0.0238, 0.0310, 0.0172, 0.0243], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0216, 0.0208, 0.0207, 0.0216, 0.0215, 0.0222, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:56:23,364 INFO [train.py:904] (2/8) Epoch 13, batch 1850, loss[loss=0.1742, simple_loss=0.2567, pruned_loss=0.04589, over 16794.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2662, pruned_loss=0.04986, over 3321480.44 frames. ], batch size: 83, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:56:47,528 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:56:52,512 INFO [optim.py:368] (2/8) 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 17:57:33,302 INFO [train.py:904] (2/8) Epoch 13, batch 1900, loss[loss=0.1614, simple_loss=0.2392, pruned_loss=0.04185, over 16853.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.265, pruned_loss=0.04871, over 3323985.06 frames. ], batch size: 42, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:57:45,229 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 17:58:17,686 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7828, 1.8530, 2.3134, 2.7053, 2.7015, 2.6800, 1.8568, 2.9556], device='cuda:2'), covar=tensor([0.0141, 0.0346, 0.0260, 0.0209, 0.0201, 0.0207, 0.0356, 0.0107], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0176, 0.0161, 0.0166, 0.0176, 0.0129, 0.0177, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:58:35,057 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3980, 3.3317, 2.6584, 2.0727, 2.1914, 2.1630, 3.3544, 3.0670], device='cuda:2'), covar=tensor([0.2606, 0.0719, 0.1664, 0.2494, 0.2402, 0.1861, 0.0586, 0.1243], device='cuda:2'), in_proj_covar=tensor([0.0306, 0.0263, 0.0290, 0.0284, 0.0283, 0.0228, 0.0273, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 17:58:39,328 INFO [train.py:904] (2/8) Epoch 13, batch 1950, loss[loss=0.2042, simple_loss=0.2854, pruned_loss=0.06148, over 12316.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2658, pruned_loss=0.0487, over 3314938.91 frames. ], batch size: 246, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:40,954 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5239, 3.7066, 4.1103, 2.1151, 4.2324, 4.2060, 3.1387, 3.0995], device='cuda:2'), covar=tensor([0.0818, 0.0198, 0.0164, 0.1138, 0.0065, 0.0205, 0.0416, 0.0483], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0102, 0.0091, 0.0139, 0.0071, 0.0111, 0.0122, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 17:59:09,937 INFO [optim.py:368] (2/8) 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 17:59:25,304 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:59:48,916 INFO [train.py:904] (2/8) Epoch 13, batch 2000, loss[loss=0.185, simple_loss=0.2604, pruned_loss=0.05479, over 16824.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2645, pruned_loss=0.04816, over 3317109.45 frames. ], batch size: 102, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 17:59:52,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1695, 5.1225, 4.9443, 4.4018, 5.0517, 2.0874, 4.7611, 4.8853], device='cuda:2'), covar=tensor([0.0084, 0.0086, 0.0165, 0.0350, 0.0081, 0.2317, 0.0131, 0.0189], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0131, 0.0178, 0.0166, 0.0149, 0.0192, 0.0167, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:00:34,152 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-29 18:00:54,543 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6566, 3.7386, 2.8059, 2.1773, 2.4527, 2.2509, 3.7397, 3.3128], device='cuda:2'), covar=tensor([0.2332, 0.0627, 0.1533, 0.2577, 0.2480, 0.1746, 0.0529, 0.1307], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0262, 0.0288, 0.0283, 0.0282, 0.0228, 0.0272, 0.0306], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:00:59,626 INFO [train.py:904] (2/8) Epoch 13, batch 2050, loss[loss=0.1389, simple_loss=0.2175, pruned_loss=0.03018, over 16741.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2646, pruned_loss=0.04875, over 3310764.85 frames. ], batch size: 39, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:01:28,711 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:01:30,782 INFO [optim.py:368] (2/8) 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 18:01:39,520 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2481, 4.1780, 4.6303, 2.5093, 4.7914, 4.8413, 3.4714, 3.8506], device='cuda:2'), covar=tensor([0.0641, 0.0178, 0.0141, 0.1011, 0.0052, 0.0128, 0.0386, 0.0331], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0101, 0.0090, 0.0138, 0.0070, 0.0111, 0.0121, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 18:02:09,974 INFO [train.py:904] (2/8) Epoch 13, batch 2100, loss[loss=0.2167, simple_loss=0.2983, pruned_loss=0.06754, over 16759.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2651, pruned_loss=0.04891, over 3313481.27 frames. ], batch size: 124, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:02:54,422 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:14,047 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-29 18:03:20,324 INFO [train.py:904] (2/8) Epoch 13, batch 2150, loss[loss=0.1983, simple_loss=0.2799, pruned_loss=0.0584, over 16720.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2664, pruned_loss=0.04994, over 3318566.79 frames. ], batch size: 134, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:03:38,128 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 18:03:44,931 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:50,024 INFO [optim.py:368] (2/8) 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 18:04:14,061 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4153, 1.6099, 2.0512, 2.1673, 2.4040, 2.3515, 1.6339, 2.4704], device='cuda:2'), covar=tensor([0.0149, 0.0369, 0.0220, 0.0235, 0.0205, 0.0203, 0.0355, 0.0090], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0177, 0.0161, 0.0166, 0.0176, 0.0130, 0.0176, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:04:30,743 INFO [train.py:904] (2/8) Epoch 13, batch 2200, loss[loss=0.1871, simple_loss=0.2748, pruned_loss=0.04966, over 17144.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2669, pruned_loss=0.05032, over 3323634.38 frames. ], batch size: 49, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:04:53,398 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:04:54,875 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6421, 1.9624, 2.2375, 4.3506, 1.9475, 2.3228, 2.0923, 2.1345], device='cuda:2'), covar=tensor([0.1138, 0.4407, 0.2706, 0.0472, 0.5079, 0.3052, 0.3715, 0.4289], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0405, 0.0340, 0.0326, 0.0417, 0.0466, 0.0370, 0.0474], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:05:01,655 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8109, 2.2520, 2.3569, 4.5586, 2.2307, 2.7730, 2.4007, 2.4747], device='cuda:2'), covar=tensor([0.0902, 0.3359, 0.2373, 0.0369, 0.3647, 0.2241, 0.3003, 0.3174], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0405, 0.0340, 0.0326, 0.0417, 0.0466, 0.0370, 0.0474], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:05:19,580 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:28,727 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5001, 4.7885, 4.5708, 4.5796, 4.3313, 4.2405, 4.2967, 4.8483], device='cuda:2'), covar=tensor([0.1110, 0.0802, 0.1055, 0.0730, 0.0818, 0.1333, 0.1077, 0.0884], device='cuda:2'), in_proj_covar=tensor([0.0579, 0.0733, 0.0588, 0.0519, 0.0463, 0.0468, 0.0609, 0.0558], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:05:40,212 INFO [train.py:904] (2/8) Epoch 13, batch 2250, loss[loss=0.1529, simple_loss=0.2454, pruned_loss=0.03021, over 17109.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2663, pruned_loss=0.0492, over 3335160.05 frames. ], batch size: 47, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:05:53,614 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124062.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:09,552 INFO [optim.py:368] (2/8) 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 18:06:22,667 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:42,882 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:46,195 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2499, 3.2852, 1.9441, 3.4476, 2.4867, 3.4340, 2.0997, 2.7003], device='cuda:2'), covar=tensor([0.0223, 0.0382, 0.1414, 0.0263, 0.0749, 0.0677, 0.1183, 0.0624], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0169, 0.0191, 0.0146, 0.0172, 0.0214, 0.0199, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 18:06:48,018 INFO [train.py:904] (2/8) Epoch 13, batch 2300, loss[loss=0.188, simple_loss=0.2795, pruned_loss=0.04832, over 17050.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2667, pruned_loss=0.04967, over 3335609.54 frames. ], batch size: 53, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:07:17,832 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:07:56,749 INFO [train.py:904] (2/8) Epoch 13, batch 2350, loss[loss=0.1772, simple_loss=0.2518, pruned_loss=0.0513, over 16828.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.267, pruned_loss=0.05028, over 3332596.32 frames. ], batch size: 96, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:08:26,417 INFO [optim.py:368] (2/8) 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 18:08:50,598 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9334, 2.5107, 2.6854, 1.8003, 2.7700, 2.7864, 2.3953, 2.3453], device='cuda:2'), covar=tensor([0.0751, 0.0237, 0.0209, 0.0931, 0.0095, 0.0249, 0.0434, 0.0449], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0102, 0.0091, 0.0139, 0.0071, 0.0112, 0.0123, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 18:08:53,070 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8027, 3.1782, 3.1301, 2.0685, 2.7504, 2.2890, 3.3529, 3.4243], device='cuda:2'), covar=tensor([0.0235, 0.0726, 0.0531, 0.1594, 0.0761, 0.0867, 0.0532, 0.0724], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0151, 0.0161, 0.0147, 0.0139, 0.0126, 0.0139, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 18:09:06,167 INFO [train.py:904] (2/8) Epoch 13, batch 2400, loss[loss=0.1827, simple_loss=0.2746, pruned_loss=0.04543, over 17246.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2684, pruned_loss=0.05118, over 3321223.57 frames. ], batch size: 45, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:09:42,796 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:09:48,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9585, 4.4298, 3.2254, 2.4329, 2.9111, 2.5837, 4.7506, 3.8525], device='cuda:2'), covar=tensor([0.2449, 0.0551, 0.1536, 0.2389, 0.2701, 0.1726, 0.0360, 0.1099], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0262, 0.0289, 0.0285, 0.0285, 0.0229, 0.0272, 0.0309], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 18:10:15,680 INFO [train.py:904] (2/8) Epoch 13, batch 2450, loss[loss=0.1659, simple_loss=0.2634, pruned_loss=0.03421, over 17122.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2685, pruned_loss=0.05044, over 3321341.30 frames. ], batch size: 48, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:10:28,304 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9231, 2.8903, 2.5178, 2.7829, 3.1474, 3.0234, 3.7070, 3.3559], device='cuda:2'), covar=tensor([0.0077, 0.0262, 0.0326, 0.0279, 0.0195, 0.0252, 0.0163, 0.0177], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0218, 0.0208, 0.0209, 0.0218, 0.0216, 0.0225, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:10:43,319 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:10:46,034 INFO [optim.py:368] (2/8) 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 18:10:46,476 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0200, 1.9947, 2.5310, 2.8720, 2.9287, 3.4501, 2.2415, 3.4099], device='cuda:2'), covar=tensor([0.0190, 0.0394, 0.0251, 0.0248, 0.0212, 0.0143, 0.0373, 0.0104], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0179, 0.0163, 0.0168, 0.0177, 0.0131, 0.0178, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:10:59,161 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9556, 4.9451, 4.7505, 4.2969, 4.8836, 2.0432, 4.6522, 4.6073], device='cuda:2'), covar=tensor([0.0095, 0.0075, 0.0160, 0.0301, 0.0077, 0.2320, 0.0127, 0.0159], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0133, 0.0179, 0.0169, 0.0152, 0.0193, 0.0169, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:11:24,236 INFO [train.py:904] (2/8) Epoch 13, batch 2500, loss[loss=0.1658, simple_loss=0.2504, pruned_loss=0.04065, over 16427.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2681, pruned_loss=0.04983, over 3330949.80 frames. ], batch size: 68, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:11:26,582 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 18:12:09,454 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:12:22,375 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2102, 1.4464, 1.9666, 2.0942, 2.2559, 2.2651, 1.5897, 2.2698], device='cuda:2'), covar=tensor([0.0184, 0.0402, 0.0205, 0.0250, 0.0219, 0.0178, 0.0396, 0.0119], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0180, 0.0163, 0.0170, 0.0179, 0.0132, 0.0179, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:12:31,501 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:12:37,095 INFO [train.py:904] (2/8) Epoch 13, batch 2550, loss[loss=0.1854, simple_loss=0.2779, pruned_loss=0.0465, over 17014.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2687, pruned_loss=0.05038, over 3321334.76 frames. ], batch size: 55, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:12:46,350 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9512, 3.1635, 2.6052, 4.5134, 3.7050, 4.1887, 1.6293, 3.0849], device='cuda:2'), covar=tensor([0.1255, 0.0578, 0.1105, 0.0159, 0.0238, 0.0437, 0.1501, 0.0731], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0162, 0.0185, 0.0158, 0.0198, 0.0211, 0.0185, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 18:13:06,416 INFO [optim.py:368] (2/8) 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 18:13:26,694 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6560, 4.6759, 5.0776, 5.0665, 5.1167, 4.7579, 4.7688, 4.5277], device='cuda:2'), covar=tensor([0.0291, 0.0469, 0.0408, 0.0428, 0.0421, 0.0338, 0.0761, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0359, 0.0377, 0.0378, 0.0358, 0.0425, 0.0402, 0.0499, 0.0321], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 18:13:26,811 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9262, 1.8948, 2.5074, 2.7969, 2.7611, 3.1296, 2.0548, 3.1556], device='cuda:2'), covar=tensor([0.0164, 0.0384, 0.0229, 0.0218, 0.0221, 0.0140, 0.0377, 0.0106], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0179, 0.0163, 0.0169, 0.0178, 0.0132, 0.0178, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:13:31,432 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:13:44,303 INFO [train.py:904] (2/8) Epoch 13, batch 2600, loss[loss=0.1891, simple_loss=0.2701, pruned_loss=0.05406, over 16491.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2683, pruned_loss=0.04937, over 3321301.71 frames. ], batch size: 75, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:54,841 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:14:06,289 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:26,630 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:54,847 INFO [train.py:904] (2/8) Epoch 13, batch 2650, loss[loss=0.1777, simple_loss=0.2711, pruned_loss=0.04214, over 16561.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2694, pruned_loss=0.0494, over 3325339.08 frames. ], batch size: 75, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:15:06,660 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9496, 3.9811, 4.4008, 4.3775, 4.4043, 4.0798, 4.1435, 4.0220], device='cuda:2'), covar=tensor([0.0346, 0.0589, 0.0359, 0.0397, 0.0404, 0.0406, 0.0739, 0.0528], device='cuda:2'), in_proj_covar=tensor([0.0363, 0.0382, 0.0381, 0.0361, 0.0431, 0.0406, 0.0504, 0.0324], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 18:15:25,988 INFO [optim.py:368] (2/8) 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 18:15:49,370 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-29 18:15:52,717 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:16:04,556 INFO [train.py:904] (2/8) Epoch 13, batch 2700, loss[loss=0.1843, simple_loss=0.2757, pruned_loss=0.04644, over 16748.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2702, pruned_loss=0.04924, over 3328162.19 frames. ], batch size: 62, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:16:40,362 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:17:13,323 INFO [train.py:904] (2/8) Epoch 13, batch 2750, loss[loss=0.1689, simple_loss=0.2713, pruned_loss=0.03326, over 17121.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2692, pruned_loss=0.04806, over 3330250.69 frames. ], batch size: 49, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:17:20,298 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8161, 2.7075, 2.3910, 2.4838, 2.9307, 2.8318, 3.5694, 3.2781], device='cuda:2'), covar=tensor([0.0095, 0.0324, 0.0384, 0.0370, 0.0252, 0.0337, 0.0178, 0.0206], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0218, 0.0209, 0.0210, 0.0220, 0.0216, 0.0227, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:17:23,331 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 18:17:32,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0425, 4.9939, 5.5149, 5.5089, 5.5308, 5.1479, 5.1216, 4.9283], device='cuda:2'), covar=tensor([0.0276, 0.0510, 0.0324, 0.0369, 0.0371, 0.0323, 0.0840, 0.0379], device='cuda:2'), in_proj_covar=tensor([0.0365, 0.0385, 0.0383, 0.0362, 0.0434, 0.0408, 0.0507, 0.0325], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 18:17:44,014 INFO [optim.py:368] (2/8) 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 18:17:46,592 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:18:10,158 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8042, 4.7572, 4.6728, 4.1835, 4.7153, 2.0020, 4.5149, 4.4757], device='cuda:2'), covar=tensor([0.0075, 0.0069, 0.0137, 0.0277, 0.0070, 0.2216, 0.0107, 0.0151], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0132, 0.0178, 0.0168, 0.0150, 0.0190, 0.0167, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:18:22,959 INFO [train.py:904] (2/8) Epoch 13, batch 2800, loss[loss=0.2071, simple_loss=0.2866, pruned_loss=0.06382, over 11988.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2694, pruned_loss=0.04845, over 3331871.63 frames. ], batch size: 248, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:18:24,376 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7747, 4.9516, 5.1352, 4.9938, 4.9248, 5.5903, 5.1308, 4.7891], device='cuda:2'), covar=tensor([0.1153, 0.1834, 0.1937, 0.2018, 0.3003, 0.1009, 0.1451, 0.2417], device='cuda:2'), in_proj_covar=tensor([0.0375, 0.0534, 0.0580, 0.0462, 0.0625, 0.0607, 0.0460, 0.0612], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 18:18:43,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0432, 4.9505, 4.8049, 4.3031, 4.8927, 1.9643, 4.6593, 4.7055], device='cuda:2'), covar=tensor([0.0074, 0.0069, 0.0152, 0.0308, 0.0072, 0.2366, 0.0117, 0.0160], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0132, 0.0179, 0.0168, 0.0151, 0.0191, 0.0168, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:18:59,127 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:12,862 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:31,049 INFO [train.py:904] (2/8) Epoch 13, batch 2850, loss[loss=0.1977, simple_loss=0.2874, pruned_loss=0.05396, over 16588.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.269, pruned_loss=0.04885, over 3322171.17 frames. ], batch size: 62, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:19:46,019 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:00,128 INFO [optim.py:368] (2/8) 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 18:20:24,483 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:33,928 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:37,524 INFO [train.py:904] (2/8) Epoch 13, batch 2900, loss[loss=0.1851, simple_loss=0.2598, pruned_loss=0.05524, over 16825.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.268, pruned_loss=0.04969, over 3317711.01 frames. ], batch size: 102, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:20:40,111 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:20:43,566 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9062, 4.1623, 2.4060, 4.7119, 3.0699, 4.6340, 2.4571, 3.3179], device='cuda:2'), covar=tensor([0.0255, 0.0323, 0.1407, 0.0141, 0.0686, 0.0398, 0.1407, 0.0639], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0170, 0.0191, 0.0145, 0.0172, 0.0216, 0.0201, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 18:20:58,998 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:07,393 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:29,781 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:46,234 INFO [train.py:904] (2/8) Epoch 13, batch 2950, loss[loss=0.1558, simple_loss=0.2472, pruned_loss=0.03225, over 17230.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2656, pruned_loss=0.04917, over 3321511.39 frames. ], batch size: 44, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:05,798 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:16,911 INFO [optim.py:368] (2/8) 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 18:22:35,658 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:55,199 INFO [train.py:904] (2/8) Epoch 13, batch 3000, loss[loss=0.1806, simple_loss=0.274, pruned_loss=0.04356, over 16650.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2657, pruned_loss=0.04918, over 3329318.16 frames. ], batch size: 62, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:55,199 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 18:23:03,995 INFO [train.py:938] (2/8) Epoch 13, validation: loss=0.1391, simple_loss=0.2452, pruned_loss=0.01648, over 944034.00 frames. 2023-04-29 18:23:03,996 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 18:23:29,484 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:23:44,260 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4871, 3.6202, 1.9820, 3.8236, 2.6564, 3.7836, 2.1555, 2.8438], device='cuda:2'), covar=tensor([0.0210, 0.0317, 0.1417, 0.0217, 0.0728, 0.0533, 0.1282, 0.0626], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0168, 0.0189, 0.0144, 0.0171, 0.0214, 0.0198, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 18:24:14,168 INFO [train.py:904] (2/8) Epoch 13, batch 3050, loss[loss=0.1831, simple_loss=0.2561, pruned_loss=0.05505, over 16958.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2662, pruned_loss=0.04951, over 3332028.84 frames. ], batch size: 90, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:24:42,572 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 18:24:44,775 INFO [optim.py:368] (2/8) 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 18:24:46,296 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1727, 4.0988, 4.2843, 4.0827, 4.1588, 4.7421, 4.3512, 3.9771], device='cuda:2'), covar=tensor([0.1929, 0.2332, 0.2161, 0.2554, 0.3022, 0.1341, 0.1697, 0.3077], device='cuda:2'), in_proj_covar=tensor([0.0378, 0.0534, 0.0581, 0.0463, 0.0627, 0.0606, 0.0461, 0.0615], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 18:24:55,353 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:25:25,280 INFO [train.py:904] (2/8) Epoch 13, batch 3100, loss[loss=0.1774, simple_loss=0.2642, pruned_loss=0.04524, over 17123.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2651, pruned_loss=0.0495, over 3328602.21 frames. ], batch size: 48, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:25:40,711 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1636, 4.5028, 4.5781, 3.3364, 3.8734, 4.4711, 3.9265, 2.8840], device='cuda:2'), covar=tensor([0.0336, 0.0046, 0.0026, 0.0254, 0.0080, 0.0066, 0.0071, 0.0313], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0073, 0.0073, 0.0128, 0.0084, 0.0093, 0.0084, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 18:25:43,683 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9241, 1.9335, 2.4519, 2.8310, 2.6281, 3.3171, 2.1296, 3.2502], device='cuda:2'), covar=tensor([0.0207, 0.0390, 0.0271, 0.0266, 0.0263, 0.0139, 0.0368, 0.0142], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0164, 0.0172, 0.0179, 0.0133, 0.0179, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:26:01,314 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:26:23,255 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-29 18:26:34,831 INFO [train.py:904] (2/8) Epoch 13, batch 3150, loss[loss=0.1763, simple_loss=0.2671, pruned_loss=0.04272, over 16673.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.264, pruned_loss=0.04853, over 3334796.59 frames. ], batch size: 57, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:05,835 INFO [optim.py:368] (2/8) 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 18:27:09,072 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124976.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:34,184 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:45,736 INFO [train.py:904] (2/8) Epoch 13, batch 3200, loss[loss=0.1444, simple_loss=0.2278, pruned_loss=0.0305, over 16740.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2623, pruned_loss=0.04818, over 3333956.86 frames. ], batch size: 39, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:48,505 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:28:10,523 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:28:28,139 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8030, 3.7343, 4.2731, 2.0873, 4.3459, 4.4102, 3.1487, 3.3524], device='cuda:2'), covar=tensor([0.0682, 0.0213, 0.0156, 0.1046, 0.0068, 0.0145, 0.0350, 0.0355], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0102, 0.0090, 0.0139, 0.0071, 0.0112, 0.0122, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 18:28:56,246 INFO [train.py:904] (2/8) Epoch 13, batch 3250, loss[loss=0.1661, simple_loss=0.2588, pruned_loss=0.03664, over 17060.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.264, pruned_loss=0.04886, over 3328558.33 frames. ], batch size: 50, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:28:56,535 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:29:27,126 INFO [optim.py:368] (2/8) 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 18:29:46,252 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:03,758 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0504, 4.1532, 2.6029, 4.7956, 3.2327, 4.7387, 2.6108, 3.3700], device='cuda:2'), covar=tensor([0.0232, 0.0345, 0.1375, 0.0150, 0.0698, 0.0393, 0.1339, 0.0670], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0170, 0.0191, 0.0146, 0.0171, 0.0216, 0.0199, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 18:30:05,563 INFO [train.py:904] (2/8) Epoch 13, batch 3300, loss[loss=0.183, simple_loss=0.2778, pruned_loss=0.04408, over 17125.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2645, pruned_loss=0.04839, over 3328019.74 frames. ], batch size: 48, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:30:05,991 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0313, 4.0953, 4.4423, 4.4473, 4.4341, 4.1684, 4.1940, 4.0575], device='cuda:2'), covar=tensor([0.0354, 0.0629, 0.0394, 0.0415, 0.0484, 0.0424, 0.0781, 0.0632], device='cuda:2'), in_proj_covar=tensor([0.0367, 0.0387, 0.0386, 0.0365, 0.0434, 0.0407, 0.0511, 0.0327], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 18:30:35,565 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 18:30:42,370 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:54,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1267, 4.8013, 5.1225, 5.3651, 5.5429, 4.8161, 5.4891, 5.5131], device='cuda:2'), covar=tensor([0.1702, 0.1361, 0.1782, 0.0642, 0.0495, 0.0749, 0.0457, 0.0610], device='cuda:2'), in_proj_covar=tensor([0.0585, 0.0737, 0.0887, 0.0742, 0.0560, 0.0584, 0.0590, 0.0681], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:31:14,752 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-29 18:31:15,063 INFO [train.py:904] (2/8) Epoch 13, batch 3350, loss[loss=0.1573, simple_loss=0.239, pruned_loss=0.03779, over 16989.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2652, pruned_loss=0.04828, over 3329754.09 frames. ], batch size: 41, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:31:23,869 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0945, 5.1829, 5.5898, 5.5212, 5.6224, 5.3373, 5.1405, 5.0267], device='cuda:2'), covar=tensor([0.0439, 0.0784, 0.0623, 0.0956, 0.0604, 0.0540, 0.1273, 0.0476], device='cuda:2'), in_proj_covar=tensor([0.0363, 0.0385, 0.0382, 0.0362, 0.0431, 0.0405, 0.0507, 0.0325], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 18:31:45,937 INFO [optim.py:368] (2/8) 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 18:31:49,062 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:07,002 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:24,429 INFO [train.py:904] (2/8) Epoch 13, batch 3400, loss[loss=0.1828, simple_loss=0.2604, pruned_loss=0.0526, over 16880.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2649, pruned_loss=0.0481, over 3331001.85 frames. ], batch size: 90, lr: 5.29e-03, grad_scale: 4.0 2023-04-29 18:32:24,829 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1363, 3.1846, 3.4000, 2.2464, 3.0884, 3.4122, 3.0854, 1.9248], device='cuda:2'), covar=tensor([0.0417, 0.0090, 0.0041, 0.0311, 0.0092, 0.0085, 0.0082, 0.0390], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0074, 0.0072, 0.0129, 0.0085, 0.0093, 0.0084, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 18:33:35,396 INFO [train.py:904] (2/8) Epoch 13, batch 3450, loss[loss=0.1591, simple_loss=0.2512, pruned_loss=0.03346, over 17278.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.263, pruned_loss=0.04702, over 3321996.41 frames. ], batch size: 52, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:34:07,280 INFO [optim.py:368] (2/8) 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 18:34:34,806 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:34:39,529 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 18:34:46,233 INFO [train.py:904] (2/8) Epoch 13, batch 3500, loss[loss=0.181, simple_loss=0.2561, pruned_loss=0.05293, over 16847.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2614, pruned_loss=0.0466, over 3323320.39 frames. ], batch size: 90, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:35:09,912 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:35:57,848 INFO [train.py:904] (2/8) Epoch 13, batch 3550, loss[loss=0.1876, simple_loss=0.2677, pruned_loss=0.05372, over 15530.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2609, pruned_loss=0.04648, over 3318723.29 frames. ], batch size: 190, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:36:03,237 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:18,666 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:20,516 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6836, 4.6254, 4.6003, 4.1196, 4.6192, 1.9368, 4.3889, 4.3754], device='cuda:2'), covar=tensor([0.0097, 0.0084, 0.0148, 0.0297, 0.0090, 0.2339, 0.0129, 0.0171], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0134, 0.0181, 0.0170, 0.0152, 0.0191, 0.0170, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:36:30,595 INFO [optim.py:368] (2/8) 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 18:36:32,213 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:37:08,976 INFO [train.py:904] (2/8) Epoch 13, batch 3600, loss[loss=0.1996, simple_loss=0.27, pruned_loss=0.06458, over 16749.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2595, pruned_loss=0.0458, over 3319166.94 frames. ], batch size: 124, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:37:24,951 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-29 18:37:31,302 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:37:35,986 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:37:59,815 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:38:21,929 INFO [train.py:904] (2/8) Epoch 13, batch 3650, loss[loss=0.1708, simple_loss=0.2513, pruned_loss=0.0452, over 16385.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2585, pruned_loss=0.04643, over 3313063.48 frames. ], batch size: 68, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:38:57,386 INFO [optim.py:368] (2/8) 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 18:38:59,575 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:07,781 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:11,351 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:36,234 INFO [train.py:904] (2/8) Epoch 13, batch 3700, loss[loss=0.1797, simple_loss=0.2531, pruned_loss=0.05317, over 16710.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2577, pruned_loss=0.04805, over 3299920.04 frames. ], batch size: 124, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:40:10,904 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:40:51,172 INFO [train.py:904] (2/8) Epoch 13, batch 3750, loss[loss=0.1837, simple_loss=0.2468, pruned_loss=0.06032, over 16911.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2592, pruned_loss=0.05018, over 3271917.54 frames. ], batch size: 116, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:40:56,499 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3891, 2.1285, 2.2445, 4.1165, 2.1716, 2.6219, 2.2441, 2.3961], device='cuda:2'), covar=tensor([0.1078, 0.3416, 0.2368, 0.0410, 0.3416, 0.2234, 0.3278, 0.2730], device='cuda:2'), in_proj_covar=tensor([0.0374, 0.0409, 0.0340, 0.0327, 0.0421, 0.0472, 0.0372, 0.0479], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:41:24,163 INFO [optim.py:368] (2/8) 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 18:42:05,166 INFO [train.py:904] (2/8) Epoch 13, batch 3800, loss[loss=0.1885, simple_loss=0.2723, pruned_loss=0.05239, over 16640.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2601, pruned_loss=0.05141, over 3283063.60 frames. ], batch size: 62, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:42:07,428 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:43:18,396 INFO [train.py:904] (2/8) Epoch 13, batch 3850, loss[loss=0.2303, simple_loss=0.2947, pruned_loss=0.08299, over 16892.00 frames. ], tot_loss[loss=0.183, simple_loss=0.261, pruned_loss=0.05245, over 3274490.38 frames. ], batch size: 116, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:43:35,105 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:43:45,235 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 18:43:50,564 INFO [optim.py:368] (2/8) 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 18:44:29,761 INFO [train.py:904] (2/8) Epoch 13, batch 3900, loss[loss=0.1802, simple_loss=0.2607, pruned_loss=0.04992, over 15601.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2614, pruned_loss=0.05343, over 3263152.61 frames. ], batch size: 190, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:44:45,082 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 18:45:03,366 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:45:14,707 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125732.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:45:42,338 INFO [train.py:904] (2/8) Epoch 13, batch 3950, loss[loss=0.1829, simple_loss=0.2665, pruned_loss=0.04963, over 17212.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2608, pruned_loss=0.05372, over 3280960.97 frames. ], batch size: 43, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:46:16,761 INFO [optim.py:368] (2/8) 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 18:46:20,057 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:21,397 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8188, 4.0220, 3.0211, 2.3314, 2.8213, 2.4134, 4.0499, 3.6143], device='cuda:2'), covar=tensor([0.2337, 0.0556, 0.1417, 0.2624, 0.2351, 0.1759, 0.0483, 0.1100], device='cuda:2'), in_proj_covar=tensor([0.0303, 0.0258, 0.0288, 0.0285, 0.0285, 0.0227, 0.0273, 0.0305], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:46:29,448 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:32,686 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:51,383 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 18:46:55,502 INFO [train.py:904] (2/8) Epoch 13, batch 4000, loss[loss=0.1749, simple_loss=0.259, pruned_loss=0.04539, over 16372.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2613, pruned_loss=0.05381, over 3269933.38 frames. ], batch size: 35, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:47:37,206 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125832.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:48:05,369 INFO [train.py:904] (2/8) Epoch 13, batch 4050, loss[loss=0.1863, simple_loss=0.2735, pruned_loss=0.04953, over 16814.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2618, pruned_loss=0.05292, over 3262947.27 frames. ], batch size: 83, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:48:17,979 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6282, 3.0672, 2.9146, 1.6549, 2.5071, 2.0005, 3.2053, 3.2945], device='cuda:2'), covar=tensor([0.0307, 0.0803, 0.0866, 0.2574, 0.1188, 0.1232, 0.0691, 0.0829], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0153, 0.0160, 0.0147, 0.0138, 0.0127, 0.0138, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 18:48:36,495 INFO [optim.py:368] (2/8) 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 18:48:45,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7107, 3.6329, 4.4894, 1.8067, 4.7290, 4.8149, 3.1944, 3.4151], device='cuda:2'), covar=tensor([0.0779, 0.0265, 0.0129, 0.1248, 0.0035, 0.0054, 0.0377, 0.0432], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0138, 0.0070, 0.0111, 0.0122, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 18:49:15,115 INFO [train.py:904] (2/8) Epoch 13, batch 4100, loss[loss=0.1996, simple_loss=0.2764, pruned_loss=0.06135, over 12044.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2625, pruned_loss=0.05201, over 3259766.93 frames. ], batch size: 248, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:50:30,153 INFO [train.py:904] (2/8) Epoch 13, batch 4150, loss[loss=0.2164, simple_loss=0.3092, pruned_loss=0.06177, over 16182.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2707, pruned_loss=0.05517, over 3244406.75 frames. ], batch size: 165, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:50:42,582 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:05,850 INFO [optim.py:368] (2/8) 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 18:51:39,625 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 18:51:40,576 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:48,708 INFO [train.py:904] (2/8) Epoch 13, batch 4200, loss[loss=0.2041, simple_loss=0.295, pruned_loss=0.05655, over 16546.00 frames. ], tot_loss[loss=0.195, simple_loss=0.277, pruned_loss=0.05649, over 3207922.56 frames. ], batch size: 75, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:52:04,543 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 18:52:08,852 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8761, 4.1273, 3.9323, 4.0298, 3.7142, 3.7274, 3.8010, 4.1028], device='cuda:2'), covar=tensor([0.1075, 0.0955, 0.1004, 0.0752, 0.0826, 0.1598, 0.0944, 0.1092], device='cuda:2'), in_proj_covar=tensor([0.0570, 0.0723, 0.0582, 0.0512, 0.0458, 0.0463, 0.0602, 0.0554], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:52:34,267 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:52:35,687 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.8732, 6.1841, 5.9243, 6.1148, 5.6686, 5.3605, 5.5907, 6.2832], device='cuda:2'), covar=tensor([0.0866, 0.0718, 0.0809, 0.0593, 0.0727, 0.0591, 0.0965, 0.0813], device='cuda:2'), in_proj_covar=tensor([0.0569, 0.0722, 0.0581, 0.0510, 0.0457, 0.0463, 0.0600, 0.0554], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:53:02,528 INFO [train.py:904] (2/8) Epoch 13, batch 4250, loss[loss=0.2225, simple_loss=0.2899, pruned_loss=0.07757, over 12519.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2807, pruned_loss=0.05685, over 3193016.79 frames. ], batch size: 247, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:53:13,098 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:14,081 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:36,049 INFO [optim.py:368] (2/8) 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 18:53:39,755 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:39,825 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1466, 3.1613, 3.5173, 1.4622, 3.6399, 3.6955, 2.7921, 2.7165], device='cuda:2'), covar=tensor([0.0875, 0.0242, 0.0182, 0.1309, 0.0069, 0.0129, 0.0410, 0.0492], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0101, 0.0088, 0.0136, 0.0069, 0.0109, 0.0121, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 18:53:43,861 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:45,084 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:54:16,797 INFO [train.py:904] (2/8) Epoch 13, batch 4300, loss[loss=0.1913, simple_loss=0.2806, pruned_loss=0.051, over 16871.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2817, pruned_loss=0.05588, over 3194197.10 frames. ], batch size: 42, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:54:51,666 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:55:00,454 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7729, 5.2704, 5.4938, 5.2667, 5.3779, 5.8607, 5.3341, 5.0390], device='cuda:2'), covar=tensor([0.0891, 0.1397, 0.1591, 0.1497, 0.1913, 0.0758, 0.1176, 0.2074], device='cuda:2'), in_proj_covar=tensor([0.0363, 0.0509, 0.0552, 0.0436, 0.0589, 0.0578, 0.0440, 0.0590], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 18:55:09,292 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 18:55:25,951 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5055, 5.5284, 5.2479, 4.7565, 5.4920, 1.9289, 5.1853, 5.1832], device='cuda:2'), covar=tensor([0.0040, 0.0029, 0.0114, 0.0244, 0.0040, 0.2296, 0.0073, 0.0130], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0128, 0.0173, 0.0162, 0.0146, 0.0184, 0.0163, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:55:30,684 INFO [train.py:904] (2/8) Epoch 13, batch 4350, loss[loss=0.2126, simple_loss=0.2998, pruned_loss=0.06269, over 17185.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2848, pruned_loss=0.05664, over 3198854.85 frames. ], batch size: 46, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:56:06,023 INFO [optim.py:368] (2/8) 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 18:56:46,665 INFO [train.py:904] (2/8) Epoch 13, batch 4400, loss[loss=0.2023, simple_loss=0.2884, pruned_loss=0.05813, over 15341.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2869, pruned_loss=0.0581, over 3164559.65 frames. ], batch size: 190, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:57:59,272 INFO [train.py:904] (2/8) Epoch 13, batch 4450, loss[loss=0.2056, simple_loss=0.298, pruned_loss=0.05657, over 15461.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2898, pruned_loss=0.05918, over 3176837.96 frames. ], batch size: 190, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:58:03,159 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5813, 4.3841, 4.4958, 4.7614, 4.9269, 4.4497, 4.9125, 4.9355], device='cuda:2'), covar=tensor([0.1473, 0.1137, 0.1712, 0.0699, 0.0543, 0.0869, 0.0543, 0.0618], device='cuda:2'), in_proj_covar=tensor([0.0546, 0.0689, 0.0822, 0.0692, 0.0522, 0.0545, 0.0550, 0.0634], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:58:10,165 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:58:33,037 INFO [optim.py:368] (2/8) 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 18:59:14,448 INFO [train.py:904] (2/8) Epoch 13, batch 4500, loss[loss=0.2004, simple_loss=0.2861, pruned_loss=0.0574, over 16940.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2903, pruned_loss=0.05916, over 3193869.89 frames. ], batch size: 109, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:59:19,073 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0035, 5.0645, 4.8591, 4.5337, 4.5168, 4.9706, 4.8224, 4.5823], device='cuda:2'), covar=tensor([0.0463, 0.0220, 0.0197, 0.0211, 0.0782, 0.0222, 0.0239, 0.0538], device='cuda:2'), in_proj_covar=tensor([0.0254, 0.0341, 0.0300, 0.0282, 0.0324, 0.0324, 0.0206, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 18:59:21,401 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:25,572 INFO [train.py:904] (2/8) Epoch 13, batch 4550, loss[loss=0.206, simple_loss=0.2999, pruned_loss=0.05609, over 16813.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2916, pruned_loss=0.06015, over 3207240.07 frames. ], batch size: 102, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:00:28,830 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:59,195 INFO [optim.py:368] (2/8) 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 19:01:06,183 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:07,318 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:35,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7112, 4.7971, 4.5958, 4.2367, 4.2080, 4.6923, 4.4403, 4.3532], device='cuda:2'), covar=tensor([0.0449, 0.0225, 0.0187, 0.0231, 0.0806, 0.0240, 0.0381, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0253, 0.0340, 0.0299, 0.0281, 0.0324, 0.0324, 0.0206, 0.0351], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:01:36,158 INFO [train.py:904] (2/8) Epoch 13, batch 4600, loss[loss=0.1764, simple_loss=0.2729, pruned_loss=0.03993, over 16905.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2922, pruned_loss=0.06013, over 3216821.72 frames. ], batch size: 96, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:02:00,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5439, 2.5312, 2.3676, 3.2903, 2.6419, 3.6504, 1.3727, 2.7647], device='cuda:2'), covar=tensor([0.1354, 0.0701, 0.1190, 0.0178, 0.0231, 0.0350, 0.1677, 0.0741], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0161, 0.0183, 0.0157, 0.0198, 0.0206, 0.0182, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-29 19:02:22,725 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:23,033 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3177, 2.3705, 2.3800, 4.3129, 2.1580, 2.8132, 2.4050, 2.4600], device='cuda:2'), covar=tensor([0.1013, 0.2868, 0.2122, 0.0360, 0.3530, 0.1853, 0.2682, 0.2894], device='cuda:2'), in_proj_covar=tensor([0.0372, 0.0405, 0.0336, 0.0319, 0.0419, 0.0470, 0.0369, 0.0473], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:02:26,936 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-29 19:02:38,625 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:53,207 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:55,168 INFO [train.py:904] (2/8) Epoch 13, batch 4650, loss[loss=0.206, simple_loss=0.2884, pruned_loss=0.06181, over 16841.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2912, pruned_loss=0.05978, over 3226805.34 frames. ], batch size: 116, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:03:27,526 INFO [optim.py:368] (2/8) 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 19:03:35,734 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 19:04:07,652 INFO [train.py:904] (2/8) Epoch 13, batch 4700, loss[loss=0.2003, simple_loss=0.2783, pruned_loss=0.0612, over 11465.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2881, pruned_loss=0.0582, over 3222470.73 frames. ], batch size: 246, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:04:21,532 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:05:13,259 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-29 19:05:21,227 INFO [train.py:904] (2/8) Epoch 13, batch 4750, loss[loss=0.1946, simple_loss=0.2791, pruned_loss=0.05507, over 15338.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2838, pruned_loss=0.056, over 3226314.84 frames. ], batch size: 190, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:05:53,853 INFO [optim.py:368] (2/8) 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 19:06:32,270 INFO [train.py:904] (2/8) Epoch 13, batch 4800, loss[loss=0.1905, simple_loss=0.2851, pruned_loss=0.04792, over 16453.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2803, pruned_loss=0.05395, over 3243582.16 frames. ], batch size: 146, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:07:26,290 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 19:07:43,472 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5656, 2.4591, 2.2675, 3.5551, 2.3465, 3.8487, 1.3389, 2.7767], device='cuda:2'), covar=tensor([0.1423, 0.0761, 0.1271, 0.0127, 0.0124, 0.0322, 0.1683, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0163, 0.0185, 0.0159, 0.0201, 0.0208, 0.0184, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 19:07:46,139 INFO [train.py:904] (2/8) Epoch 13, batch 4850, loss[loss=0.2283, simple_loss=0.3109, pruned_loss=0.07288, over 12180.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2814, pruned_loss=0.05342, over 3232726.01 frames. ], batch size: 247, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:07:50,109 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 19:08:36,864 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1565, 3.8396, 3.9224, 2.4738, 3.4100, 3.7761, 3.4792, 2.1128], device='cuda:2'), covar=tensor([0.0469, 0.0029, 0.0027, 0.0343, 0.0075, 0.0100, 0.0071, 0.0405], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0070, 0.0071, 0.0126, 0.0083, 0.0091, 0.0082, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 19:08:59,235 INFO [train.py:904] (2/8) Epoch 13, batch 4900, loss[loss=0.1867, simple_loss=0.2752, pruned_loss=0.04909, over 16723.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.28, pruned_loss=0.05174, over 3238319.25 frames. ], batch size: 89, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:08:59,538 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:09:49,772 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:10:12,983 INFO [train.py:904] (2/8) Epoch 13, batch 4950, loss[loss=0.1705, simple_loss=0.2652, pruned_loss=0.03792, over 16729.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2794, pruned_loss=0.05134, over 3251957.05 frames. ], batch size: 89, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:10:39,209 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8355, 4.7884, 4.7351, 3.9564, 4.7570, 1.6633, 4.5082, 4.5279], device='cuda:2'), covar=tensor([0.0066, 0.0072, 0.0119, 0.0434, 0.0088, 0.2435, 0.0109, 0.0181], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0124, 0.0168, 0.0159, 0.0141, 0.0180, 0.0158, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:10:45,732 INFO [optim.py:368] (2/8) 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 19:11:26,260 INFO [train.py:904] (2/8) Epoch 13, batch 5000, loss[loss=0.2043, simple_loss=0.2956, pruned_loss=0.05654, over 16912.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2812, pruned_loss=0.05191, over 3241587.97 frames. ], batch size: 116, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:11:32,726 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:11:47,312 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:12:38,197 INFO [train.py:904] (2/8) Epoch 13, batch 5050, loss[loss=0.1855, simple_loss=0.2772, pruned_loss=0.04687, over 16869.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2821, pruned_loss=0.05196, over 3231799.47 frames. ], batch size: 42, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:13:11,123 INFO [optim.py:368] (2/8) 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 19:13:14,594 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:13:51,463 INFO [train.py:904] (2/8) Epoch 13, batch 5100, loss[loss=0.175, simple_loss=0.2548, pruned_loss=0.04762, over 16999.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2801, pruned_loss=0.0511, over 3233994.55 frames. ], batch size: 55, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:14:02,185 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-04-29 19:15:04,908 INFO [train.py:904] (2/8) Epoch 13, batch 5150, loss[loss=0.1869, simple_loss=0.2902, pruned_loss=0.04181, over 16797.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2802, pruned_loss=0.05061, over 3223223.68 frames. ], batch size: 102, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:15:14,468 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0990, 3.6637, 3.7745, 2.4218, 3.3656, 3.7108, 3.4239, 2.1188], device='cuda:2'), covar=tensor([0.0477, 0.0034, 0.0032, 0.0334, 0.0075, 0.0091, 0.0070, 0.0381], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0071, 0.0072, 0.0128, 0.0085, 0.0093, 0.0083, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 19:15:37,786 INFO [optim.py:368] (2/8) 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 19:16:05,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0737, 1.5475, 1.7989, 2.0278, 2.2085, 2.2803, 1.5648, 2.3075], device='cuda:2'), covar=tensor([0.0168, 0.0397, 0.0229, 0.0255, 0.0240, 0.0157, 0.0436, 0.0093], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0177, 0.0162, 0.0165, 0.0176, 0.0131, 0.0177, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:16:17,737 INFO [train.py:904] (2/8) Epoch 13, batch 5200, loss[loss=0.1896, simple_loss=0.2756, pruned_loss=0.05181, over 16621.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2788, pruned_loss=0.05009, over 3238658.21 frames. ], batch size: 134, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:16:25,736 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1660, 3.2707, 3.6569, 1.8066, 3.7025, 3.7654, 2.8168, 2.7081], device='cuda:2'), covar=tensor([0.0844, 0.0218, 0.0110, 0.1177, 0.0062, 0.0103, 0.0381, 0.0491], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0103, 0.0088, 0.0137, 0.0070, 0.0108, 0.0122, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 19:16:58,468 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6498, 4.9424, 4.6583, 4.7118, 4.3991, 4.3930, 4.3509, 4.9802], device='cuda:2'), covar=tensor([0.1144, 0.0856, 0.1055, 0.0746, 0.0906, 0.1080, 0.1024, 0.0911], device='cuda:2'), in_proj_covar=tensor([0.0562, 0.0705, 0.0571, 0.0498, 0.0447, 0.0451, 0.0586, 0.0543], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:17:08,362 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:17:31,265 INFO [train.py:904] (2/8) Epoch 13, batch 5250, loss[loss=0.1884, simple_loss=0.2784, pruned_loss=0.04915, over 16318.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2774, pruned_loss=0.05016, over 3225494.37 frames. ], batch size: 165, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:04,188 INFO [optim.py:368] (2/8) 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 19:18:18,851 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:18:44,212 INFO [train.py:904] (2/8) Epoch 13, batch 5300, loss[loss=0.1779, simple_loss=0.2593, pruned_loss=0.04822, over 16916.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.274, pruned_loss=0.04931, over 3215027.05 frames. ], batch size: 109, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:50,591 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:18:56,498 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 19:19:50,734 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8469, 4.8481, 4.7132, 3.9963, 4.7386, 1.9500, 4.4867, 4.5340], device='cuda:2'), covar=tensor([0.0082, 0.0068, 0.0117, 0.0407, 0.0082, 0.2218, 0.0109, 0.0189], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0125, 0.0170, 0.0162, 0.0143, 0.0183, 0.0160, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:19:58,319 INFO [train.py:904] (2/8) Epoch 13, batch 5350, loss[loss=0.2172, simple_loss=0.3061, pruned_loss=0.06418, over 16935.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2721, pruned_loss=0.04858, over 3227690.28 frames. ], batch size: 109, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:20:01,842 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:27,747 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:32,661 INFO [optim.py:368] (2/8) 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 19:21:10,971 INFO [train.py:904] (2/8) Epoch 13, batch 5400, loss[loss=0.1958, simple_loss=0.2867, pruned_loss=0.05238, over 16583.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2743, pruned_loss=0.04912, over 3209757.92 frames. ], batch size: 62, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:21:52,007 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0977, 4.9929, 5.5362, 5.4952, 5.5097, 5.1284, 5.0601, 4.7983], device='cuda:2'), covar=tensor([0.0246, 0.0438, 0.0300, 0.0358, 0.0478, 0.0306, 0.1034, 0.0418], device='cuda:2'), in_proj_covar=tensor([0.0349, 0.0366, 0.0364, 0.0354, 0.0416, 0.0389, 0.0487, 0.0313], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 19:22:11,351 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0017, 3.9913, 3.8964, 3.2441, 3.9385, 1.7651, 3.7259, 3.5040], device='cuda:2'), covar=tensor([0.0109, 0.0088, 0.0152, 0.0311, 0.0088, 0.2376, 0.0130, 0.0220], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0125, 0.0171, 0.0162, 0.0143, 0.0183, 0.0159, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:22:20,873 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:22:26,302 INFO [train.py:904] (2/8) Epoch 13, batch 5450, loss[loss=0.2401, simple_loss=0.3196, pruned_loss=0.0803, over 16642.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2776, pruned_loss=0.05098, over 3199434.52 frames. ], batch size: 134, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:22:45,202 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0611, 3.0720, 3.3195, 1.7185, 3.4334, 3.4960, 2.6653, 2.5758], device='cuda:2'), covar=tensor([0.0828, 0.0239, 0.0168, 0.1167, 0.0074, 0.0158, 0.0456, 0.0465], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0102, 0.0088, 0.0138, 0.0070, 0.0109, 0.0122, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 19:23:02,325 INFO [optim.py:368] (2/8) 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 19:23:43,866 INFO [train.py:904] (2/8) Epoch 13, batch 5500, loss[loss=0.273, simple_loss=0.3349, pruned_loss=0.1055, over 11597.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.286, pruned_loss=0.05682, over 3163662.85 frames. ], batch size: 247, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:57,072 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:25:00,716 INFO [train.py:904] (2/8) Epoch 13, batch 5550, loss[loss=0.2667, simple_loss=0.3287, pruned_loss=0.1023, over 11217.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2939, pruned_loss=0.06258, over 3128261.82 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:25:12,955 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4121, 1.6046, 2.0435, 2.3031, 2.3152, 2.6853, 1.7672, 2.6134], device='cuda:2'), covar=tensor([0.0157, 0.0386, 0.0230, 0.0237, 0.0237, 0.0138, 0.0367, 0.0087], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0176, 0.0161, 0.0164, 0.0174, 0.0130, 0.0175, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 19:25:15,934 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9048, 4.1398, 3.9285, 3.9664, 3.6922, 3.7464, 3.8350, 4.1127], device='cuda:2'), covar=tensor([0.0975, 0.0904, 0.1008, 0.0790, 0.0723, 0.1651, 0.0864, 0.1089], device='cuda:2'), in_proj_covar=tensor([0.0561, 0.0703, 0.0571, 0.0497, 0.0443, 0.0450, 0.0583, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:25:36,673 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9100, 3.3660, 3.2599, 1.9566, 2.7580, 2.2265, 3.4211, 3.5016], device='cuda:2'), covar=tensor([0.0281, 0.0646, 0.0589, 0.1855, 0.0826, 0.0932, 0.0673, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0149, 0.0160, 0.0146, 0.0139, 0.0126, 0.0139, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 19:25:38,573 INFO [optim.py:368] (2/8) 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 19:26:20,884 INFO [train.py:904] (2/8) Epoch 13, batch 5600, loss[loss=0.3041, simple_loss=0.3529, pruned_loss=0.1277, over 11628.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2986, pruned_loss=0.06669, over 3095733.48 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:27:02,207 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:27:13,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9328, 2.6074, 2.5091, 1.9293, 2.4788, 2.6347, 2.5447, 1.8816], device='cuda:2'), covar=tensor([0.0296, 0.0058, 0.0062, 0.0275, 0.0102, 0.0107, 0.0086, 0.0316], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0070, 0.0072, 0.0128, 0.0085, 0.0092, 0.0082, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 19:27:41,559 INFO [train.py:904] (2/8) Epoch 13, batch 5650, loss[loss=0.2327, simple_loss=0.3171, pruned_loss=0.07412, over 16771.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3041, pruned_loss=0.07136, over 3070263.09 frames. ], batch size: 83, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:27:55,509 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7302, 2.4761, 2.3127, 3.3099, 2.4053, 3.5911, 1.4480, 2.6249], device='cuda:2'), covar=tensor([0.1198, 0.0675, 0.1135, 0.0180, 0.0186, 0.0385, 0.1516, 0.0792], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0163, 0.0184, 0.0158, 0.0201, 0.0208, 0.0185, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 19:28:12,178 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:28:17,739 INFO [optim.py:368] (2/8) 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 19:28:21,242 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 19:28:31,241 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3502, 3.2989, 3.3743, 3.4962, 3.5107, 3.2420, 3.4914, 3.5561], device='cuda:2'), covar=tensor([0.1074, 0.0928, 0.1022, 0.0569, 0.0613, 0.2470, 0.1011, 0.0770], device='cuda:2'), in_proj_covar=tensor([0.0543, 0.0682, 0.0813, 0.0686, 0.0519, 0.0538, 0.0549, 0.0631], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:28:36,964 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:28:44,594 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1000, 3.0620, 3.2937, 1.7576, 3.4434, 3.5262, 2.6826, 2.6001], device='cuda:2'), covar=tensor([0.0843, 0.0248, 0.0176, 0.1128, 0.0070, 0.0148, 0.0462, 0.0447], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0102, 0.0088, 0.0136, 0.0069, 0.0109, 0.0121, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 19:28:58,352 INFO [train.py:904] (2/8) Epoch 13, batch 5700, loss[loss=0.2798, simple_loss=0.337, pruned_loss=0.1113, over 11451.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3057, pruned_loss=0.07349, over 3043092.58 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:29:27,678 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:29:41,971 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:29:51,082 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 19:30:00,075 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3130, 5.2956, 5.0279, 4.4386, 5.2226, 1.9678, 4.9673, 4.9463], device='cuda:2'), covar=tensor([0.0064, 0.0060, 0.0142, 0.0326, 0.0066, 0.2263, 0.0095, 0.0149], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0124, 0.0169, 0.0160, 0.0142, 0.0182, 0.0158, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:30:02,773 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.35 vs. limit=5.0 2023-04-29 19:30:06,644 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8205, 2.5950, 2.3807, 3.1464, 2.3197, 3.5765, 1.3598, 2.8070], device='cuda:2'), covar=tensor([0.1272, 0.0613, 0.1111, 0.0153, 0.0165, 0.0372, 0.1723, 0.0702], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0164, 0.0184, 0.0157, 0.0201, 0.0208, 0.0185, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 19:30:18,040 INFO [train.py:904] (2/8) Epoch 13, batch 5750, loss[loss=0.2516, simple_loss=0.3083, pruned_loss=0.09746, over 11379.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3093, pruned_loss=0.0752, over 3042106.76 frames. ], batch size: 247, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:30:19,227 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0678, 2.0113, 2.2004, 3.5235, 1.9912, 2.3804, 2.1573, 2.1494], device='cuda:2'), covar=tensor([0.1081, 0.3331, 0.2296, 0.0509, 0.3756, 0.2123, 0.3003, 0.3234], device='cuda:2'), in_proj_covar=tensor([0.0364, 0.0400, 0.0332, 0.0315, 0.0412, 0.0462, 0.0366, 0.0466], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:30:56,348 INFO [optim.py:368] (2/8) 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 19:31:21,840 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:31:26,467 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8936, 5.3744, 5.5565, 5.3552, 5.4684, 5.9461, 5.4580, 5.2171], device='cuda:2'), covar=tensor([0.0893, 0.1852, 0.2479, 0.1760, 0.2237, 0.0839, 0.1386, 0.2216], device='cuda:2'), in_proj_covar=tensor([0.0363, 0.0507, 0.0549, 0.0432, 0.0588, 0.0577, 0.0439, 0.0589], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 19:31:39,675 INFO [train.py:904] (2/8) Epoch 13, batch 5800, loss[loss=0.245, simple_loss=0.3085, pruned_loss=0.09076, over 12097.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3089, pruned_loss=0.07459, over 3016354.42 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:31:44,371 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:31:49,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0276, 3.4345, 3.4704, 2.3229, 3.1224, 3.4892, 3.2886, 1.9777], device='cuda:2'), covar=tensor([0.0469, 0.0044, 0.0043, 0.0344, 0.0086, 0.0086, 0.0073, 0.0402], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0070, 0.0071, 0.0128, 0.0084, 0.0092, 0.0082, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 19:32:57,618 INFO [train.py:904] (2/8) Epoch 13, batch 5850, loss[loss=0.2039, simple_loss=0.2942, pruned_loss=0.05679, over 16172.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3059, pruned_loss=0.07198, over 3045651.23 frames. ], batch size: 165, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:33:15,140 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 13, batch 5900, loss[loss=0.187, simple_loss=0.2827, pruned_loss=0.04565, over 16741.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3047, pruned_loss=0.07065, over 3064921.59 frames. ], batch size: 57, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:34:30,676 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 19:34:57,881 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:35:36,317 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3903, 4.3642, 4.2906, 3.6294, 4.3508, 1.6070, 4.0542, 3.9876], device='cuda:2'), covar=tensor([0.0096, 0.0086, 0.0148, 0.0318, 0.0082, 0.2437, 0.0124, 0.0208], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0124, 0.0170, 0.0160, 0.0141, 0.0182, 0.0157, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:35:42,724 INFO [train.py:904] (2/8) Epoch 13, batch 5950, loss[loss=0.234, simple_loss=0.3208, pruned_loss=0.07359, over 17039.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.306, pruned_loss=0.0697, over 3073293.44 frames. ], batch size: 50, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:36:16,782 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9100, 5.4112, 5.6306, 5.3539, 5.4348, 5.9394, 5.4133, 5.2303], device='cuda:2'), covar=tensor([0.0912, 0.1707, 0.1804, 0.1903, 0.2129, 0.0887, 0.1426, 0.2397], device='cuda:2'), in_proj_covar=tensor([0.0367, 0.0512, 0.0556, 0.0437, 0.0593, 0.0581, 0.0444, 0.0592], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 19:36:21,830 INFO [optim.py:368] (2/8) 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 19:36:26,886 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:36:27,101 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 19:36:32,789 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 19:36:41,242 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 19:37:04,075 INFO [train.py:904] (2/8) Epoch 13, batch 6000, loss[loss=0.2091, simple_loss=0.2899, pruned_loss=0.06412, over 16305.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3041, pruned_loss=0.06843, over 3084467.31 frames. ], batch size: 165, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:37:04,075 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 19:37:14,210 INFO [train.py:938] (2/8) Epoch 13, validation: loss=0.1599, simple_loss=0.2726, pruned_loss=0.02359, over 944034.00 frames. 2023-04-29 19:37:14,211 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 19:38:13,295 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:32,204 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127850.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:35,158 INFO [train.py:904] (2/8) Epoch 13, batch 6050, loss[loss=0.2394, simple_loss=0.3063, pruned_loss=0.08624, over 12001.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3018, pruned_loss=0.06719, over 3097277.21 frames. ], batch size: 247, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:38:44,341 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-29 19:39:07,481 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:14,938 INFO [optim.py:368] (2/8) 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 19:39:28,467 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:55,360 INFO [train.py:904] (2/8) Epoch 13, batch 6100, loss[loss=0.2116, simple_loss=0.2971, pruned_loss=0.06301, over 16890.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.3005, pruned_loss=0.06607, over 3099321.50 frames. ], batch size: 116, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:40:01,518 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:12,616 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:23,020 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 19:40:47,064 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:15,198 INFO [train.py:904] (2/8) Epoch 13, batch 6150, loss[loss=0.2275, simple_loss=0.2988, pruned_loss=0.07806, over 12018.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.299, pruned_loss=0.06578, over 3107316.87 frames. ], batch size: 247, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:41:17,426 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:56,815 INFO [optim.py:368] (2/8) 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 19:42:18,116 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7377, 3.7986, 4.2588, 2.0680, 4.3732, 4.4199, 3.1046, 3.2340], device='cuda:2'), covar=tensor([0.0799, 0.0212, 0.0122, 0.1176, 0.0047, 0.0118, 0.0356, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0102, 0.0088, 0.0138, 0.0069, 0.0109, 0.0121, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 19:42:19,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8682, 5.2949, 5.5450, 5.2637, 5.2997, 5.8996, 5.3920, 5.1031], device='cuda:2'), covar=tensor([0.0946, 0.1728, 0.1640, 0.1865, 0.2187, 0.0907, 0.1386, 0.2356], device='cuda:2'), in_proj_covar=tensor([0.0367, 0.0515, 0.0559, 0.0439, 0.0596, 0.0583, 0.0448, 0.0595], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 19:42:39,406 INFO [train.py:904] (2/8) Epoch 13, batch 6200, loss[loss=0.1912, simple_loss=0.282, pruned_loss=0.05024, over 16737.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2972, pruned_loss=0.06539, over 3107153.39 frames. ], batch size: 124, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:42:39,885 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1146, 1.5787, 1.8356, 2.0520, 2.2259, 2.3243, 1.6224, 2.2853], device='cuda:2'), covar=tensor([0.0191, 0.0385, 0.0216, 0.0255, 0.0214, 0.0142, 0.0381, 0.0098], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0163, 0.0174, 0.0130, 0.0176, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 19:43:05,219 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:43:57,667 INFO [train.py:904] (2/8) Epoch 13, batch 6250, loss[loss=0.2041, simple_loss=0.2968, pruned_loss=0.05573, over 17123.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2972, pruned_loss=0.06546, over 3092794.52 frames. ], batch size: 49, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:44:08,643 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6930, 3.7739, 2.1686, 4.1899, 2.7676, 4.1265, 2.1588, 2.8256], device='cuda:2'), covar=tensor([0.0212, 0.0333, 0.1517, 0.0204, 0.0718, 0.0551, 0.1601, 0.0754], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0167, 0.0190, 0.0137, 0.0168, 0.0207, 0.0197, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 19:44:37,187 INFO [optim.py:368] (2/8) 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 19:44:45,115 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 19:45:13,038 INFO [train.py:904] (2/8) Epoch 13, batch 6300, loss[loss=0.2234, simple_loss=0.3101, pruned_loss=0.06839, over 16764.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2974, pruned_loss=0.06483, over 3105542.88 frames. ], batch size: 83, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:45:51,344 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-04-29 19:46:02,062 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:46:08,188 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:46:14,635 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0465, 3.0752, 1.8468, 3.2884, 2.2532, 3.3166, 2.0468, 2.5041], device='cuda:2'), covar=tensor([0.0245, 0.0371, 0.1571, 0.0168, 0.0874, 0.0493, 0.1476, 0.0749], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0169, 0.0191, 0.0138, 0.0169, 0.0208, 0.0199, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 19:46:33,873 INFO [train.py:904] (2/8) Epoch 13, batch 6350, loss[loss=0.2274, simple_loss=0.3095, pruned_loss=0.07267, over 16669.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2984, pruned_loss=0.06603, over 3106860.07 frames. ], batch size: 57, lr: 5.22e-03, grad_scale: 4.0 2023-04-29 19:47:13,674 INFO [optim.py:368] (2/8) 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 19:47:24,718 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:47:49,923 INFO [train.py:904] (2/8) Epoch 13, batch 6400, loss[loss=0.2019, simple_loss=0.2895, pruned_loss=0.05712, over 17115.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2991, pruned_loss=0.06729, over 3100063.95 frames. ], batch size: 49, lr: 5.22e-03, grad_scale: 8.0 2023-04-29 19:47:56,307 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:14,694 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 19:48:29,323 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:36,341 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:51,021 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2628, 1.4111, 1.8668, 2.1528, 2.2662, 2.5191, 1.4372, 2.3972], device='cuda:2'), covar=tensor([0.0175, 0.0422, 0.0249, 0.0232, 0.0230, 0.0127, 0.0452, 0.0106], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0176, 0.0159, 0.0162, 0.0173, 0.0130, 0.0176, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 19:49:04,024 INFO [train.py:904] (2/8) Epoch 13, batch 6450, loss[loss=0.2162, simple_loss=0.3079, pruned_loss=0.06222, over 16353.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2991, pruned_loss=0.0664, over 3101626.61 frames. ], batch size: 146, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:49:48,882 INFO [optim.py:368] (2/8) 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 19:49:59,302 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8053, 4.0336, 3.1001, 2.2641, 2.8758, 2.4478, 4.3053, 3.7006], device='cuda:2'), covar=tensor([0.2552, 0.0632, 0.1557, 0.2316, 0.2414, 0.1823, 0.0427, 0.1017], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0259, 0.0288, 0.0286, 0.0284, 0.0228, 0.0273, 0.0305], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:50:21,886 INFO [train.py:904] (2/8) Epoch 13, batch 6500, loss[loss=0.1681, simple_loss=0.2554, pruned_loss=0.04043, over 16664.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2967, pruned_loss=0.06482, over 3134157.10 frames. ], batch size: 89, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:50:45,301 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:50:47,662 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128320.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:51:39,454 INFO [train.py:904] (2/8) Epoch 13, batch 6550, loss[loss=0.2406, simple_loss=0.3294, pruned_loss=0.07593, over 17042.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.3, pruned_loss=0.06569, over 3137241.09 frames. ], batch size: 53, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:51:43,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3102, 2.1034, 2.1530, 4.1044, 2.0019, 2.5800, 2.1623, 2.1881], device='cuda:2'), covar=tensor([0.1065, 0.3542, 0.2577, 0.0433, 0.4091, 0.2323, 0.3515, 0.3320], device='cuda:2'), in_proj_covar=tensor([0.0366, 0.0401, 0.0332, 0.0314, 0.0415, 0.0462, 0.0367, 0.0468], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:52:01,585 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:01,926 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.28 vs. limit=5.0 2023-04-29 19:52:12,860 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0121, 1.8562, 2.5257, 2.8890, 2.7689, 3.2429, 1.9176, 3.1970], device='cuda:2'), covar=tensor([0.0133, 0.0405, 0.0223, 0.0217, 0.0220, 0.0128, 0.0437, 0.0118], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0175, 0.0159, 0.0161, 0.0173, 0.0130, 0.0176, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 19:52:22,963 INFO [optim.py:368] (2/8) 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 19:52:25,024 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:56,160 INFO [train.py:904] (2/8) Epoch 13, batch 6600, loss[loss=0.2354, simple_loss=0.3141, pruned_loss=0.0784, over 16842.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3019, pruned_loss=0.06611, over 3137628.05 frames. ], batch size: 116, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:53:13,984 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:53:48,769 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:04,851 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 19:54:14,487 INFO [train.py:904] (2/8) Epoch 13, batch 6650, loss[loss=0.2147, simple_loss=0.2948, pruned_loss=0.06734, over 16857.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3021, pruned_loss=0.06708, over 3141633.78 frames. ], batch size: 109, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:54:24,229 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9020, 3.4571, 3.2990, 2.0084, 2.7764, 2.3918, 3.2395, 3.6646], device='cuda:2'), covar=tensor([0.0402, 0.0716, 0.0648, 0.1888, 0.0894, 0.0907, 0.1030, 0.0916], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0149, 0.0160, 0.0145, 0.0138, 0.0126, 0.0138, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 19:54:48,008 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:56,949 INFO [optim.py:368] (2/8) 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 19:55:02,265 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:08,750 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-29 19:55:30,393 INFO [train.py:904] (2/8) Epoch 13, batch 6700, loss[loss=0.2191, simple_loss=0.2947, pruned_loss=0.07174, over 16458.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3015, pruned_loss=0.06803, over 3125882.04 frames. ], batch size: 35, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:55:36,475 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:57,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5051, 4.4706, 4.3454, 3.6849, 4.4054, 1.5501, 4.1433, 4.0753], device='cuda:2'), covar=tensor([0.0073, 0.0065, 0.0142, 0.0297, 0.0077, 0.2520, 0.0116, 0.0183], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0124, 0.0170, 0.0160, 0.0142, 0.0184, 0.0159, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 19:56:09,998 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:13,692 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 19:56:46,042 INFO [train.py:904] (2/8) Epoch 13, batch 6750, loss[loss=0.2151, simple_loss=0.2915, pruned_loss=0.06936, over 17025.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3005, pruned_loss=0.06791, over 3126965.14 frames. ], batch size: 50, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:56:49,431 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:55,919 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 19:57:22,972 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:57:28,178 INFO [optim.py:368] (2/8) 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 19:57:36,866 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 19:58:01,812 INFO [train.py:904] (2/8) Epoch 13, batch 6800, loss[loss=0.2061, simple_loss=0.3041, pruned_loss=0.05404, over 16782.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3003, pruned_loss=0.06707, over 3142067.95 frames. ], batch size: 83, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:58:54,591 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7557, 1.2822, 1.6292, 1.6287, 1.7854, 1.9110, 1.5333, 1.7850], device='cuda:2'), covar=tensor([0.0187, 0.0291, 0.0150, 0.0217, 0.0187, 0.0131, 0.0318, 0.0088], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0161, 0.0172, 0.0129, 0.0175, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 19:59:09,056 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 19:59:19,100 INFO [train.py:904] (2/8) Epoch 13, batch 6850, loss[loss=0.2172, simple_loss=0.3105, pruned_loss=0.06191, over 16319.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3009, pruned_loss=0.06707, over 3136953.75 frames. ], batch size: 165, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:59:54,879 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:00:00,842 INFO [optim.py:368] (2/8) 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 20:00:34,560 INFO [train.py:904] (2/8) Epoch 13, batch 6900, loss[loss=0.2904, simple_loss=0.3445, pruned_loss=0.1181, over 11411.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3026, pruned_loss=0.06642, over 3137841.08 frames. ], batch size: 248, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:00:35,036 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2251, 5.2236, 5.0500, 4.6843, 4.6025, 5.0883, 5.0659, 4.7591], device='cuda:2'), covar=tensor([0.0855, 0.0923, 0.0359, 0.0364, 0.1112, 0.0727, 0.0416, 0.0915], device='cuda:2'), in_proj_covar=tensor([0.0246, 0.0336, 0.0295, 0.0275, 0.0313, 0.0319, 0.0200, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:00:38,592 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-29 20:01:50,151 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 20:01:52,386 INFO [train.py:904] (2/8) Epoch 13, batch 6950, loss[loss=0.2189, simple_loss=0.3045, pruned_loss=0.06667, over 16525.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3047, pruned_loss=0.0685, over 3119126.46 frames. ], batch size: 75, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:02:18,898 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:02:36,166 INFO [optim.py:368] (2/8) 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 20:02:38,366 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:03:08,751 INFO [train.py:904] (2/8) Epoch 13, batch 7000, loss[loss=0.2318, simple_loss=0.3251, pruned_loss=0.06929, over 16387.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3055, pruned_loss=0.06866, over 3103834.14 frames. ], batch size: 146, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:04:01,227 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7321, 2.4177, 2.4682, 4.7921, 2.3414, 2.8742, 2.4335, 2.6729], device='cuda:2'), covar=tensor([0.0953, 0.3350, 0.2452, 0.0317, 0.3738, 0.2242, 0.3267, 0.2758], device='cuda:2'), in_proj_covar=tensor([0.0365, 0.0400, 0.0333, 0.0313, 0.0415, 0.0460, 0.0367, 0.0467], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:04:07,283 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:04:20,573 INFO [train.py:904] (2/8) Epoch 13, batch 7050, loss[loss=0.2214, simple_loss=0.3158, pruned_loss=0.06345, over 17106.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.305, pruned_loss=0.06784, over 3099741.21 frames. ], batch size: 49, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:05:02,586 INFO [optim.py:368] (2/8) 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 20:05:03,755 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3969, 2.5440, 2.0181, 2.2882, 2.9571, 2.6178, 3.1689, 3.1260], device='cuda:2'), covar=tensor([0.0073, 0.0321, 0.0407, 0.0337, 0.0181, 0.0288, 0.0150, 0.0176], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0211, 0.0204, 0.0205, 0.0208, 0.0208, 0.0214, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:05:05,667 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7011, 4.0474, 3.5933, 3.8545, 3.5062, 3.6440, 3.6227, 4.0014], device='cuda:2'), covar=tensor([0.2336, 0.1514, 0.2787, 0.1429, 0.1789, 0.2742, 0.2236, 0.1793], device='cuda:2'), in_proj_covar=tensor([0.0566, 0.0697, 0.0571, 0.0498, 0.0441, 0.0449, 0.0585, 0.0537], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:05:19,782 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:05:33,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8085, 3.8541, 2.2760, 4.5654, 2.8659, 4.4210, 2.3604, 3.0542], device='cuda:2'), covar=tensor([0.0223, 0.0336, 0.1595, 0.0100, 0.0750, 0.0366, 0.1429, 0.0694], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0165, 0.0188, 0.0136, 0.0167, 0.0206, 0.0194, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:05:38,297 INFO [train.py:904] (2/8) Epoch 13, batch 7100, loss[loss=0.2097, simple_loss=0.2974, pruned_loss=0.06104, over 16872.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3037, pruned_loss=0.06803, over 3086522.79 frames. ], batch size: 102, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:06:56,807 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:06:57,449 INFO [train.py:904] (2/8) Epoch 13, batch 7150, loss[loss=0.2246, simple_loss=0.309, pruned_loss=0.07017, over 16782.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3014, pruned_loss=0.06747, over 3094021.97 frames. ], batch size: 83, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:07:33,976 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128976.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:07:39,435 INFO [optim.py:368] (2/8) 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 20:08:12,085 INFO [train.py:904] (2/8) Epoch 13, batch 7200, loss[loss=0.1798, simple_loss=0.269, pruned_loss=0.0453, over 16686.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2999, pruned_loss=0.0662, over 3081886.61 frames. ], batch size: 134, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:08:45,130 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:09:29,930 INFO [train.py:904] (2/8) Epoch 13, batch 7250, loss[loss=0.2049, simple_loss=0.2851, pruned_loss=0.06229, over 16858.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2977, pruned_loss=0.06537, over 3090716.59 frames. ], batch size: 116, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:09:56,884 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:10:12,181 INFO [optim.py:368] (2/8) 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 20:10:45,716 INFO [train.py:904] (2/8) Epoch 13, batch 7300, loss[loss=0.241, simple_loss=0.3193, pruned_loss=0.08135, over 15260.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.297, pruned_loss=0.06491, over 3090761.40 frames. ], batch size: 190, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:11:09,071 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:11:29,115 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5713, 2.4939, 1.8061, 2.6824, 2.0139, 2.6648, 2.0199, 2.2498], device='cuda:2'), covar=tensor([0.0230, 0.0273, 0.1184, 0.0166, 0.0624, 0.0364, 0.1121, 0.0547], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0165, 0.0190, 0.0136, 0.0169, 0.0207, 0.0197, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:11:40,200 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:12:02,352 INFO [train.py:904] (2/8) Epoch 13, batch 7350, loss[loss=0.1982, simple_loss=0.2873, pruned_loss=0.05452, over 15318.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.298, pruned_loss=0.06578, over 3063230.24 frames. ], batch size: 191, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:12:46,221 INFO [optim.py:368] (2/8) 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 20:13:18,328 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9413, 2.9482, 2.7253, 4.8106, 3.6157, 4.2048, 1.6699, 3.1321], device='cuda:2'), covar=tensor([0.1237, 0.0662, 0.1079, 0.0142, 0.0260, 0.0365, 0.1478, 0.0756], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0165, 0.0185, 0.0158, 0.0205, 0.0209, 0.0188, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:13:21,035 INFO [train.py:904] (2/8) Epoch 13, batch 7400, loss[loss=0.2579, simple_loss=0.3331, pruned_loss=0.09133, over 11189.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2992, pruned_loss=0.06639, over 3048522.46 frames. ], batch size: 248, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:14:32,916 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:14:41,688 INFO [train.py:904] (2/8) Epoch 13, batch 7450, loss[loss=0.236, simple_loss=0.3189, pruned_loss=0.07652, over 15274.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2996, pruned_loss=0.06694, over 3055394.67 frames. ], batch size: 190, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:15:21,310 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 20:15:30,917 INFO [optim.py:368] (2/8) 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 20:15:47,299 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0147, 3.0635, 1.8085, 3.2368, 2.2503, 3.3108, 2.0036, 2.4639], device='cuda:2'), covar=tensor([0.0266, 0.0358, 0.1550, 0.0223, 0.0817, 0.0611, 0.1403, 0.0714], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0165, 0.0190, 0.0136, 0.0168, 0.0207, 0.0197, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:16:05,626 INFO [train.py:904] (2/8) Epoch 13, batch 7500, loss[loss=0.201, simple_loss=0.2903, pruned_loss=0.05588, over 16404.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3005, pruned_loss=0.067, over 3052473.23 frames. ], batch size: 146, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:17:24,560 INFO [train.py:904] (2/8) Epoch 13, batch 7550, loss[loss=0.243, simple_loss=0.3055, pruned_loss=0.09022, over 11832.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2994, pruned_loss=0.06696, over 3055974.25 frames. ], batch size: 246, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:17:41,525 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 20:18:03,728 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 20:18:07,707 INFO [optim.py:368] (2/8) 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 20:18:41,434 INFO [train.py:904] (2/8) Epoch 13, batch 7600, loss[loss=0.2077, simple_loss=0.293, pruned_loss=0.06122, over 17172.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2984, pruned_loss=0.06664, over 3078595.68 frames. ], batch size: 46, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:18:45,249 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 20:18:50,543 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5979, 2.5832, 1.8172, 2.7290, 2.1005, 2.7819, 2.0167, 2.3035], device='cuda:2'), covar=tensor([0.0281, 0.0352, 0.1249, 0.0193, 0.0682, 0.0451, 0.1169, 0.0613], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0166, 0.0190, 0.0136, 0.0168, 0.0208, 0.0196, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:18:53,019 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0887, 2.4066, 2.5863, 1.8545, 2.6951, 2.7888, 2.3978, 2.3694], device='cuda:2'), covar=tensor([0.0640, 0.0205, 0.0195, 0.0924, 0.0091, 0.0204, 0.0399, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0101, 0.0088, 0.0137, 0.0069, 0.0109, 0.0120, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 20:18:59,717 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:19:26,252 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6012, 2.5894, 1.9033, 2.7470, 2.1589, 2.7736, 2.0511, 2.3235], device='cuda:2'), covar=tensor([0.0295, 0.0417, 0.1171, 0.0203, 0.0624, 0.0523, 0.1128, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0165, 0.0190, 0.0136, 0.0168, 0.0208, 0.0196, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:19:37,485 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:20:00,005 INFO [train.py:904] (2/8) Epoch 13, batch 7650, loss[loss=0.2188, simple_loss=0.3013, pruned_loss=0.06811, over 15388.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2991, pruned_loss=0.0674, over 3073785.15 frames. ], batch size: 190, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:20:35,780 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:20:44,160 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:21:18,576 INFO [train.py:904] (2/8) Epoch 13, batch 7700, loss[loss=0.2113, simple_loss=0.2971, pruned_loss=0.06275, over 16387.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2987, pruned_loss=0.06728, over 3066402.22 frames. ], batch size: 35, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:21:48,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8655, 2.8822, 2.6206, 4.8729, 3.6801, 4.3230, 1.6664, 3.1647], device='cuda:2'), covar=tensor([0.1331, 0.0777, 0.1251, 0.0163, 0.0406, 0.0388, 0.1589, 0.0853], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0186, 0.0159, 0.0205, 0.0210, 0.0189, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:22:16,184 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2943, 4.1531, 4.3612, 4.5130, 4.6713, 4.2260, 4.5885, 4.6648], device='cuda:2'), covar=tensor([0.1757, 0.1089, 0.1437, 0.0632, 0.0512, 0.0997, 0.0699, 0.0565], device='cuda:2'), in_proj_covar=tensor([0.0539, 0.0665, 0.0798, 0.0678, 0.0517, 0.0526, 0.0545, 0.0627], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:22:26,989 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:22:35,877 INFO [train.py:904] (2/8) Epoch 13, batch 7750, loss[loss=0.2359, simple_loss=0.3269, pruned_loss=0.0725, over 16965.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2989, pruned_loss=0.06724, over 3059574.54 frames. ], batch size: 109, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:23:20,357 INFO [optim.py:368] (2/8) 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 20:23:27,593 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6051, 1.6977, 2.2103, 2.5670, 2.4990, 2.9067, 1.8369, 2.8501], device='cuda:2'), covar=tensor([0.0174, 0.0411, 0.0245, 0.0223, 0.0236, 0.0140, 0.0373, 0.0112], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0174, 0.0157, 0.0162, 0.0172, 0.0128, 0.0174, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 20:23:40,232 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:23:52,177 INFO [train.py:904] (2/8) Epoch 13, batch 7800, loss[loss=0.2638, simple_loss=0.319, pruned_loss=0.1043, over 11610.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3001, pruned_loss=0.06819, over 3065287.87 frames. ], batch size: 248, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:24:09,260 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:25:09,850 INFO [train.py:904] (2/8) Epoch 13, batch 7850, loss[loss=0.1923, simple_loss=0.2762, pruned_loss=0.05419, over 16757.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.301, pruned_loss=0.06818, over 3056408.22 frames. ], batch size: 124, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:25:34,795 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 20:25:39,007 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1608, 2.9629, 3.1054, 1.7888, 3.2420, 3.3194, 2.6132, 2.6134], device='cuda:2'), covar=tensor([0.0804, 0.0261, 0.0185, 0.1131, 0.0088, 0.0172, 0.0450, 0.0466], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0102, 0.0089, 0.0139, 0.0069, 0.0110, 0.0122, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 20:25:40,261 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:25:52,269 INFO [optim.py:368] (2/8) 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 20:26:22,707 INFO [train.py:904] (2/8) Epoch 13, batch 7900, loss[loss=0.2213, simple_loss=0.3086, pruned_loss=0.06698, over 16753.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2997, pruned_loss=0.06716, over 3066908.81 frames. ], batch size: 124, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:26:24,449 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8670, 4.9068, 4.7382, 4.4413, 4.3710, 4.7905, 4.7039, 4.4539], device='cuda:2'), covar=tensor([0.0590, 0.0429, 0.0258, 0.0262, 0.0941, 0.0419, 0.0289, 0.0648], device='cuda:2'), in_proj_covar=tensor([0.0245, 0.0336, 0.0293, 0.0274, 0.0311, 0.0317, 0.0201, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:26:36,796 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2982, 4.3551, 4.1387, 3.8990, 3.8138, 4.2305, 3.9831, 3.9367], device='cuda:2'), covar=tensor([0.0526, 0.0477, 0.0279, 0.0288, 0.0825, 0.0444, 0.0600, 0.0655], device='cuda:2'), in_proj_covar=tensor([0.0245, 0.0337, 0.0293, 0.0274, 0.0311, 0.0317, 0.0201, 0.0343], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:27:14,835 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-04-29 20:27:36,767 INFO [train.py:904] (2/8) Epoch 13, batch 7950, loss[loss=0.2162, simple_loss=0.3007, pruned_loss=0.06583, over 17233.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3001, pruned_loss=0.06794, over 3052487.24 frames. ], batch size: 44, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:28:02,254 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:28:18,208 INFO [optim.py:368] (2/8) 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 20:28:33,246 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129791.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:28:49,439 INFO [train.py:904] (2/8) Epoch 13, batch 8000, loss[loss=0.204, simple_loss=0.289, pruned_loss=0.05945, over 16611.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3009, pruned_loss=0.06915, over 3032823.32 frames. ], batch size: 57, lr: 5.19e-03, grad_scale: 8.0 2023-04-29 20:30:02,332 INFO [train.py:904] (2/8) Epoch 13, batch 8050, loss[loss=0.217, simple_loss=0.3046, pruned_loss=0.06472, over 16740.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3013, pruned_loss=0.06927, over 3028928.68 frames. ], batch size: 124, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:30:02,915 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:30:45,830 INFO [optim.py:368] (2/8) 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 20:30:46,851 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 20:31:02,162 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-29 20:31:15,197 INFO [train.py:904] (2/8) Epoch 13, batch 8100, loss[loss=0.2018, simple_loss=0.287, pruned_loss=0.05835, over 16998.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3008, pruned_loss=0.0685, over 3047616.79 frames. ], batch size: 41, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:29,503 INFO [train.py:904] (2/8) Epoch 13, batch 8150, loss[loss=0.1919, simple_loss=0.2824, pruned_loss=0.05072, over 16705.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2976, pruned_loss=0.06682, over 3063941.70 frames. ], batch size: 89, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:53,000 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:14,447 INFO [optim.py:368] (2/8) 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 20:33:28,576 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4610, 4.2963, 4.4833, 4.6698, 4.8320, 4.3660, 4.7931, 4.7893], device='cuda:2'), covar=tensor([0.1706, 0.1149, 0.1560, 0.0685, 0.0518, 0.0986, 0.0537, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0544, 0.0672, 0.0802, 0.0684, 0.0524, 0.0530, 0.0551, 0.0638], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:33:46,556 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130000.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:48,621 INFO [train.py:904] (2/8) Epoch 13, batch 8200, loss[loss=0.1878, simple_loss=0.2817, pruned_loss=0.04694, over 16820.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2948, pruned_loss=0.066, over 3039988.07 frames. ], batch size: 102, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:33:49,234 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:52,966 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8534, 3.8245, 4.2604, 2.0619, 4.4401, 4.4610, 3.1187, 3.3001], device='cuda:2'), covar=tensor([0.0687, 0.0191, 0.0165, 0.1166, 0.0043, 0.0104, 0.0385, 0.0389], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0139, 0.0069, 0.0110, 0.0121, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 20:35:09,262 INFO [train.py:904] (2/8) Epoch 13, batch 8250, loss[loss=0.1819, simple_loss=0.2804, pruned_loss=0.04164, over 16857.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2934, pruned_loss=0.06331, over 3034069.14 frames. ], batch size: 96, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:35:24,054 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:35:27,853 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:35:37,696 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:35:57,079 INFO [optim.py:368] (2/8) 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 20:36:15,005 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-29 20:36:16,163 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 20:36:29,984 INFO [train.py:904] (2/8) Epoch 13, batch 8300, loss[loss=0.1895, simple_loss=0.2838, pruned_loss=0.04758, over 16916.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.291, pruned_loss=0.06033, over 3048001.95 frames. ], batch size: 116, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:36:55,890 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:37:18,197 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7447, 3.6958, 3.8915, 3.6801, 3.8164, 4.2406, 3.9544, 3.6196], device='cuda:2'), covar=tensor([0.2282, 0.2561, 0.2197, 0.2631, 0.3147, 0.1700, 0.1489, 0.2830], device='cuda:2'), in_proj_covar=tensor([0.0358, 0.0505, 0.0552, 0.0432, 0.0582, 0.0576, 0.0442, 0.0579], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 20:37:43,814 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:37:51,334 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5753, 3.5938, 3.3819, 3.1410, 3.1878, 3.5025, 3.3270, 3.3408], device='cuda:2'), covar=tensor([0.0513, 0.0566, 0.0243, 0.0219, 0.0483, 0.0391, 0.1164, 0.0466], device='cuda:2'), in_proj_covar=tensor([0.0243, 0.0335, 0.0293, 0.0271, 0.0306, 0.0314, 0.0200, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:37:51,985 INFO [train.py:904] (2/8) Epoch 13, batch 8350, loss[loss=0.2417, simple_loss=0.3084, pruned_loss=0.08753, over 12057.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2906, pruned_loss=0.05845, over 3043038.79 frames. ], batch size: 247, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:38:31,835 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6213, 3.7910, 2.0778, 4.1518, 2.8369, 4.1102, 2.1693, 2.9713], device='cuda:2'), covar=tensor([0.0208, 0.0296, 0.1441, 0.0140, 0.0671, 0.0397, 0.1508, 0.0621], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0161, 0.0184, 0.0132, 0.0163, 0.0201, 0.0192, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 20:38:39,963 INFO [optim.py:368] (2/8) 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 20:39:12,412 INFO [train.py:904] (2/8) Epoch 13, batch 8400, loss[loss=0.1942, simple_loss=0.2818, pruned_loss=0.05335, over 16150.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2878, pruned_loss=0.05608, over 3041471.43 frames. ], batch size: 165, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:29,224 INFO [train.py:904] (2/8) Epoch 13, batch 8450, loss[loss=0.1698, simple_loss=0.2649, pruned_loss=0.03739, over 17053.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2864, pruned_loss=0.05477, over 3045895.83 frames. ], batch size: 55, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:55,984 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:41:17,432 INFO [optim.py:368] (2/8) 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 20:41:49,390 INFO [train.py:904] (2/8) Epoch 13, batch 8500, loss[loss=0.1596, simple_loss=0.2529, pruned_loss=0.03309, over 15255.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2821, pruned_loss=0.05163, over 3064179.45 frames. ], batch size: 191, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:42:12,542 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:43:10,837 INFO [train.py:904] (2/8) Epoch 13, batch 8550, loss[loss=0.1939, simple_loss=0.2821, pruned_loss=0.05288, over 16919.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2795, pruned_loss=0.05065, over 3043152.57 frames. ], batch size: 109, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:43:19,679 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:43:23,658 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:44:07,712 INFO [optim.py:368] (2/8) 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 20:44:12,597 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:44:50,607 INFO [train.py:904] (2/8) Epoch 13, batch 8600, loss[loss=0.2011, simple_loss=0.2779, pruned_loss=0.0621, over 12606.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2798, pruned_loss=0.05006, over 3025523.84 frames. ], batch size: 247, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:45:42,651 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0983, 2.5420, 2.5983, 1.9262, 2.7816, 2.9031, 2.5145, 2.4746], device='cuda:2'), covar=tensor([0.0673, 0.0200, 0.0183, 0.0913, 0.0081, 0.0169, 0.0377, 0.0408], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0097, 0.0085, 0.0133, 0.0066, 0.0104, 0.0116, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 20:45:51,019 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:04,162 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:04,441 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.34 vs. limit=5.0 2023-04-29 20:46:16,972 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:20,506 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130447.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:29,995 INFO [train.py:904] (2/8) Epoch 13, batch 8650, loss[loss=0.1891, simple_loss=0.2838, pruned_loss=0.04713, over 16678.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2778, pruned_loss=0.04858, over 3015632.57 frames. ], batch size: 134, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:47:40,991 INFO [optim.py:368] (2/8) 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 20:47:45,117 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9165, 2.5035, 2.2939, 3.1672, 1.9642, 3.3753, 1.7209, 2.6938], device='cuda:2'), covar=tensor([0.1114, 0.0493, 0.0982, 0.0145, 0.0074, 0.0373, 0.1290, 0.0618], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0162, 0.0182, 0.0154, 0.0198, 0.0206, 0.0185, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 20:47:46,543 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8327, 3.6769, 3.8820, 3.9796, 4.0886, 3.6646, 4.0535, 4.0824], device='cuda:2'), covar=tensor([0.1310, 0.1105, 0.1212, 0.0642, 0.0548, 0.1609, 0.0591, 0.0588], device='cuda:2'), in_proj_covar=tensor([0.0524, 0.0648, 0.0772, 0.0664, 0.0505, 0.0512, 0.0528, 0.0617], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:48:01,195 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:04,696 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:12,952 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130499.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:17,482 INFO [train.py:904] (2/8) Epoch 13, batch 8700, loss[loss=0.176, simple_loss=0.2586, pruned_loss=0.04666, over 12532.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2747, pruned_loss=0.04712, over 3022877.74 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:49:46,594 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:49:54,308 INFO [train.py:904] (2/8) Epoch 13, batch 8750, loss[loss=0.1721, simple_loss=0.2614, pruned_loss=0.04141, over 12385.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2738, pruned_loss=0.04637, over 3040888.35 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:49:56,879 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 20:50:30,084 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:51:07,894 INFO [optim.py:368] (2/8) 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 20:51:48,202 INFO [train.py:904] (2/8) Epoch 13, batch 8800, loss[loss=0.1785, simple_loss=0.2743, pruned_loss=0.0414, over 16224.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2721, pruned_loss=0.04498, over 3051638.54 frames. ], batch size: 166, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:52:02,456 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:52:39,096 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:31,851 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1290, 3.2420, 1.6908, 3.5018, 2.1508, 3.4563, 1.8695, 2.5382], device='cuda:2'), covar=tensor([0.0286, 0.0339, 0.1752, 0.0155, 0.1032, 0.0425, 0.1700, 0.0848], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0160, 0.0185, 0.0130, 0.0164, 0.0198, 0.0192, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 20:53:32,460 INFO [train.py:904] (2/8) Epoch 13, batch 8850, loss[loss=0.176, simple_loss=0.2831, pruned_loss=0.03449, over 16176.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2748, pruned_loss=0.04473, over 3041658.41 frames. ], batch size: 165, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:53:39,824 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6106, 2.6462, 1.6182, 2.8103, 1.9860, 2.7689, 1.8127, 2.3042], device='cuda:2'), covar=tensor([0.0299, 0.0355, 0.1616, 0.0194, 0.0827, 0.0401, 0.1639, 0.0751], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0160, 0.0184, 0.0130, 0.0164, 0.0198, 0.0192, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 20:53:41,312 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:45,881 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:54:37,528 INFO [optim.py:368] (2/8) 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 20:55:17,114 INFO [train.py:904] (2/8) Epoch 13, batch 8900, loss[loss=0.1931, simple_loss=0.2826, pruned_loss=0.05183, over 15337.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2749, pruned_loss=0.04392, over 3054678.29 frames. ], batch size: 191, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:55:22,705 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:55:26,794 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:56:20,254 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 20:56:54,223 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:57:01,901 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7897, 3.7828, 4.1765, 2.2701, 4.3778, 4.4649, 3.1994, 3.2376], device='cuda:2'), covar=tensor([0.0713, 0.0190, 0.0145, 0.1062, 0.0037, 0.0062, 0.0323, 0.0393], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0097, 0.0084, 0.0133, 0.0065, 0.0103, 0.0116, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 20:57:21,721 INFO [train.py:904] (2/8) Epoch 13, batch 8950, loss[loss=0.1622, simple_loss=0.2538, pruned_loss=0.03526, over 15210.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.275, pruned_loss=0.0443, over 3060332.25 frames. ], batch size: 190, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:57:33,975 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 20:57:50,136 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-29 20:58:29,463 INFO [optim.py:368] (2/8) 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 20:58:32,637 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5651, 2.0952, 1.5691, 1.7494, 2.4274, 2.0327, 2.4331, 2.5606], device='cuda:2'), covar=tensor([0.0142, 0.0428, 0.0577, 0.0488, 0.0280, 0.0419, 0.0187, 0.0259], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0205, 0.0199, 0.0199, 0.0203, 0.0203, 0.0202, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 20:58:41,730 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:58:54,558 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:59:04,205 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-29 20:59:11,360 INFO [train.py:904] (2/8) Epoch 13, batch 9000, loss[loss=0.1755, simple_loss=0.2629, pruned_loss=0.04402, over 12210.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2714, pruned_loss=0.04291, over 3057350.11 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:59:11,361 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 20:59:22,049 INFO [train.py:938] (2/8) Epoch 13, validation: loss=0.1517, simple_loss=0.2561, pruned_loss=0.02371, over 944034.00 frames. 2023-04-29 20:59:22,050 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 21:00:24,623 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 21:01:06,010 INFO [train.py:904] (2/8) Epoch 13, batch 9050, loss[loss=0.1858, simple_loss=0.2774, pruned_loss=0.04704, over 16788.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2728, pruned_loss=0.04352, over 3070889.74 frames. ], batch size: 83, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:07,082 INFO [optim.py:368] (2/8) 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 21:02:14,435 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8587, 3.9608, 2.4368, 4.4729, 2.9736, 4.3778, 2.6668, 3.2370], device='cuda:2'), covar=tensor([0.0193, 0.0259, 0.1329, 0.0152, 0.0690, 0.0373, 0.1217, 0.0604], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0160, 0.0185, 0.0129, 0.0163, 0.0198, 0.0191, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 21:02:52,493 INFO [train.py:904] (2/8) Epoch 13, batch 9100, loss[loss=0.1695, simple_loss=0.2536, pruned_loss=0.04264, over 12259.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2723, pruned_loss=0.04383, over 3075106.21 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:58,157 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 21:03:33,377 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130921.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:04:49,577 INFO [train.py:904] (2/8) Epoch 13, batch 9150, loss[loss=0.1715, simple_loss=0.2665, pruned_loss=0.03823, over 15401.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2725, pruned_loss=0.04328, over 3072683.34 frames. ], batch size: 191, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:05:52,920 INFO [optim.py:368] (2/8) 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 21:06:31,914 INFO [train.py:904] (2/8) Epoch 13, batch 9200, loss[loss=0.1652, simple_loss=0.2462, pruned_loss=0.04211, over 12341.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2679, pruned_loss=0.04203, over 3073487.74 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:06:33,932 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3797, 3.3427, 3.4436, 3.5158, 3.5700, 3.3019, 3.5604, 3.6161], device='cuda:2'), covar=tensor([0.1087, 0.0839, 0.0939, 0.0620, 0.0537, 0.1991, 0.0745, 0.0668], device='cuda:2'), in_proj_covar=tensor([0.0526, 0.0646, 0.0772, 0.0663, 0.0503, 0.0509, 0.0529, 0.0611], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:07:31,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2484, 3.4633, 2.0363, 3.7506, 2.4851, 3.6520, 2.0750, 2.6157], device='cuda:2'), covar=tensor([0.0287, 0.0311, 0.1524, 0.0176, 0.0853, 0.0540, 0.1498, 0.0786], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0159, 0.0183, 0.0128, 0.0162, 0.0197, 0.0190, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-29 21:07:43,557 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:08:07,049 INFO [train.py:904] (2/8) Epoch 13, batch 9250, loss[loss=0.1627, simple_loss=0.2562, pruned_loss=0.03464, over 12301.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2678, pruned_loss=0.04213, over 3054581.06 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:08:25,143 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-29 21:09:00,099 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 21:09:12,850 INFO [optim.py:368] (2/8) 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 21:09:24,455 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:24,512 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:39,322 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:55,037 INFO [train.py:904] (2/8) Epoch 13, batch 9300, loss[loss=0.1801, simple_loss=0.2531, pruned_loss=0.05355, over 12649.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2661, pruned_loss=0.0416, over 3047515.91 frames. ], batch size: 249, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:10:00,910 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 21:10:29,869 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6690, 2.4680, 2.2860, 3.5836, 2.1934, 3.6717, 1.3064, 2.8330], device='cuda:2'), covar=tensor([0.1430, 0.0735, 0.1205, 0.0197, 0.0112, 0.0399, 0.1756, 0.0751], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0162, 0.0184, 0.0154, 0.0193, 0.0206, 0.0187, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 21:10:40,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5153, 3.7327, 2.8347, 2.0525, 2.3782, 2.2795, 4.0354, 3.3873], device='cuda:2'), covar=tensor([0.3080, 0.0676, 0.1716, 0.2731, 0.2514, 0.2043, 0.0396, 0.1027], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0252, 0.0284, 0.0280, 0.0265, 0.0225, 0.0266, 0.0294], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:11:11,076 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:23,562 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:40,550 INFO [train.py:904] (2/8) Epoch 13, batch 9350, loss[loss=0.1674, simple_loss=0.2621, pruned_loss=0.03638, over 16800.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2662, pruned_loss=0.0418, over 3049565.07 frames. ], batch size: 83, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:12:25,574 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131174.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:12:41,633 INFO [optim.py:368] (2/8) 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 21:13:20,691 INFO [train.py:904] (2/8) Epoch 13, batch 9400, loss[loss=0.1679, simple_loss=0.2495, pruned_loss=0.04309, over 12470.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2653, pruned_loss=0.04115, over 3037670.26 frames. ], batch size: 246, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:13:25,622 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:13:59,481 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:14:27,933 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:14:59,709 INFO [train.py:904] (2/8) Epoch 13, batch 9450, loss[loss=0.1864, simple_loss=0.2692, pruned_loss=0.0518, over 12482.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2675, pruned_loss=0.04161, over 3040306.74 frames. ], batch size: 247, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:15:00,865 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131252.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:15:34,984 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:15:42,232 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-29 21:16:03,420 INFO [optim.py:368] (2/8) 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 21:16:19,273 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 21:16:40,125 INFO [train.py:904] (2/8) Epoch 13, batch 9500, loss[loss=0.1619, simple_loss=0.2551, pruned_loss=0.03433, over 17039.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2659, pruned_loss=0.04071, over 3058889.63 frames. ], batch size: 55, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:17:13,866 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-29 21:18:25,254 INFO [train.py:904] (2/8) Epoch 13, batch 9550, loss[loss=0.2073, simple_loss=0.3001, pruned_loss=0.05724, over 16606.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2657, pruned_loss=0.04056, over 3069441.43 frames. ], batch size: 134, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:19:29,577 INFO [optim.py:368] (2/8) 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 21:20:03,674 INFO [train.py:904] (2/8) Epoch 13, batch 9600, loss[loss=0.1865, simple_loss=0.2668, pruned_loss=0.05311, over 12253.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2681, pruned_loss=0.04185, over 3060917.60 frames. ], batch size: 250, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:20:17,931 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:20:29,767 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:21:12,054 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8684, 4.1886, 4.0095, 4.0402, 3.6541, 3.7204, 3.8017, 4.1681], device='cuda:2'), covar=tensor([0.1029, 0.0928, 0.1014, 0.0693, 0.0754, 0.1737, 0.0944, 0.1074], device='cuda:2'), in_proj_covar=tensor([0.0547, 0.0686, 0.0549, 0.0486, 0.0428, 0.0440, 0.0569, 0.0521], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:21:49,828 INFO [train.py:904] (2/8) Epoch 13, batch 9650, loss[loss=0.1765, simple_loss=0.2688, pruned_loss=0.04216, over 16208.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2699, pruned_loss=0.04208, over 3072733.38 frames. ], batch size: 165, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:22:33,802 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:22:47,901 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 21:23:18,185 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131492.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:23:38,319 INFO [train.py:904] (2/8) Epoch 13, batch 9700, loss[loss=0.1815, simple_loss=0.2721, pruned_loss=0.04544, over 16992.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2691, pruned_loss=0.04201, over 3066053.18 frames. ], batch size: 109, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:24:36,287 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131530.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:25:20,235 INFO [train.py:904] (2/8) Epoch 13, batch 9750, loss[loss=0.1783, simple_loss=0.2753, pruned_loss=0.04064, over 16704.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2678, pruned_loss=0.04199, over 3072179.11 frames. ], batch size: 134, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:25:22,334 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:26:22,843 INFO [optim.py:368] (2/8) 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 21:26:57,813 INFO [train.py:904] (2/8) Epoch 13, batch 9800, loss[loss=0.1776, simple_loss=0.2786, pruned_loss=0.03827, over 15189.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2678, pruned_loss=0.04114, over 3079634.62 frames. ], batch size: 190, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:28:39,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5090, 4.0896, 4.2162, 3.1601, 3.6559, 4.2157, 3.9678, 2.4481], device='cuda:2'), covar=tensor([0.0439, 0.0029, 0.0026, 0.0246, 0.0082, 0.0056, 0.0049, 0.0394], device='cuda:2'), in_proj_covar=tensor([0.0127, 0.0067, 0.0069, 0.0125, 0.0081, 0.0088, 0.0078, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 21:28:39,800 INFO [train.py:904] (2/8) Epoch 13, batch 9850, loss[loss=0.1619, simple_loss=0.2577, pruned_loss=0.03307, over 16532.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2698, pruned_loss=0.04099, over 3089563.78 frames. ], batch size: 68, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:28:46,235 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5696, 3.6024, 3.4082, 3.1016, 3.2342, 3.5222, 3.3001, 3.3116], device='cuda:2'), covar=tensor([0.0510, 0.0460, 0.0254, 0.0218, 0.0438, 0.0404, 0.1240, 0.0413], device='cuda:2'), in_proj_covar=tensor([0.0236, 0.0320, 0.0286, 0.0263, 0.0293, 0.0306, 0.0193, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-29 21:29:46,601 INFO [optim.py:368] (2/8) 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 21:30:18,583 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 21:30:19,960 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7853, 5.0743, 4.8934, 4.9096, 4.6287, 4.5651, 4.5557, 5.1316], device='cuda:2'), covar=tensor([0.1028, 0.0825, 0.0840, 0.0628, 0.0694, 0.1032, 0.0929, 0.0877], device='cuda:2'), in_proj_covar=tensor([0.0548, 0.0688, 0.0552, 0.0486, 0.0431, 0.0441, 0.0569, 0.0524], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:30:32,565 INFO [train.py:904] (2/8) Epoch 13, batch 9900, loss[loss=0.1844, simple_loss=0.2887, pruned_loss=0.04007, over 16288.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2697, pruned_loss=0.04078, over 3064081.00 frames. ], batch size: 146, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:32:30,669 INFO [train.py:904] (2/8) Epoch 13, batch 9950, loss[loss=0.1606, simple_loss=0.2576, pruned_loss=0.03181, over 16549.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.272, pruned_loss=0.04131, over 3058266.09 frames. ], batch size: 68, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:33:02,012 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:02,140 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2833, 4.2885, 4.1568, 3.6560, 4.2065, 1.6126, 3.9657, 3.9679], device='cuda:2'), covar=tensor([0.0095, 0.0096, 0.0155, 0.0267, 0.0096, 0.2424, 0.0134, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0119, 0.0160, 0.0147, 0.0135, 0.0181, 0.0152, 0.0146], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:33:20,130 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:47,201 INFO [optim.py:368] (2/8) 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 21:34:31,155 INFO [train.py:904] (2/8) Epoch 13, batch 10000, loss[loss=0.1855, simple_loss=0.2784, pruned_loss=0.04626, over 12901.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2708, pruned_loss=0.04106, over 3082019.00 frames. ], batch size: 248, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:35:27,559 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:00,350 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 21:36:06,298 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:11,635 INFO [train.py:904] (2/8) Epoch 13, batch 10050, loss[loss=0.1934, simple_loss=0.287, pruned_loss=0.04991, over 16681.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2702, pruned_loss=0.04055, over 3079570.57 frames. ], batch size: 134, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:36:23,306 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:04,199 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:11,036 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 21:37:14,088 INFO [optim.py:368] (2/8) 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 21:37:45,821 INFO [train.py:904] (2/8) Epoch 13, batch 10100, loss[loss=0.1747, simple_loss=0.2587, pruned_loss=0.04537, over 12610.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2706, pruned_loss=0.04103, over 3073364.80 frames. ], batch size: 248, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:38:16,883 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:39:29,405 INFO [train.py:904] (2/8) Epoch 14, batch 0, loss[loss=0.1792, simple_loss=0.2747, pruned_loss=0.04188, over 17117.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2747, pruned_loss=0.04188, over 17117.00 frames. ], batch size: 48, lr: 4.96e-03, grad_scale: 8.0 2023-04-29 21:39:29,405 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 21:39:36,895 INFO [train.py:938] (2/8) Epoch 14, validation: loss=0.1515, simple_loss=0.2551, pruned_loss=0.024, over 944034.00 frames. 2023-04-29 21:39:36,895 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 21:40:22,370 INFO [optim.py:368] (2/8) 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 21:40:46,681 INFO [train.py:904] (2/8) Epoch 14, batch 50, loss[loss=0.1613, simple_loss=0.2488, pruned_loss=0.03695, over 16857.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2847, pruned_loss=0.06025, over 744623.64 frames. ], batch size: 42, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:27,136 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132031.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:41:55,078 INFO [train.py:904] (2/8) Epoch 14, batch 100, loss[loss=0.2081, simple_loss=0.278, pruned_loss=0.06906, over 16911.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2784, pruned_loss=0.05524, over 1310288.52 frames. ], batch size: 109, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:58,495 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:04,234 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 21:42:13,316 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:23,839 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:28,032 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8074, 4.7969, 5.3452, 5.3395, 5.3261, 4.8822, 4.8906, 4.6314], device='cuda:2'), covar=tensor([0.0332, 0.0422, 0.0390, 0.0387, 0.0544, 0.0428, 0.1000, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0337, 0.0350, 0.0351, 0.0338, 0.0395, 0.0378, 0.0462, 0.0301], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 21:42:44,829 INFO [optim.py:368] (2/8) 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 21:42:51,077 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:04,251 INFO [train.py:904] (2/8) Epoch 14, batch 150, loss[loss=0.1538, simple_loss=0.2465, pruned_loss=0.03051, over 17181.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2749, pruned_loss=0.05286, over 1757068.79 frames. ], batch size: 46, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:43:19,278 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9971, 2.0030, 2.1778, 3.5387, 2.0146, 2.3116, 2.1336, 2.1835], device='cuda:2'), covar=tensor([0.1166, 0.3344, 0.2522, 0.0630, 0.3807, 0.2298, 0.3233, 0.2950], device='cuda:2'), in_proj_covar=tensor([0.0368, 0.0402, 0.0338, 0.0316, 0.0415, 0.0456, 0.0366, 0.0466], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:43:20,790 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:23,963 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 21:43:30,565 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:33,503 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-29 21:43:43,995 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8795, 1.7770, 2.4353, 2.8261, 2.6450, 3.2299, 1.9654, 3.2514], device='cuda:2'), covar=tensor([0.0202, 0.0441, 0.0277, 0.0236, 0.0245, 0.0148, 0.0431, 0.0129], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0176, 0.0159, 0.0162, 0.0173, 0.0129, 0.0178, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 21:44:07,519 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0630, 3.8502, 4.1302, 4.2576, 4.3331, 3.9365, 4.1640, 4.3522], device='cuda:2'), covar=tensor([0.1358, 0.1042, 0.1135, 0.0627, 0.0537, 0.1290, 0.1604, 0.0571], device='cuda:2'), in_proj_covar=tensor([0.0543, 0.0666, 0.0799, 0.0682, 0.0515, 0.0523, 0.0545, 0.0628], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:44:08,564 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:44:15,117 INFO [train.py:904] (2/8) Epoch 14, batch 200, loss[loss=0.1771, simple_loss=0.264, pruned_loss=0.04505, over 17150.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2749, pruned_loss=0.05314, over 2101249.53 frames. ], batch size: 46, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:44:15,669 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0819, 1.8111, 2.5736, 3.0574, 2.7244, 3.3791, 1.9119, 3.4611], device='cuda:2'), covar=tensor([0.0151, 0.0460, 0.0237, 0.0195, 0.0223, 0.0144, 0.0486, 0.0094], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0177, 0.0159, 0.0162, 0.0173, 0.0129, 0.0178, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 21:44:27,941 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 21:45:02,143 INFO [optim.py:368] (2/8) 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 21:45:15,515 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:45:21,481 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 21:45:23,102 INFO [train.py:904] (2/8) Epoch 14, batch 250, loss[loss=0.216, simple_loss=0.2867, pruned_loss=0.07263, over 12076.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2729, pruned_loss=0.05269, over 2372312.42 frames. ], batch size: 246, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:45:38,785 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:46:33,333 INFO [train.py:904] (2/8) Epoch 14, batch 300, loss[loss=0.153, simple_loss=0.2325, pruned_loss=0.0367, over 15912.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2693, pruned_loss=0.0519, over 2586424.90 frames. ], batch size: 35, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:47:16,167 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7929, 2.7689, 2.3211, 2.6749, 3.1337, 2.9743, 3.6013, 3.4038], device='cuda:2'), covar=tensor([0.0092, 0.0351, 0.0453, 0.0355, 0.0228, 0.0311, 0.0189, 0.0217], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0216, 0.0209, 0.0208, 0.0214, 0.0215, 0.0216, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:47:22,677 INFO [optim.py:368] (2/8) 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 21:47:42,658 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2819, 4.2024, 4.6428, 4.6421, 4.7010, 4.3446, 4.3829, 4.2526], device='cuda:2'), covar=tensor([0.0357, 0.0620, 0.0395, 0.0399, 0.0412, 0.0416, 0.0835, 0.0538], device='cuda:2'), in_proj_covar=tensor([0.0350, 0.0366, 0.0366, 0.0352, 0.0409, 0.0393, 0.0481, 0.0313], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 21:47:43,454 INFO [train.py:904] (2/8) Epoch 14, batch 350, loss[loss=0.2253, simple_loss=0.284, pruned_loss=0.08336, over 16858.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2677, pruned_loss=0.05088, over 2745112.29 frames. ], batch size: 109, lr: 4.95e-03, grad_scale: 1.0 2023-04-29 21:48:04,630 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 21:48:47,469 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0766, 5.5251, 5.6500, 5.4679, 5.5524, 6.0743, 5.5205, 5.2981], device='cuda:2'), covar=tensor([0.0891, 0.1930, 0.2445, 0.1893, 0.2651, 0.0993, 0.1514, 0.2520], device='cuda:2'), in_proj_covar=tensor([0.0363, 0.0515, 0.0565, 0.0437, 0.0594, 0.0588, 0.0448, 0.0582], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 21:48:51,146 INFO [train.py:904] (2/8) Epoch 14, batch 400, loss[loss=0.161, simple_loss=0.2525, pruned_loss=0.03477, over 17216.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2663, pruned_loss=0.05021, over 2870221.15 frames. ], batch size: 46, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:49:38,425 INFO [optim.py:368] (2/8) 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 21:49:39,466 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:49:43,179 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7106, 3.8241, 2.8898, 2.1989, 2.5376, 2.3123, 3.9308, 3.3718], device='cuda:2'), covar=tensor([0.2506, 0.0534, 0.1522, 0.2722, 0.2504, 0.1863, 0.0465, 0.1335], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0256, 0.0288, 0.0283, 0.0272, 0.0228, 0.0271, 0.0302], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:49:54,341 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9873, 2.0351, 2.4915, 3.0059, 2.6960, 3.3518, 2.2071, 3.3284], device='cuda:2'), covar=tensor([0.0194, 0.0369, 0.0234, 0.0202, 0.0237, 0.0135, 0.0349, 0.0118], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0177, 0.0160, 0.0163, 0.0174, 0.0130, 0.0178, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 21:50:00,132 INFO [train.py:904] (2/8) Epoch 14, batch 450, loss[loss=0.1793, simple_loss=0.2714, pruned_loss=0.04359, over 17139.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2644, pruned_loss=0.04877, over 2976002.15 frames. ], batch size: 49, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:50:11,959 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:50:35,456 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8284, 1.3491, 1.6831, 1.7107, 1.7615, 1.9254, 1.5761, 1.8003], device='cuda:2'), covar=tensor([0.0197, 0.0315, 0.0158, 0.0200, 0.0204, 0.0145, 0.0330, 0.0097], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0178, 0.0160, 0.0163, 0.0173, 0.0130, 0.0178, 0.0120], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-29 21:51:11,603 INFO [train.py:904] (2/8) Epoch 14, batch 500, loss[loss=0.1565, simple_loss=0.2332, pruned_loss=0.03985, over 15503.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2629, pruned_loss=0.04825, over 3056067.95 frames. ], batch size: 190, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:51:58,582 INFO [optim.py:368] (2/8) 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 21:52:19,293 INFO [train.py:904] (2/8) Epoch 14, batch 550, loss[loss=0.1907, simple_loss=0.2879, pruned_loss=0.04676, over 16725.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2624, pruned_loss=0.04748, over 3118766.29 frames. ], batch size: 57, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:52:24,160 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 21:52:34,267 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:52:48,586 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6240, 4.4011, 4.5851, 4.7902, 4.8996, 4.4526, 4.7845, 4.8649], device='cuda:2'), covar=tensor([0.1496, 0.1100, 0.1408, 0.0730, 0.0631, 0.1063, 0.1846, 0.0805], device='cuda:2'), in_proj_covar=tensor([0.0576, 0.0704, 0.0848, 0.0723, 0.0548, 0.0552, 0.0572, 0.0667], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:52:51,112 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9306, 3.0154, 3.0829, 2.0306, 2.7362, 2.3063, 3.4147, 3.2854], device='cuda:2'), covar=tensor([0.0222, 0.0825, 0.0601, 0.1739, 0.0793, 0.0924, 0.0515, 0.0836], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0144, 0.0157, 0.0144, 0.0136, 0.0124, 0.0134, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 21:53:14,752 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0854, 5.6475, 5.8068, 5.5612, 5.6062, 6.1772, 5.7270, 5.5094], device='cuda:2'), covar=tensor([0.0901, 0.1834, 0.2122, 0.2141, 0.2724, 0.0986, 0.1448, 0.2436], device='cuda:2'), in_proj_covar=tensor([0.0370, 0.0522, 0.0574, 0.0444, 0.0603, 0.0597, 0.0453, 0.0592], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 21:53:28,147 INFO [train.py:904] (2/8) Epoch 14, batch 600, loss[loss=0.1762, simple_loss=0.2683, pruned_loss=0.04205, over 17024.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2612, pruned_loss=0.04778, over 3153534.52 frames. ], batch size: 50, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:53:41,853 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:54:14,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1663, 4.2800, 4.5860, 4.6025, 4.6387, 4.3582, 4.2222, 4.1857], device='cuda:2'), covar=tensor([0.0545, 0.0807, 0.0561, 0.0543, 0.0623, 0.0574, 0.1158, 0.0685], device='cuda:2'), in_proj_covar=tensor([0.0358, 0.0374, 0.0374, 0.0358, 0.0417, 0.0400, 0.0492, 0.0319], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 21:54:17,820 INFO [optim.py:368] (2/8) 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 21:54:39,141 INFO [train.py:904] (2/8) Epoch 14, batch 650, loss[loss=0.1504, simple_loss=0.2338, pruned_loss=0.03354, over 16817.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2598, pruned_loss=0.04802, over 3179049.53 frames. ], batch size: 39, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:54:39,597 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0265, 4.1628, 2.5997, 4.7000, 3.1886, 4.7130, 2.7606, 3.4791], device='cuda:2'), covar=tensor([0.0200, 0.0325, 0.1329, 0.0238, 0.0680, 0.0345, 0.1299, 0.0541], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0166, 0.0190, 0.0141, 0.0169, 0.0207, 0.0199, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 21:54:58,015 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3236, 5.7209, 5.5168, 5.5220, 5.0779, 5.1198, 5.1989, 5.8190], device='cuda:2'), covar=tensor([0.1277, 0.1081, 0.1124, 0.0755, 0.0932, 0.0774, 0.1048, 0.1028], device='cuda:2'), in_proj_covar=tensor([0.0593, 0.0741, 0.0603, 0.0527, 0.0472, 0.0472, 0.0621, 0.0565], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 21:55:49,508 INFO [train.py:904] (2/8) Epoch 14, batch 700, loss[loss=0.1715, simple_loss=0.2609, pruned_loss=0.04104, over 17281.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2593, pruned_loss=0.04794, over 3204968.31 frames. ], batch size: 52, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:55:55,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5724, 3.3139, 3.7712, 2.6622, 3.4646, 3.7719, 3.6146, 2.0541], device='cuda:2'), covar=tensor([0.0389, 0.0182, 0.0040, 0.0320, 0.0096, 0.0096, 0.0075, 0.0427], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0074, 0.0074, 0.0131, 0.0086, 0.0095, 0.0084, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 21:56:01,321 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:56:38,037 INFO [optim.py:368] (2/8) 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 21:56:39,072 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132687.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:56:58,534 INFO [train.py:904] (2/8) Epoch 14, batch 750, loss[loss=0.1831, simple_loss=0.2569, pruned_loss=0.0546, over 16198.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2592, pruned_loss=0.04741, over 3232620.30 frames. ], batch size: 165, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:57:09,169 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:57:24,361 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:43,135 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:59,785 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:07,420 INFO [train.py:904] (2/8) Epoch 14, batch 800, loss[loss=0.1669, simple_loss=0.2551, pruned_loss=0.03938, over 17153.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2593, pruned_loss=0.04719, over 3246231.58 frames. ], batch size: 46, lr: 4.95e-03, grad_scale: 4.0 2023-04-29 21:58:16,055 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:54,910 INFO [optim.py:368] (2/8) 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 21:59:15,846 INFO [train.py:904] (2/8) Epoch 14, batch 850, loss[loss=0.192, simple_loss=0.2621, pruned_loss=0.06095, over 16552.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2588, pruned_loss=0.04737, over 3261160.18 frames. ], batch size: 146, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 21:59:23,478 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:59:51,248 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-29 22:00:24,478 INFO [train.py:904] (2/8) Epoch 14, batch 900, loss[loss=0.1469, simple_loss=0.2269, pruned_loss=0.03342, over 16317.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2587, pruned_loss=0.04687, over 3277786.66 frames. ], batch size: 36, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:00:48,323 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1582, 5.1669, 4.9049, 4.4392, 4.9916, 1.7198, 4.7514, 4.8604], device='cuda:2'), covar=tensor([0.0071, 0.0071, 0.0183, 0.0375, 0.0089, 0.2738, 0.0123, 0.0190], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0130, 0.0175, 0.0161, 0.0146, 0.0191, 0.0165, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:01:14,368 INFO [optim.py:368] (2/8) 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 22:01:34,126 INFO [train.py:904] (2/8) Epoch 14, batch 950, loss[loss=0.1617, simple_loss=0.2437, pruned_loss=0.03983, over 17010.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2581, pruned_loss=0.04648, over 3288013.72 frames. ], batch size: 41, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:02:33,252 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0367, 5.0065, 4.8102, 4.3036, 4.8926, 2.0760, 4.6481, 4.6881], device='cuda:2'), covar=tensor([0.0083, 0.0079, 0.0168, 0.0327, 0.0080, 0.2307, 0.0123, 0.0177], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0131, 0.0176, 0.0162, 0.0147, 0.0192, 0.0166, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:02:41,756 INFO [train.py:904] (2/8) Epoch 14, batch 1000, loss[loss=0.1443, simple_loss=0.2272, pruned_loss=0.03072, over 16772.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2574, pruned_loss=0.04639, over 3299167.10 frames. ], batch size: 39, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:03:29,371 INFO [optim.py:368] (2/8) 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 22:03:48,002 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2494, 4.6520, 4.1465, 4.6057, 4.3047, 4.1306, 4.1829, 4.7538], device='cuda:2'), covar=tensor([0.2529, 0.2087, 0.2900, 0.1384, 0.1722, 0.2488, 0.2385, 0.2209], device='cuda:2'), in_proj_covar=tensor([0.0598, 0.0747, 0.0607, 0.0531, 0.0476, 0.0477, 0.0625, 0.0571], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:03:50,011 INFO [train.py:904] (2/8) Epoch 14, batch 1050, loss[loss=0.1774, simple_loss=0.2581, pruned_loss=0.04837, over 16773.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2571, pruned_loss=0.04576, over 3313857.66 frames. ], batch size: 102, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:04:06,809 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2163, 5.1866, 4.9257, 4.3932, 5.0199, 1.9440, 4.7641, 4.8625], device='cuda:2'), covar=tensor([0.0081, 0.0076, 0.0175, 0.0358, 0.0083, 0.2534, 0.0122, 0.0168], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0132, 0.0177, 0.0164, 0.0148, 0.0193, 0.0167, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:04:10,800 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:04:59,790 INFO [train.py:904] (2/8) Epoch 14, batch 1100, loss[loss=0.1704, simple_loss=0.2513, pruned_loss=0.04471, over 16710.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2567, pruned_loss=0.04519, over 3322309.06 frames. ], batch size: 134, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:05:47,408 INFO [optim.py:368] (2/8) 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 22:06:08,349 INFO [train.py:904] (2/8) Epoch 14, batch 1150, loss[loss=0.1761, simple_loss=0.2524, pruned_loss=0.04995, over 16394.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2559, pruned_loss=0.04506, over 3327590.40 frames. ], batch size: 75, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:06:08,688 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:06:25,418 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 22:07:03,960 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0673, 2.0388, 1.6918, 1.7344, 2.2549, 2.0522, 2.1359, 2.3828], device='cuda:2'), covar=tensor([0.0210, 0.0314, 0.0418, 0.0415, 0.0186, 0.0281, 0.0189, 0.0237], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0218, 0.0212, 0.0211, 0.0218, 0.0220, 0.0224, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:07:16,495 INFO [train.py:904] (2/8) Epoch 14, batch 1200, loss[loss=0.1969, simple_loss=0.2686, pruned_loss=0.06261, over 15383.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2557, pruned_loss=0.04514, over 3325467.92 frames. ], batch size: 190, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:07:40,476 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 22:08:05,990 INFO [optim.py:368] (2/8) 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 22:08:28,180 INFO [train.py:904] (2/8) Epoch 14, batch 1250, loss[loss=0.1865, simple_loss=0.253, pruned_loss=0.05995, over 16673.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2562, pruned_loss=0.04577, over 3320741.29 frames. ], batch size: 134, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:09:33,425 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6845, 1.7906, 2.1407, 2.4681, 2.5302, 2.3694, 1.7185, 2.7223], device='cuda:2'), covar=tensor([0.0151, 0.0408, 0.0283, 0.0243, 0.0222, 0.0259, 0.0440, 0.0125], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0178, 0.0163, 0.0166, 0.0175, 0.0133, 0.0178, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:09:37,992 INFO [train.py:904] (2/8) Epoch 14, batch 1300, loss[loss=0.1737, simple_loss=0.2477, pruned_loss=0.04989, over 12047.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2564, pruned_loss=0.04577, over 3328256.06 frames. ], batch size: 247, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:10:27,126 INFO [optim.py:368] (2/8) 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 22:10:48,345 INFO [train.py:904] (2/8) Epoch 14, batch 1350, loss[loss=0.1599, simple_loss=0.2573, pruned_loss=0.03124, over 17118.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2548, pruned_loss=0.04446, over 3333504.16 frames. ], batch size: 48, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:11:06,796 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:11:09,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9030, 5.2994, 5.3918, 5.2459, 5.2245, 5.8890, 5.3863, 5.0880], device='cuda:2'), covar=tensor([0.1062, 0.1825, 0.2245, 0.1852, 0.2960, 0.1003, 0.1492, 0.2481], device='cuda:2'), in_proj_covar=tensor([0.0376, 0.0533, 0.0587, 0.0453, 0.0617, 0.0611, 0.0464, 0.0607], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:11:14,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6609, 2.7038, 2.1928, 2.4982, 2.9850, 2.8148, 3.4288, 3.2318], device='cuda:2'), covar=tensor([0.0106, 0.0317, 0.0425, 0.0366, 0.0222, 0.0303, 0.0191, 0.0196], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0218, 0.0211, 0.0210, 0.0216, 0.0219, 0.0224, 0.0210], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:11:52,375 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 22:11:56,459 INFO [train.py:904] (2/8) Epoch 14, batch 1400, loss[loss=0.163, simple_loss=0.2509, pruned_loss=0.03753, over 17231.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2551, pruned_loss=0.04486, over 3330394.05 frames. ], batch size: 45, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:11:58,026 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3770, 5.3633, 5.1772, 4.6899, 4.7114, 5.2702, 5.2696, 4.8393], device='cuda:2'), covar=tensor([0.0592, 0.0386, 0.0327, 0.0326, 0.1138, 0.0392, 0.0328, 0.0759], device='cuda:2'), in_proj_covar=tensor([0.0268, 0.0365, 0.0325, 0.0303, 0.0338, 0.0350, 0.0220, 0.0375], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:12:12,798 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:26,526 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 22:12:29,376 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:44,804 INFO [optim.py:368] (2/8) 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 22:13:06,248 INFO [train.py:904] (2/8) Epoch 14, batch 1450, loss[loss=0.1606, simple_loss=0.2412, pruned_loss=0.04002, over 16498.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2539, pruned_loss=0.04473, over 3328558.98 frames. ], batch size: 146, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:13:06,533 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:13:36,002 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-29 22:13:52,453 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 22:13:55,207 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:12,863 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:14,809 INFO [train.py:904] (2/8) Epoch 14, batch 1500, loss[loss=0.1616, simple_loss=0.2385, pruned_loss=0.04234, over 16799.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2543, pruned_loss=0.04527, over 3324809.60 frames. ], batch size: 39, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:14:35,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4317, 4.2018, 4.4389, 4.6157, 4.6961, 4.2919, 4.4930, 4.6840], device='cuda:2'), covar=tensor([0.1362, 0.1069, 0.1341, 0.0599, 0.0568, 0.1071, 0.1863, 0.0653], device='cuda:2'), in_proj_covar=tensor([0.0588, 0.0727, 0.0873, 0.0743, 0.0560, 0.0572, 0.0586, 0.0687], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:14:46,614 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:15:03,755 INFO [optim.py:368] (2/8) 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 22:15:22,928 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 22:15:23,303 INFO [train.py:904] (2/8) Epoch 14, batch 1550, loss[loss=0.2102, simple_loss=0.2742, pruned_loss=0.07311, over 16474.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2564, pruned_loss=0.04699, over 3322721.54 frames. ], batch size: 75, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:16:10,885 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:16:33,567 INFO [train.py:904] (2/8) Epoch 14, batch 1600, loss[loss=0.17, simple_loss=0.2715, pruned_loss=0.03424, over 17054.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2585, pruned_loss=0.04766, over 3314147.57 frames. ], batch size: 50, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:16:37,067 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8431, 2.4962, 1.9280, 2.2854, 2.9407, 2.7656, 2.9955, 3.0029], device='cuda:2'), covar=tensor([0.0143, 0.0337, 0.0459, 0.0418, 0.0170, 0.0277, 0.0211, 0.0228], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0220, 0.0213, 0.0212, 0.0218, 0.0222, 0.0226, 0.0212], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:16:43,337 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-29 22:17:21,876 INFO [optim.py:368] (2/8) 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 22:17:43,027 INFO [train.py:904] (2/8) Epoch 14, batch 1650, loss[loss=0.1871, simple_loss=0.2543, pruned_loss=0.05994, over 16841.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2597, pruned_loss=0.04822, over 3321381.46 frames. ], batch size: 116, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:37,756 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 22:18:52,510 INFO [train.py:904] (2/8) Epoch 14, batch 1700, loss[loss=0.1773, simple_loss=0.2645, pruned_loss=0.04504, over 15546.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2616, pruned_loss=0.04844, over 3323946.02 frames. ], batch size: 191, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:53,076 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133652.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:19:25,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7614, 4.7133, 5.1741, 5.1166, 5.1862, 4.7837, 4.7928, 4.4962], device='cuda:2'), covar=tensor([0.0303, 0.0563, 0.0355, 0.0425, 0.0425, 0.0385, 0.0902, 0.0505], device='cuda:2'), in_proj_covar=tensor([0.0371, 0.0389, 0.0389, 0.0372, 0.0430, 0.0416, 0.0508, 0.0330], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 22:19:42,412 INFO [optim.py:368] (2/8) 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 22:19:56,986 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6196, 5.9976, 5.7641, 5.8235, 5.4542, 5.1821, 5.4272, 6.1125], device='cuda:2'), covar=tensor([0.1226, 0.0868, 0.1036, 0.0694, 0.0808, 0.0705, 0.1111, 0.0868], device='cuda:2'), in_proj_covar=tensor([0.0599, 0.0750, 0.0606, 0.0535, 0.0477, 0.0478, 0.0629, 0.0574], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:20:03,031 INFO [train.py:904] (2/8) Epoch 14, batch 1750, loss[loss=0.1637, simple_loss=0.2508, pruned_loss=0.03835, over 17200.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2626, pruned_loss=0.04827, over 3315707.97 frames. ], batch size: 44, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:20:19,381 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:20:42,444 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 22:20:44,907 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133732.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:21:12,608 INFO [train.py:904] (2/8) Epoch 14, batch 1800, loss[loss=0.2008, simple_loss=0.284, pruned_loss=0.0588, over 17043.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2625, pruned_loss=0.04756, over 3323058.09 frames. ], batch size: 53, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:22:00,848 INFO [optim.py:368] (2/8) 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 22:22:22,400 INFO [train.py:904] (2/8) Epoch 14, batch 1850, loss[loss=0.1831, simple_loss=0.2739, pruned_loss=0.04619, over 17068.00 frames. ], tot_loss[loss=0.18, simple_loss=0.264, pruned_loss=0.04804, over 3314429.90 frames. ], batch size: 53, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:22:54,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0313, 4.7425, 4.9981, 5.2427, 5.4427, 4.7151, 5.4035, 5.3980], device='cuda:2'), covar=tensor([0.1603, 0.1267, 0.1852, 0.0785, 0.0487, 0.0802, 0.0475, 0.0521], device='cuda:2'), in_proj_covar=tensor([0.0601, 0.0742, 0.0891, 0.0759, 0.0571, 0.0586, 0.0598, 0.0704], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:23:03,017 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:23:32,964 INFO [train.py:904] (2/8) Epoch 14, batch 1900, loss[loss=0.1681, simple_loss=0.2501, pruned_loss=0.04308, over 16396.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2622, pruned_loss=0.04715, over 3316299.54 frames. ], batch size: 75, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:23:44,586 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9352, 4.9807, 5.4123, 5.4242, 5.4317, 5.0600, 5.0242, 4.7941], device='cuda:2'), covar=tensor([0.0299, 0.0466, 0.0376, 0.0379, 0.0423, 0.0355, 0.0869, 0.0407], device='cuda:2'), in_proj_covar=tensor([0.0375, 0.0391, 0.0394, 0.0375, 0.0435, 0.0418, 0.0512, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 22:24:23,480 INFO [optim.py:368] (2/8) 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 22:24:30,978 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0837, 5.4088, 5.1553, 5.1813, 4.8225, 4.7801, 4.8370, 5.4927], device='cuda:2'), covar=tensor([0.1131, 0.0919, 0.1032, 0.0656, 0.0802, 0.0851, 0.1075, 0.0941], device='cuda:2'), in_proj_covar=tensor([0.0598, 0.0748, 0.0606, 0.0533, 0.0475, 0.0476, 0.0623, 0.0571], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:24:32,290 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4218, 3.2748, 2.6690, 2.1182, 2.2684, 2.1209, 3.3528, 3.0451], device='cuda:2'), covar=tensor([0.2517, 0.0747, 0.1548, 0.2408, 0.2281, 0.1928, 0.0525, 0.1168], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0263, 0.0290, 0.0288, 0.0282, 0.0232, 0.0275, 0.0311], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:24:42,152 INFO [train.py:904] (2/8) Epoch 14, batch 1950, loss[loss=0.1722, simple_loss=0.2592, pruned_loss=0.04259, over 17241.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2635, pruned_loss=0.04761, over 3304583.71 frames. ], batch size: 45, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:25:52,905 INFO [train.py:904] (2/8) Epoch 14, batch 2000, loss[loss=0.1966, simple_loss=0.2781, pruned_loss=0.05754, over 16954.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2627, pruned_loss=0.04714, over 3304392.11 frames. ], batch size: 90, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:26:18,939 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8485, 4.3471, 3.0709, 2.3126, 2.8070, 2.5351, 4.6420, 3.7591], device='cuda:2'), covar=tensor([0.2760, 0.0627, 0.1676, 0.2621, 0.2802, 0.1874, 0.0413, 0.1168], device='cuda:2'), in_proj_covar=tensor([0.0310, 0.0262, 0.0290, 0.0289, 0.0283, 0.0233, 0.0276, 0.0312], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:26:43,619 INFO [optim.py:368] (2/8) 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 22:26:57,063 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:06,903 INFO [train.py:904] (2/8) Epoch 14, batch 2050, loss[loss=0.1933, simple_loss=0.2701, pruned_loss=0.05828, over 16707.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2634, pruned_loss=0.04781, over 3304687.43 frames. ], batch size: 124, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:27:09,235 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5186, 3.5543, 3.8505, 2.7807, 3.5332, 3.9448, 3.6288, 2.3084], device='cuda:2'), covar=tensor([0.0388, 0.0166, 0.0046, 0.0284, 0.0083, 0.0072, 0.0076, 0.0394], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0072, 0.0073, 0.0128, 0.0084, 0.0094, 0.0083, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:27:15,545 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:21,606 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1787, 5.1324, 4.8875, 4.3171, 4.9788, 1.8098, 4.7529, 4.8077], device='cuda:2'), covar=tensor([0.0078, 0.0075, 0.0186, 0.0396, 0.0093, 0.2643, 0.0135, 0.0192], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0136, 0.0184, 0.0170, 0.0154, 0.0196, 0.0173, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:27:46,414 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 22:27:49,333 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:28:16,827 INFO [train.py:904] (2/8) Epoch 14, batch 2100, loss[loss=0.1706, simple_loss=0.2735, pruned_loss=0.03385, over 17262.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2641, pruned_loss=0.04779, over 3303846.46 frames. ], batch size: 52, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:28:25,578 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:28:54,050 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:28:56,311 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:29:09,011 INFO [optim.py:368] (2/8) 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 22:29:26,680 INFO [train.py:904] (2/8) Epoch 14, batch 2150, loss[loss=0.1743, simple_loss=0.26, pruned_loss=0.04427, over 16823.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2648, pruned_loss=0.04851, over 3298957.15 frames. ], batch size: 39, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:29:59,786 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 22:30:05,041 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7423, 2.7791, 2.5116, 4.2071, 3.4740, 4.1183, 1.5282, 2.9177], device='cuda:2'), covar=tensor([0.1317, 0.0613, 0.1013, 0.0158, 0.0125, 0.0334, 0.1393, 0.0744], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0163, 0.0181, 0.0161, 0.0198, 0.0210, 0.0186, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:30:06,157 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:30:30,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6547, 2.2489, 2.3805, 4.3546, 2.2178, 2.6397, 2.3399, 2.4746], device='cuda:2'), covar=tensor([0.1052, 0.3664, 0.2592, 0.0431, 0.3987, 0.2546, 0.3113, 0.3551], device='cuda:2'), in_proj_covar=tensor([0.0380, 0.0412, 0.0345, 0.0328, 0.0421, 0.0476, 0.0378, 0.0484], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:30:35,960 INFO [train.py:904] (2/8) Epoch 14, batch 2200, loss[loss=0.2084, simple_loss=0.2808, pruned_loss=0.06807, over 16888.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2642, pruned_loss=0.04796, over 3304565.48 frames. ], batch size: 109, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:31:12,754 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134178.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:31:27,737 INFO [optim.py:368] (2/8) 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 22:31:41,782 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-29 22:31:45,970 INFO [train.py:904] (2/8) Epoch 14, batch 2250, loss[loss=0.1982, simple_loss=0.2742, pruned_loss=0.06116, over 16464.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2661, pruned_loss=0.04871, over 3301260.53 frames. ], batch size: 146, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:32:07,851 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4429, 2.2638, 2.3690, 4.2778, 2.2131, 2.6259, 2.2706, 2.4078], device='cuda:2'), covar=tensor([0.1157, 0.3429, 0.2564, 0.0486, 0.3832, 0.2332, 0.3382, 0.3224], device='cuda:2'), in_proj_covar=tensor([0.0380, 0.0412, 0.0344, 0.0328, 0.0420, 0.0476, 0.0378, 0.0484], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:32:56,451 INFO [train.py:904] (2/8) Epoch 14, batch 2300, loss[loss=0.1935, simple_loss=0.2673, pruned_loss=0.05985, over 16915.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2659, pruned_loss=0.04878, over 3303117.80 frames. ], batch size: 96, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:33:48,101 INFO [optim.py:368] (2/8) 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 22:34:06,358 INFO [train.py:904] (2/8) Epoch 14, batch 2350, loss[loss=0.1553, simple_loss=0.2415, pruned_loss=0.03457, over 16997.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2662, pruned_loss=0.04888, over 3302034.60 frames. ], batch size: 41, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:34:09,718 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:13,085 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-29 22:34:15,132 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:16,976 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:17,012 INFO [train.py:904] (2/8) Epoch 14, batch 2400, loss[loss=0.2003, simple_loss=0.2713, pruned_loss=0.06468, over 16858.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2669, pruned_loss=0.04899, over 3297112.02 frames. ], batch size: 109, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:35:19,204 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:23,355 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:35,549 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:42,551 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:36:08,841 INFO [optim.py:368] (2/8) 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 22:36:25,718 INFO [train.py:904] (2/8) Epoch 14, batch 2450, loss[loss=0.2104, simple_loss=0.3039, pruned_loss=0.0584, over 16658.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2674, pruned_loss=0.04871, over 3298343.74 frames. ], batch size: 62, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:37:35,054 INFO [train.py:904] (2/8) Epoch 14, batch 2500, loss[loss=0.1801, simple_loss=0.269, pruned_loss=0.04565, over 16697.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2666, pruned_loss=0.04824, over 3310359.90 frames. ], batch size: 62, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:37:57,947 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2706, 5.2118, 5.1326, 4.6826, 4.7172, 5.1555, 5.1641, 4.7450], device='cuda:2'), covar=tensor([0.0561, 0.0393, 0.0256, 0.0309, 0.1072, 0.0368, 0.0279, 0.0706], device='cuda:2'), in_proj_covar=tensor([0.0271, 0.0372, 0.0329, 0.0309, 0.0345, 0.0356, 0.0223, 0.0383], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:38:20,575 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1579, 5.4987, 5.1733, 5.2436, 4.9532, 4.8151, 4.9299, 5.6041], device='cuda:2'), covar=tensor([0.1189, 0.0886, 0.1121, 0.0803, 0.0815, 0.0860, 0.1102, 0.0831], device='cuda:2'), in_proj_covar=tensor([0.0603, 0.0755, 0.0615, 0.0542, 0.0479, 0.0481, 0.0629, 0.0578], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:38:25,228 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9989, 2.7330, 2.7480, 2.1584, 2.5804, 2.1409, 2.7454, 2.8561], device='cuda:2'), covar=tensor([0.0285, 0.0676, 0.0521, 0.1536, 0.0727, 0.0889, 0.0598, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0154, 0.0163, 0.0149, 0.0141, 0.0128, 0.0140, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:38:28,206 INFO [optim.py:368] (2/8) 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 22:38:45,442 INFO [train.py:904] (2/8) Epoch 14, batch 2550, loss[loss=0.1919, simple_loss=0.2677, pruned_loss=0.05808, over 16359.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2674, pruned_loss=0.04866, over 3306676.05 frames. ], batch size: 145, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:39:55,150 INFO [train.py:904] (2/8) Epoch 14, batch 2600, loss[loss=0.2082, simple_loss=0.2772, pruned_loss=0.06963, over 16489.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2679, pruned_loss=0.04882, over 3299683.41 frames. ], batch size: 146, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:40:46,917 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 14, batch 2650, loss[loss=0.1923, simple_loss=0.2736, pruned_loss=0.05553, over 15498.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2673, pruned_loss=0.04831, over 3310402.26 frames. ], batch size: 190, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:12,178 INFO [train.py:904] (2/8) Epoch 14, batch 2700, loss[loss=0.1769, simple_loss=0.2687, pruned_loss=0.04251, over 17026.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2672, pruned_loss=0.04753, over 3317662.25 frames. ], batch size: 53, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:13,612 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:18,882 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-04-29 22:42:23,623 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:23,996 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-29 22:42:31,376 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:35,002 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0381, 2.5258, 2.6698, 1.8459, 2.7426, 2.7780, 2.4407, 2.3021], device='cuda:2'), covar=tensor([0.0779, 0.0244, 0.0252, 0.1011, 0.0107, 0.0250, 0.0477, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0104, 0.0090, 0.0139, 0.0072, 0.0115, 0.0123, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 22:42:55,360 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8857, 4.8868, 5.3199, 5.3264, 5.3574, 4.9506, 4.9210, 4.6038], device='cuda:2'), covar=tensor([0.0321, 0.0528, 0.0414, 0.0425, 0.0417, 0.0387, 0.0888, 0.0476], device='cuda:2'), in_proj_covar=tensor([0.0380, 0.0396, 0.0400, 0.0378, 0.0440, 0.0423, 0.0515, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 22:43:04,246 INFO [optim.py:368] (2/8) 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 22:43:20,427 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:43:21,289 INFO [train.py:904] (2/8) Epoch 14, batch 2750, loss[loss=0.1701, simple_loss=0.2612, pruned_loss=0.03952, over 17084.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2677, pruned_loss=0.04704, over 3321199.44 frames. ], batch size: 47, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:43:21,849 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 22:43:59,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7852, 4.4163, 4.3492, 3.2795, 3.6243, 4.4265, 3.8478, 2.6658], device='cuda:2'), covar=tensor([0.0401, 0.0050, 0.0033, 0.0266, 0.0104, 0.0074, 0.0072, 0.0352], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0074, 0.0074, 0.0129, 0.0086, 0.0095, 0.0084, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:44:29,010 INFO [train.py:904] (2/8) Epoch 14, batch 2800, loss[loss=0.2065, simple_loss=0.2814, pruned_loss=0.06579, over 12260.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2671, pruned_loss=0.0469, over 3326469.56 frames. ], batch size: 246, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:44:34,877 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8583, 2.8662, 2.4633, 4.4203, 3.3998, 4.1306, 1.7014, 2.9809], device='cuda:2'), covar=tensor([0.1388, 0.0811, 0.1247, 0.0223, 0.0366, 0.0499, 0.1546, 0.0867], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0163, 0.0183, 0.0164, 0.0200, 0.0212, 0.0187, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:45:20,154 INFO [optim.py:368] (2/8) 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 22:45:20,602 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5317, 3.6727, 2.2795, 3.8851, 2.7981, 3.8566, 2.2957, 2.9569], device='cuda:2'), covar=tensor([0.0238, 0.0351, 0.1305, 0.0254, 0.0686, 0.0637, 0.1324, 0.0612], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0171, 0.0190, 0.0148, 0.0169, 0.0213, 0.0199, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:45:37,176 INFO [train.py:904] (2/8) Epoch 14, batch 2850, loss[loss=0.1856, simple_loss=0.2784, pruned_loss=0.04637, over 17120.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2667, pruned_loss=0.04719, over 3330033.27 frames. ], batch size: 49, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:45,299 INFO [train.py:904] (2/8) Epoch 14, batch 2900, loss[loss=0.1689, simple_loss=0.2601, pruned_loss=0.03889, over 16574.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2656, pruned_loss=0.047, over 3333436.60 frames. ], batch size: 62, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:58,784 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:47:28,213 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9974, 3.2513, 3.0073, 5.1466, 4.2736, 4.6090, 1.5478, 3.3630], device='cuda:2'), covar=tensor([0.1260, 0.0644, 0.0975, 0.0175, 0.0261, 0.0362, 0.1552, 0.0690], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0163, 0.0182, 0.0163, 0.0200, 0.0211, 0.0186, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:47:36,338 INFO [optim.py:368] (2/8) 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 22:47:54,145 INFO [train.py:904] (2/8) Epoch 14, batch 2950, loss[loss=0.244, simple_loss=0.2964, pruned_loss=0.09574, over 11836.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2663, pruned_loss=0.04808, over 3315180.72 frames. ], batch size: 247, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:48:06,738 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7039, 3.6938, 2.9114, 2.2433, 2.5055, 2.2985, 3.7599, 3.3538], device='cuda:2'), covar=tensor([0.2397, 0.0576, 0.1396, 0.2682, 0.2326, 0.1860, 0.0497, 0.1261], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0262, 0.0290, 0.0289, 0.0283, 0.0232, 0.0275, 0.0314], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:48:24,014 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:02,687 INFO [train.py:904] (2/8) Epoch 14, batch 3000, loss[loss=0.149, simple_loss=0.2301, pruned_loss=0.0339, over 16745.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2658, pruned_loss=0.04824, over 3313196.82 frames. ], batch size: 39, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:49:02,687 INFO [train.py:929] (2/8) Computing validation loss 2023-04-29 22:49:12,417 INFO [train.py:938] (2/8) Epoch 14, validation: loss=0.1382, simple_loss=0.244, pruned_loss=0.01621, over 944034.00 frames. 2023-04-29 22:49:12,417 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-29 22:49:13,956 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2495, 4.9858, 5.2122, 5.4302, 5.6531, 4.8506, 5.5926, 5.6031], device='cuda:2'), covar=tensor([0.1368, 0.1161, 0.1590, 0.0732, 0.0453, 0.0758, 0.0450, 0.0521], device='cuda:2'), in_proj_covar=tensor([0.0599, 0.0740, 0.0891, 0.0766, 0.0571, 0.0588, 0.0601, 0.0699], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:49:24,938 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:31,518 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:37,650 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0067, 4.1240, 2.7676, 4.7093, 3.3613, 4.7177, 2.7296, 3.4057], device='cuda:2'), covar=tensor([0.0224, 0.0364, 0.1252, 0.0239, 0.0633, 0.0431, 0.1296, 0.0614], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0171, 0.0189, 0.0147, 0.0169, 0.0213, 0.0199, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:50:06,979 INFO [optim.py:368] (2/8) 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 22:50:24,227 INFO [train.py:904] (2/8) Epoch 14, batch 3050, loss[loss=0.1733, simple_loss=0.2635, pruned_loss=0.04153, over 17223.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2664, pruned_loss=0.0491, over 3305357.29 frames. ], batch size: 46, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:50:33,225 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:40,114 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:51:32,224 INFO [train.py:904] (2/8) Epoch 14, batch 3100, loss[loss=0.1661, simple_loss=0.2512, pruned_loss=0.04047, over 16751.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2657, pruned_loss=0.04835, over 3308092.95 frames. ], batch size: 39, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:52:07,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9741, 4.4977, 4.4100, 3.3573, 3.7780, 4.5367, 4.0462, 2.6002], device='cuda:2'), covar=tensor([0.0373, 0.0036, 0.0036, 0.0277, 0.0100, 0.0064, 0.0060, 0.0374], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0073, 0.0073, 0.0128, 0.0084, 0.0094, 0.0083, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:52:24,972 INFO [optim.py:368] (2/8) 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 22:52:40,980 INFO [train.py:904] (2/8) Epoch 14, batch 3150, loss[loss=0.1907, simple_loss=0.2818, pruned_loss=0.04979, over 17094.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2643, pruned_loss=0.04808, over 3309367.24 frames. ], batch size: 53, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:53:46,462 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4152, 5.8046, 5.3322, 5.7721, 5.2415, 5.0465, 5.4591, 5.9069], device='cuda:2'), covar=tensor([0.2135, 0.1657, 0.2441, 0.1165, 0.1612, 0.1244, 0.1891, 0.1926], device='cuda:2'), in_proj_covar=tensor([0.0607, 0.0758, 0.0619, 0.0546, 0.0482, 0.0483, 0.0634, 0.0585], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:53:50,739 INFO [train.py:904] (2/8) Epoch 14, batch 3200, loss[loss=0.1532, simple_loss=0.2352, pruned_loss=0.0356, over 16724.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2633, pruned_loss=0.0478, over 3321124.82 frames. ], batch size: 89, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:54:17,365 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:54:41,146 INFO [optim.py:368] (2/8) 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 22:54:59,014 INFO [train.py:904] (2/8) Epoch 14, batch 3250, loss[loss=0.1834, simple_loss=0.2583, pruned_loss=0.0543, over 16472.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.264, pruned_loss=0.04776, over 3318951.66 frames. ], batch size: 75, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:55:22,124 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:55:41,172 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 14, batch 3300, loss[loss=0.182, simple_loss=0.2645, pruned_loss=0.04975, over 16814.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2646, pruned_loss=0.048, over 3314695.05 frames. ], batch size: 83, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:56:21,361 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0695, 4.4614, 3.3398, 2.3545, 2.9376, 2.7367, 4.7305, 3.9051], device='cuda:2'), covar=tensor([0.2419, 0.0680, 0.1593, 0.2645, 0.2869, 0.1828, 0.0409, 0.1240], device='cuda:2'), in_proj_covar=tensor([0.0310, 0.0264, 0.0291, 0.0290, 0.0286, 0.0233, 0.0276, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:56:28,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3609, 4.1545, 4.3854, 4.5590, 4.6509, 4.2128, 4.4552, 4.6318], device='cuda:2'), covar=tensor([0.1484, 0.1024, 0.1339, 0.0588, 0.0544, 0.1184, 0.1981, 0.0577], device='cuda:2'), in_proj_covar=tensor([0.0602, 0.0743, 0.0896, 0.0764, 0.0573, 0.0587, 0.0602, 0.0702], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 22:56:52,865 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7412, 3.8600, 2.1787, 4.4426, 2.9574, 4.4390, 2.5064, 3.1683], device='cuda:2'), covar=tensor([0.0245, 0.0329, 0.1645, 0.0246, 0.0754, 0.0460, 0.1349, 0.0633], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0173, 0.0191, 0.0149, 0.0170, 0.0215, 0.0200, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:57:01,617 INFO [optim.py:368] (2/8) 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 22:57:05,004 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9263, 3.3861, 2.9293, 5.1815, 4.3636, 4.6664, 1.7242, 3.4003], device='cuda:2'), covar=tensor([0.1266, 0.0658, 0.1021, 0.0162, 0.0218, 0.0323, 0.1462, 0.0690], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0164, 0.0184, 0.0164, 0.0202, 0.0212, 0.0187, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:57:16,993 INFO [train.py:904] (2/8) Epoch 14, batch 3350, loss[loss=0.1881, simple_loss=0.2712, pruned_loss=0.05249, over 16797.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2642, pruned_loss=0.04743, over 3316779.62 frames. ], batch size: 57, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:57:46,168 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7059, 4.7361, 4.9341, 4.7791, 4.7313, 5.3729, 4.9115, 4.5775], device='cuda:2'), covar=tensor([0.1367, 0.1919, 0.1820, 0.2000, 0.2626, 0.0969, 0.1447, 0.2272], device='cuda:2'), in_proj_covar=tensor([0.0383, 0.0544, 0.0591, 0.0462, 0.0625, 0.0621, 0.0471, 0.0615], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:58:06,460 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 22:58:24,246 INFO [train.py:904] (2/8) Epoch 14, batch 3400, loss[loss=0.1666, simple_loss=0.2461, pruned_loss=0.04351, over 16927.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2641, pruned_loss=0.04741, over 3316320.45 frames. ], batch size: 96, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:58:56,544 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5786, 2.9641, 2.8217, 4.9307, 4.0119, 4.4605, 1.5606, 3.1882], device='cuda:2'), covar=tensor([0.1409, 0.0718, 0.1025, 0.0197, 0.0218, 0.0374, 0.1557, 0.0705], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0164, 0.0183, 0.0164, 0.0201, 0.0212, 0.0187, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 22:59:12,544 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6517, 4.5766, 4.5559, 4.2790, 4.2512, 4.6450, 4.4293, 4.3681], device='cuda:2'), covar=tensor([0.0658, 0.0704, 0.0289, 0.0295, 0.0851, 0.0435, 0.0511, 0.0647], device='cuda:2'), in_proj_covar=tensor([0.0272, 0.0374, 0.0330, 0.0309, 0.0346, 0.0359, 0.0225, 0.0386], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 22:59:16,982 INFO [optim.py:368] (2/8) 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 22:59:32,657 INFO [train.py:904] (2/8) Epoch 14, batch 3450, loss[loss=0.1858, simple_loss=0.2549, pruned_loss=0.05841, over 16750.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2627, pruned_loss=0.04684, over 3318210.89 frames. ], batch size: 124, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:59:49,255 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 23:00:41,081 INFO [train.py:904] (2/8) Epoch 14, batch 3500, loss[loss=0.1856, simple_loss=0.2724, pruned_loss=0.0494, over 16492.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2617, pruned_loss=0.04637, over 3318015.95 frames. ], batch size: 68, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:00:54,365 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5610, 2.5751, 1.8161, 2.7286, 2.1145, 2.7644, 2.1076, 2.3585], device='cuda:2'), covar=tensor([0.0269, 0.0354, 0.1344, 0.0261, 0.0647, 0.0406, 0.1087, 0.0578], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0175, 0.0193, 0.0151, 0.0173, 0.0218, 0.0203, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 23:01:37,086 INFO [optim.py:368] (2/8) 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 23:01:51,741 INFO [train.py:904] (2/8) Epoch 14, batch 3550, loss[loss=0.1835, simple_loss=0.263, pruned_loss=0.05195, over 16431.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2616, pruned_loss=0.04624, over 3319359.66 frames. ], batch size: 75, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:02:15,148 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:27,272 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:41,911 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9462, 2.9608, 2.5166, 2.7638, 3.2265, 3.0418, 3.6444, 3.4479], device='cuda:2'), covar=tensor([0.0104, 0.0304, 0.0385, 0.0333, 0.0218, 0.0283, 0.0198, 0.0216], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0220, 0.0215, 0.0214, 0.0221, 0.0222, 0.0232, 0.0216], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:02:41,939 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9613, 3.4104, 3.0128, 5.2190, 4.3293, 4.6684, 1.8443, 3.5058], device='cuda:2'), covar=tensor([0.1215, 0.0622, 0.0920, 0.0142, 0.0193, 0.0329, 0.1391, 0.0591], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0183, 0.0164, 0.0201, 0.0212, 0.0187, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 23:03:02,250 INFO [train.py:904] (2/8) Epoch 14, batch 3600, loss[loss=0.1972, simple_loss=0.2721, pruned_loss=0.06112, over 16737.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2608, pruned_loss=0.04603, over 3325360.33 frames. ], batch size: 134, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:03:06,352 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8117, 4.3405, 3.1436, 2.2345, 2.6736, 2.6315, 4.6646, 3.5710], device='cuda:2'), covar=tensor([0.2596, 0.0534, 0.1550, 0.2481, 0.2760, 0.1719, 0.0335, 0.1224], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0262, 0.0289, 0.0288, 0.0285, 0.0232, 0.0274, 0.0313], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:03:22,564 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:03:58,635 INFO [optim.py:368] (2/8) 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 23:04:14,253 INFO [train.py:904] (2/8) Epoch 14, batch 3650, loss[loss=0.1788, simple_loss=0.2681, pruned_loss=0.0447, over 16480.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.26, pruned_loss=0.04629, over 3289897.17 frames. ], batch size: 62, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:04:57,263 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:05:31,030 INFO [train.py:904] (2/8) Epoch 14, batch 3700, loss[loss=0.1652, simple_loss=0.2404, pruned_loss=0.045, over 16620.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.259, pruned_loss=0.04787, over 3275539.18 frames. ], batch size: 89, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:06:30,608 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:06:32,146 INFO [optim.py:368] (2/8) 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 23:06:46,314 INFO [train.py:904] (2/8) Epoch 14, batch 3750, loss[loss=0.1884, simple_loss=0.2611, pruned_loss=0.05786, over 16504.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.259, pruned_loss=0.04896, over 3276932.08 frames. ], batch size: 75, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:08:01,576 INFO [train.py:904] (2/8) Epoch 14, batch 3800, loss[loss=0.2093, simple_loss=0.2866, pruned_loss=0.06604, over 15549.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.26, pruned_loss=0.0502, over 3284178.34 frames. ], batch size: 190, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:00,855 INFO [optim.py:368] (2/8) 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 23:09:16,028 INFO [train.py:904] (2/8) Epoch 14, batch 3850, loss[loss=0.1811, simple_loss=0.2605, pruned_loss=0.05089, over 16458.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2599, pruned_loss=0.05089, over 3293747.10 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:26,535 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2088, 4.1806, 4.1147, 3.9242, 3.8738, 4.1963, 3.8831, 3.9902], device='cuda:2'), covar=tensor([0.0559, 0.0562, 0.0249, 0.0233, 0.0712, 0.0419, 0.0771, 0.0533], device='cuda:2'), in_proj_covar=tensor([0.0272, 0.0374, 0.0327, 0.0307, 0.0344, 0.0358, 0.0223, 0.0383], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:09:53,600 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:10:03,985 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5047, 4.1344, 4.1671, 2.8620, 3.8034, 4.2585, 3.9007, 2.4469], device='cuda:2'), covar=tensor([0.0406, 0.0063, 0.0033, 0.0309, 0.0068, 0.0068, 0.0054, 0.0344], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0072, 0.0072, 0.0127, 0.0085, 0.0094, 0.0083, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:10:30,026 INFO [train.py:904] (2/8) Epoch 14, batch 3900, loss[loss=0.1926, simple_loss=0.2657, pruned_loss=0.05979, over 16853.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2595, pruned_loss=0.05128, over 3292077.53 frames. ], batch size: 116, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:11:05,162 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:11:30,216 INFO [optim.py:368] (2/8) 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 23:11:45,241 INFO [train.py:904] (2/8) Epoch 14, batch 3950, loss[loss=0.1816, simple_loss=0.2504, pruned_loss=0.0564, over 16949.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2596, pruned_loss=0.05218, over 3284757.21 frames. ], batch size: 90, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:12:05,742 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7427, 5.0731, 4.8278, 4.8247, 4.5851, 4.5271, 4.5125, 5.1222], device='cuda:2'), covar=tensor([0.1121, 0.0860, 0.0964, 0.0781, 0.0804, 0.1070, 0.1057, 0.0880], device='cuda:2'), in_proj_covar=tensor([0.0600, 0.0741, 0.0606, 0.0538, 0.0477, 0.0476, 0.0623, 0.0574], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:12:58,425 INFO [train.py:904] (2/8) Epoch 14, batch 4000, loss[loss=0.1627, simple_loss=0.2459, pruned_loss=0.03974, over 16877.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2595, pruned_loss=0.05269, over 3293696.10 frames. ], batch size: 90, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:13:41,762 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:47,646 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:55,966 INFO [optim.py:368] (2/8) 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 23:13:58,925 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:14:13,530 INFO [train.py:904] (2/8) Epoch 14, batch 4050, loss[loss=0.1774, simple_loss=0.2543, pruned_loss=0.05024, over 16368.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2594, pruned_loss=0.05152, over 3305092.54 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:12,434 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:15:21,409 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-04-29 23:15:24,937 INFO [train.py:904] (2/8) Epoch 14, batch 4100, loss[loss=0.1867, simple_loss=0.2688, pruned_loss=0.05228, over 17205.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2607, pruned_loss=0.05093, over 3301723.86 frames. ], batch size: 46, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:29,858 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:15:44,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8615, 5.1110, 5.3307, 5.1696, 5.2017, 5.7703, 5.2291, 4.9771], device='cuda:2'), covar=tensor([0.0871, 0.1647, 0.1599, 0.1593, 0.2055, 0.0757, 0.1325, 0.2096], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0533, 0.0575, 0.0450, 0.0610, 0.0604, 0.0462, 0.0600], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:16:15,615 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7113, 1.3736, 1.6720, 1.7231, 1.7583, 1.9143, 1.5341, 1.8134], device='cuda:2'), covar=tensor([0.0200, 0.0302, 0.0150, 0.0229, 0.0222, 0.0144, 0.0317, 0.0110], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0179, 0.0164, 0.0172, 0.0180, 0.0136, 0.0181, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:16:24,202 INFO [optim.py:368] (2/8) 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 23:16:40,524 INFO [train.py:904] (2/8) Epoch 14, batch 4150, loss[loss=0.2134, simple_loss=0.2975, pruned_loss=0.06467, over 16515.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2684, pruned_loss=0.05372, over 3256709.07 frames. ], batch size: 68, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:17:41,238 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:17:55,761 INFO [train.py:904] (2/8) Epoch 14, batch 4200, loss[loss=0.215, simple_loss=0.3083, pruned_loss=0.06089, over 16689.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2758, pruned_loss=0.05558, over 3241343.82 frames. ], batch size: 134, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:18:55,502 INFO [optim.py:368] (2/8) 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 23:19:10,103 INFO [train.py:904] (2/8) Epoch 14, batch 4250, loss[loss=0.1722, simple_loss=0.265, pruned_loss=0.03975, over 16391.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2793, pruned_loss=0.05572, over 3224820.61 frames. ], batch size: 146, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:19:12,553 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:19:16,705 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4956, 2.6015, 2.0419, 2.3715, 3.0501, 2.6979, 3.2097, 3.2394], device='cuda:2'), covar=tensor([0.0088, 0.0322, 0.0452, 0.0383, 0.0170, 0.0282, 0.0167, 0.0193], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0215, 0.0211, 0.0211, 0.0217, 0.0217, 0.0224, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:20:23,645 INFO [train.py:904] (2/8) Epoch 14, batch 4300, loss[loss=0.1892, simple_loss=0.2805, pruned_loss=0.04897, over 16731.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2803, pruned_loss=0.05472, over 3212321.18 frames. ], batch size: 134, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:21:14,584 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:21:23,663 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 14, batch 4350, loss[loss=0.2033, simple_loss=0.2931, pruned_loss=0.05674, over 16867.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.283, pruned_loss=0.05542, over 3211517.53 frames. ], batch size: 90, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:22:00,651 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8279, 5.0875, 4.8369, 4.8188, 4.5793, 4.5211, 4.5141, 5.1664], device='cuda:2'), covar=tensor([0.0963, 0.0744, 0.1001, 0.0815, 0.0777, 0.0968, 0.1019, 0.0805], device='cuda:2'), in_proj_covar=tensor([0.0588, 0.0728, 0.0595, 0.0530, 0.0468, 0.0468, 0.0611, 0.0564], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:22:18,327 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8016, 2.8947, 2.9627, 1.4448, 3.0686, 3.2025, 2.5359, 2.3049], device='cuda:2'), covar=tensor([0.1151, 0.0250, 0.0224, 0.1410, 0.0090, 0.0137, 0.0530, 0.0623], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0105, 0.0089, 0.0138, 0.0071, 0.0113, 0.0123, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-29 23:22:26,307 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:29,245 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:33,213 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:51,180 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:53,919 INFO [train.py:904] (2/8) Epoch 14, batch 4400, loss[loss=0.206, simple_loss=0.2946, pruned_loss=0.05872, over 16789.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.285, pruned_loss=0.05657, over 3207741.40 frames. ], batch size: 124, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:23:51,149 INFO [optim.py:368] (2/8) 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 23:23:57,445 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:24:06,498 INFO [train.py:904] (2/8) Epoch 14, batch 4450, loss[loss=0.2172, simple_loss=0.3031, pruned_loss=0.06559, over 17039.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2884, pruned_loss=0.05813, over 3204043.54 frames. ], batch size: 53, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:18,607 INFO [train.py:904] (2/8) Epoch 14, batch 4500, loss[loss=0.2019, simple_loss=0.2782, pruned_loss=0.06278, over 16221.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2882, pruned_loss=0.05857, over 3202156.29 frames. ], batch size: 165, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:26:01,686 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7279, 4.9935, 5.2357, 4.8971, 4.9917, 5.6691, 5.1035, 4.8276], device='cuda:2'), covar=tensor([0.1025, 0.1807, 0.1949, 0.2097, 0.2524, 0.0889, 0.1406, 0.2427], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0530, 0.0570, 0.0448, 0.0604, 0.0599, 0.0458, 0.0598], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:26:18,140 INFO [optim.py:368] (2/8) 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 23:26:27,108 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 14, batch 4550, loss[loss=0.2019, simple_loss=0.2768, pruned_loss=0.06353, over 11845.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2899, pruned_loss=0.05954, over 3212024.77 frames. ], batch size: 248, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:27:45,692 INFO [train.py:904] (2/8) Epoch 14, batch 4600, loss[loss=0.1948, simple_loss=0.285, pruned_loss=0.05231, over 16690.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.291, pruned_loss=0.05968, over 3215801.61 frames. ], batch size: 62, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:28:11,327 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8902, 5.3586, 5.5543, 5.2559, 5.2691, 5.9590, 5.4272, 5.1215], device='cuda:2'), covar=tensor([0.0929, 0.1654, 0.1737, 0.2071, 0.2572, 0.0878, 0.1233, 0.2391], device='cuda:2'), in_proj_covar=tensor([0.0370, 0.0525, 0.0566, 0.0443, 0.0597, 0.0595, 0.0453, 0.0592], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:28:42,905 INFO [optim.py:368] (2/8) 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 23:28:57,101 INFO [train.py:904] (2/8) Epoch 14, batch 4650, loss[loss=0.2122, simple_loss=0.3016, pruned_loss=0.06137, over 16485.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2896, pruned_loss=0.05913, over 3227542.05 frames. ], batch size: 75, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:29:03,026 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5606, 4.5900, 4.9229, 4.8793, 4.9429, 4.5467, 4.5393, 4.2549], device='cuda:2'), covar=tensor([0.0257, 0.0361, 0.0292, 0.0390, 0.0351, 0.0304, 0.0881, 0.0512], device='cuda:2'), in_proj_covar=tensor([0.0350, 0.0368, 0.0366, 0.0351, 0.0414, 0.0392, 0.0483, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 23:29:37,580 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 23:29:52,298 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:29:58,789 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:07,966 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:10,584 INFO [train.py:904] (2/8) Epoch 14, batch 4700, loss[loss=0.1835, simple_loss=0.2787, pruned_loss=0.04415, over 16913.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2867, pruned_loss=0.05779, over 3227931.83 frames. ], batch size: 109, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:30:29,987 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4892, 1.8073, 2.1175, 2.4648, 2.4109, 2.7636, 1.7230, 2.5721], device='cuda:2'), covar=tensor([0.0189, 0.0399, 0.0302, 0.0281, 0.0268, 0.0183, 0.0455, 0.0129], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0163, 0.0171, 0.0179, 0.0135, 0.0180, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:31:01,304 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:08,483 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:09,910 INFO [optim.py:368] (2/8) 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 23:31:13,120 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 23:31:18,048 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:23,574 INFO [train.py:904] (2/8) Epoch 14, batch 4750, loss[loss=0.2531, simple_loss=0.3172, pruned_loss=0.09446, over 11850.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2835, pruned_loss=0.05644, over 3216628.05 frames. ], batch size: 248, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:27,814 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:32:03,425 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2784, 2.2790, 2.7978, 3.1167, 2.9686, 3.7866, 2.1702, 3.5499], device='cuda:2'), covar=tensor([0.0157, 0.0346, 0.0242, 0.0252, 0.0230, 0.0105, 0.0429, 0.0097], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0177, 0.0161, 0.0170, 0.0177, 0.0134, 0.0179, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:32:36,916 INFO [train.py:904] (2/8) Epoch 14, batch 4800, loss[loss=0.1792, simple_loss=0.2684, pruned_loss=0.04499, over 16599.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2795, pruned_loss=0.05439, over 3212402.37 frames. ], batch size: 62, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:21,051 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 23:33:36,431 INFO [optim.py:368] (2/8) 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 23:33:46,372 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:33:52,185 INFO [train.py:904] (2/8) Epoch 14, batch 4850, loss[loss=0.1677, simple_loss=0.2658, pruned_loss=0.03485, over 16388.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2805, pruned_loss=0.05358, over 3206857.21 frames. ], batch size: 146, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:58,207 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:35:05,044 INFO [train.py:904] (2/8) Epoch 14, batch 4900, loss[loss=0.2101, simple_loss=0.3032, pruned_loss=0.05854, over 15535.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2802, pruned_loss=0.05248, over 3197934.93 frames. ], batch size: 191, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:35:10,011 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8909, 4.8711, 4.7860, 4.0567, 4.7575, 1.7726, 4.5333, 4.6072], device='cuda:2'), covar=tensor([0.0067, 0.0071, 0.0116, 0.0458, 0.0097, 0.2392, 0.0120, 0.0180], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0130, 0.0174, 0.0165, 0.0148, 0.0188, 0.0165, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:35:28,805 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:36:02,817 INFO [optim.py:368] (2/8) 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 23:36:18,496 INFO [train.py:904] (2/8) Epoch 14, batch 4950, loss[loss=0.1937, simple_loss=0.28, pruned_loss=0.05364, over 16955.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2791, pruned_loss=0.05159, over 3195163.15 frames. ], batch size: 41, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:37:30,155 INFO [train.py:904] (2/8) Epoch 14, batch 5000, loss[loss=0.1787, simple_loss=0.2729, pruned_loss=0.04222, over 16896.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2808, pruned_loss=0.05145, over 3190804.72 frames. ], batch size: 42, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:38:23,434 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 23:38:24,230 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136991.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:38:25,065 INFO [optim.py:368] (2/8) 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 23:38:36,206 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136999.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:38:39,688 INFO [train.py:904] (2/8) Epoch 14, batch 5050, loss[loss=0.176, simple_loss=0.2585, pruned_loss=0.04674, over 16698.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2814, pruned_loss=0.05129, over 3193684.17 frames. ], batch size: 57, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:39:19,854 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 23:39:31,296 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:39:41,688 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-29 23:39:49,977 INFO [train.py:904] (2/8) Epoch 14, batch 5100, loss[loss=0.1695, simple_loss=0.2537, pruned_loss=0.04266, over 16596.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.279, pruned_loss=0.05033, over 3204890.38 frames. ], batch size: 62, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:39:59,237 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 23:40:45,951 INFO [optim.py:368] (2/8) 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 23:41:00,263 INFO [train.py:904] (2/8) Epoch 14, batch 5150, loss[loss=0.2205, simple_loss=0.3207, pruned_loss=0.06012, over 16769.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2794, pruned_loss=0.05002, over 3201992.51 frames. ], batch size: 124, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:42:08,569 INFO [train.py:904] (2/8) Epoch 14, batch 5200, loss[loss=0.2257, simple_loss=0.3022, pruned_loss=0.07458, over 12501.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2785, pruned_loss=0.04985, over 3177978.44 frames. ], batch size: 247, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:42:22,389 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:43:03,334 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 14, batch 5250, loss[loss=0.1849, simple_loss=0.2793, pruned_loss=0.04524, over 16897.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2759, pruned_loss=0.04954, over 3193414.06 frames. ], batch size: 109, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:44:26,110 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 23:44:26,543 INFO [train.py:904] (2/8) Epoch 14, batch 5300, loss[loss=0.154, simple_loss=0.2444, pruned_loss=0.03175, over 16929.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2721, pruned_loss=0.04831, over 3209436.55 frames. ], batch size: 96, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:45:23,602 INFO [optim.py:368] (2/8) 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 23:45:33,777 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 14, batch 5350, loss[loss=0.2112, simple_loss=0.2978, pruned_loss=0.06232, over 16249.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.271, pruned_loss=0.04782, over 3205605.67 frames. ], batch size: 165, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:46:40,458 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:46:46,692 INFO [train.py:904] (2/8) Epoch 14, batch 5400, loss[loss=0.2297, simple_loss=0.305, pruned_loss=0.07725, over 11847.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2735, pruned_loss=0.04845, over 3210255.26 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:47:05,296 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 23:47:46,867 INFO [optim.py:368] (2/8) 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 23:48:02,464 INFO [train.py:904] (2/8) Epoch 14, batch 5450, loss[loss=0.2138, simple_loss=0.2905, pruned_loss=0.06854, over 11835.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2762, pruned_loss=0.04957, over 3210752.18 frames. ], batch size: 246, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:48:44,273 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9072, 2.0977, 2.3943, 3.1645, 2.1179, 2.2992, 2.3098, 2.1825], device='cuda:2'), covar=tensor([0.1050, 0.2897, 0.1947, 0.0576, 0.3631, 0.2009, 0.2764, 0.2960], device='cuda:2'), in_proj_covar=tensor([0.0369, 0.0407, 0.0339, 0.0317, 0.0414, 0.0468, 0.0371, 0.0474], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:49:15,128 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3643, 3.3653, 2.6753, 2.0420, 2.2658, 2.2274, 3.5011, 3.2083], device='cuda:2'), covar=tensor([0.3070, 0.0803, 0.1717, 0.2578, 0.2609, 0.1975, 0.0601, 0.1162], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0259, 0.0288, 0.0289, 0.0283, 0.0229, 0.0276, 0.0307], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:49:19,958 INFO [train.py:904] (2/8) Epoch 14, batch 5500, loss[loss=0.2136, simple_loss=0.2931, pruned_loss=0.06712, over 17051.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2836, pruned_loss=0.0542, over 3182553.29 frames. ], batch size: 55, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:35,322 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:49:47,822 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6754, 4.6308, 4.4629, 3.8284, 4.5535, 1.8423, 4.3570, 4.3065], device='cuda:2'), covar=tensor([0.0087, 0.0070, 0.0140, 0.0359, 0.0087, 0.2386, 0.0122, 0.0192], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0130, 0.0176, 0.0166, 0.0148, 0.0189, 0.0166, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:50:23,833 INFO [optim.py:368] (2/8) 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 23:50:34,946 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6433, 4.7485, 4.9688, 4.7820, 4.7741, 5.3160, 4.8418, 4.6351], device='cuda:2'), covar=tensor([0.1077, 0.1792, 0.1923, 0.1751, 0.2278, 0.0922, 0.1464, 0.2344], device='cuda:2'), in_proj_covar=tensor([0.0369, 0.0517, 0.0563, 0.0439, 0.0593, 0.0588, 0.0448, 0.0592], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:50:36,992 INFO [train.py:904] (2/8) Epoch 14, batch 5550, loss[loss=0.2346, simple_loss=0.3129, pruned_loss=0.07819, over 17184.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2914, pruned_loss=0.05955, over 3172991.90 frames. ], batch size: 46, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:50:45,214 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6048, 4.0598, 3.9786, 2.7652, 3.5639, 4.0348, 3.7022, 2.2203], device='cuda:2'), covar=tensor([0.0396, 0.0031, 0.0039, 0.0309, 0.0090, 0.0079, 0.0064, 0.0385], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0073, 0.0073, 0.0129, 0.0086, 0.0096, 0.0084, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-29 23:50:51,278 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:51:55,242 INFO [train.py:904] (2/8) Epoch 14, batch 5600, loss[loss=0.2184, simple_loss=0.3004, pruned_loss=0.06815, over 16689.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2969, pruned_loss=0.06429, over 3134773.23 frames. ], batch size: 134, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:52:26,601 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5958, 4.6216, 4.4790, 4.2219, 4.1568, 4.5534, 4.3980, 4.2798], device='cuda:2'), covar=tensor([0.0587, 0.0453, 0.0272, 0.0297, 0.0962, 0.0438, 0.0411, 0.0663], device='cuda:2'), in_proj_covar=tensor([0.0256, 0.0351, 0.0306, 0.0288, 0.0324, 0.0339, 0.0207, 0.0361], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:52:53,507 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1764, 4.1812, 4.5432, 4.5086, 4.5606, 4.2166, 4.2475, 4.1559], device='cuda:2'), covar=tensor([0.0336, 0.0593, 0.0422, 0.0466, 0.0497, 0.0414, 0.0982, 0.0536], device='cuda:2'), in_proj_covar=tensor([0.0352, 0.0373, 0.0372, 0.0356, 0.0418, 0.0398, 0.0491, 0.0317], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 23:53:03,058 INFO [optim.py:368] (2/8) 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 23:53:17,475 INFO [train.py:904] (2/8) Epoch 14, batch 5650, loss[loss=0.2288, simple_loss=0.3197, pruned_loss=0.06894, over 16903.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3014, pruned_loss=0.06827, over 3120105.97 frames. ], batch size: 96, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:53:21,964 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2096, 3.3690, 3.5676, 3.5372, 3.5567, 3.3439, 3.3982, 3.4302], device='cuda:2'), covar=tensor([0.0395, 0.0661, 0.0453, 0.0474, 0.0502, 0.0572, 0.0833, 0.0571], device='cuda:2'), in_proj_covar=tensor([0.0350, 0.0370, 0.0370, 0.0354, 0.0416, 0.0396, 0.0487, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-29 23:53:41,028 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 23:53:41,139 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 23:53:49,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5759, 2.8184, 2.4653, 4.2114, 2.7686, 3.9696, 1.5365, 2.6869], device='cuda:2'), covar=tensor([0.1624, 0.0797, 0.1326, 0.0206, 0.0397, 0.0444, 0.1840, 0.1011], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0162, 0.0184, 0.0160, 0.0200, 0.0208, 0.0187, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 23:54:23,689 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9564, 2.7890, 2.7938, 2.1361, 2.6588, 2.2515, 2.7339, 2.9425], device='cuda:2'), covar=tensor([0.0277, 0.0659, 0.0500, 0.1599, 0.0734, 0.0872, 0.0518, 0.0688], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0151, 0.0162, 0.0146, 0.0139, 0.0126, 0.0139, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 23:54:28,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6184, 3.7802, 2.0924, 4.3447, 2.8210, 4.2524, 2.4038, 2.9283], device='cuda:2'), covar=tensor([0.0273, 0.0354, 0.1789, 0.0154, 0.0809, 0.0410, 0.1522, 0.0809], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0168, 0.0189, 0.0139, 0.0168, 0.0207, 0.0197, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-29 23:54:38,731 INFO [train.py:904] (2/8) Epoch 14, batch 5700, loss[loss=0.2856, simple_loss=0.3426, pruned_loss=0.1143, over 11495.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.303, pruned_loss=0.07021, over 3093904.12 frames. ], batch size: 248, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:54:51,525 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 23:55:45,156 INFO [optim.py:368] (2/8) 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 23:55:59,932 INFO [train.py:904] (2/8) Epoch 14, batch 5750, loss[loss=0.2288, simple_loss=0.3248, pruned_loss=0.06645, over 15345.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3066, pruned_loss=0.07207, over 3086041.52 frames. ], batch size: 191, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:56:07,280 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 23:56:24,984 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7499, 1.3543, 1.6685, 1.6364, 1.7332, 1.8640, 1.5799, 1.7210], device='cuda:2'), covar=tensor([0.0203, 0.0307, 0.0169, 0.0216, 0.0228, 0.0144, 0.0335, 0.0103], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0175, 0.0160, 0.0167, 0.0175, 0.0132, 0.0177, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:57:21,150 INFO [train.py:904] (2/8) Epoch 14, batch 5800, loss[loss=0.2298, simple_loss=0.3255, pruned_loss=0.06702, over 16677.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3064, pruned_loss=0.07144, over 3059468.86 frames. ], batch size: 134, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:57:34,936 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 23:57:56,996 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0271, 2.7102, 2.7481, 2.0454, 2.6173, 2.1562, 2.7989, 2.8875], device='cuda:2'), covar=tensor([0.0323, 0.0736, 0.0563, 0.1708, 0.0835, 0.0922, 0.0636, 0.0699], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0149, 0.0161, 0.0145, 0.0138, 0.0126, 0.0138, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-29 23:58:15,805 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1731, 4.2358, 4.0482, 3.8018, 3.7697, 4.1523, 3.8752, 3.8990], device='cuda:2'), covar=tensor([0.0581, 0.0526, 0.0298, 0.0292, 0.0844, 0.0446, 0.0752, 0.0667], device='cuda:2'), in_proj_covar=tensor([0.0254, 0.0348, 0.0302, 0.0284, 0.0321, 0.0332, 0.0206, 0.0357], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-29 23:58:26,264 INFO [optim.py:368] (2/8) 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 23:58:41,173 INFO [train.py:904] (2/8) Epoch 14, batch 5850, loss[loss=0.2353, simple_loss=0.3059, pruned_loss=0.08233, over 11662.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.304, pruned_loss=0.06982, over 3044161.82 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:00:03,509 INFO [train.py:904] (2/8) Epoch 14, batch 5900, loss[loss=0.2027, simple_loss=0.2937, pruned_loss=0.05584, over 17123.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3035, pruned_loss=0.06937, over 3055153.70 frames. ], batch size: 48, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:00:32,250 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 00:01:10,248 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:10,914 INFO [optim.py:368] (2/8) 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 00:01:15,009 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:24,299 INFO [train.py:904] (2/8) Epoch 14, batch 5950, loss[loss=0.212, simple_loss=0.3045, pruned_loss=0.05972, over 16867.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3044, pruned_loss=0.06816, over 3059587.41 frames. ], batch size: 90, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:08,522 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6814, 4.0587, 3.9561, 2.3553, 3.3740, 2.7049, 3.9860, 4.2806], device='cuda:2'), covar=tensor([0.0244, 0.0600, 0.0522, 0.1768, 0.0719, 0.0845, 0.0592, 0.0791], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0151, 0.0161, 0.0146, 0.0138, 0.0126, 0.0139, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:02:18,660 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6352, 2.5709, 2.3274, 3.8685, 2.8133, 3.9372, 1.4793, 2.8143], device='cuda:2'), covar=tensor([0.1371, 0.0763, 0.1224, 0.0160, 0.0258, 0.0390, 0.1548, 0.0806], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0164, 0.0186, 0.0161, 0.0203, 0.0210, 0.0189, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:02:40,371 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1008, 2.4044, 2.2588, 2.7450, 2.1027, 3.2580, 1.7806, 2.6656], device='cuda:2'), covar=tensor([0.0972, 0.0503, 0.0973, 0.0150, 0.0119, 0.0414, 0.1151, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0161, 0.0202, 0.0209, 0.0188, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:02:44,551 INFO [train.py:904] (2/8) Epoch 14, batch 6000, loss[loss=0.185, simple_loss=0.2691, pruned_loss=0.05044, over 16698.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3024, pruned_loss=0.06699, over 3082585.43 frames. ], batch size: 134, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:44,552 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 00:02:55,344 INFO [train.py:938] (2/8) Epoch 14, validation: loss=0.1574, simple_loss=0.2706, pruned_loss=0.0221, over 944034.00 frames. 2023-04-30 00:02:55,344 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 00:02:56,896 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:01,576 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:29,773 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 00:03:58,418 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4039, 4.0528, 4.0544, 2.8749, 3.6992, 4.1144, 3.8007, 2.1890], device='cuda:2'), covar=tensor([0.0467, 0.0032, 0.0043, 0.0303, 0.0076, 0.0097, 0.0056, 0.0402], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0072, 0.0073, 0.0128, 0.0086, 0.0096, 0.0083, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 00:03:59,081 INFO [optim.py:368] (2/8) 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 00:04:18,770 INFO [train.py:904] (2/8) Epoch 14, batch 6050, loss[loss=0.2224, simple_loss=0.2946, pruned_loss=0.07509, over 11469.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3011, pruned_loss=0.06623, over 3084612.87 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:05:40,154 INFO [train.py:904] (2/8) Epoch 14, batch 6100, loss[loss=0.1985, simple_loss=0.2826, pruned_loss=0.05717, over 15493.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2999, pruned_loss=0.0647, over 3101518.46 frames. ], batch size: 191, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:06:45,097 INFO [optim.py:368] (2/8) 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 00:06:59,047 INFO [train.py:904] (2/8) Epoch 14, batch 6150, loss[loss=0.2291, simple_loss=0.3059, pruned_loss=0.07618, over 11733.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2987, pruned_loss=0.06461, over 3104665.21 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:07:25,675 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 00:08:16,180 INFO [train.py:904] (2/8) Epoch 14, batch 6200, loss[loss=0.2054, simple_loss=0.2769, pruned_loss=0.06698, over 11302.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2973, pruned_loss=0.0642, over 3106536.80 frames. ], batch size: 248, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:09:17,889 INFO [optim.py:368] (2/8) 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 00:09:32,460 INFO [train.py:904] (2/8) Epoch 14, batch 6250, loss[loss=0.2619, simple_loss=0.3232, pruned_loss=0.1003, over 11734.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2965, pruned_loss=0.06416, over 3097503.88 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:10:42,267 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:10:46,667 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 14, batch 6300, loss[loss=0.2259, simple_loss=0.2964, pruned_loss=0.07767, over 11513.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2962, pruned_loss=0.06354, over 3098378.04 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:11:17,759 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:11:52,435 INFO [optim.py:368] (2/8) 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 00:12:05,889 INFO [train.py:904] (2/8) Epoch 14, batch 6350, loss[loss=0.1996, simple_loss=0.2815, pruned_loss=0.05881, over 16468.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2976, pruned_loss=0.06519, over 3080622.49 frames. ], batch size: 62, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:12:06,479 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7885, 3.0161, 2.3038, 4.3033, 3.3224, 4.0184, 1.5696, 2.8472], device='cuda:2'), covar=tensor([0.1178, 0.0581, 0.1258, 0.0157, 0.0247, 0.0457, 0.1428, 0.0820], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0164, 0.0186, 0.0163, 0.0203, 0.0211, 0.0188, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:12:36,892 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:12:40,973 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-30 00:12:50,034 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:13:22,144 INFO [train.py:904] (2/8) Epoch 14, batch 6400, loss[loss=0.1972, simple_loss=0.2819, pruned_loss=0.05628, over 16906.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2977, pruned_loss=0.06637, over 3077920.15 frames. ], batch size: 116, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:14:09,292 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:14:23,618 INFO [optim.py:368] (2/8) 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 00:14:37,173 INFO [train.py:904] (2/8) Epoch 14, batch 6450, loss[loss=0.2178, simple_loss=0.2988, pruned_loss=0.06846, over 15246.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2977, pruned_loss=0.06538, over 3084281.53 frames. ], batch size: 190, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:15:54,740 INFO [train.py:904] (2/8) Epoch 14, batch 6500, loss[loss=0.2216, simple_loss=0.3021, pruned_loss=0.07051, over 16365.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2958, pruned_loss=0.06497, over 3070691.26 frames. ], batch size: 146, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:16:59,451 INFO [optim.py:368] (2/8) 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 00:17:12,086 INFO [train.py:904] (2/8) Epoch 14, batch 6550, loss[loss=0.2197, simple_loss=0.3144, pruned_loss=0.06253, over 16661.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.298, pruned_loss=0.06587, over 3063577.88 frames. ], batch size: 134, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:18:22,257 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:25,774 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:26,459 INFO [train.py:904] (2/8) Epoch 14, batch 6600, loss[loss=0.2291, simple_loss=0.313, pruned_loss=0.07256, over 15328.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3006, pruned_loss=0.06657, over 3073781.12 frames. ], batch size: 191, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:19:29,570 INFO [optim.py:368] (2/8) 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 00:19:32,844 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6702, 3.9083, 4.1847, 4.1058, 4.1364, 3.8714, 3.5923, 3.8825], device='cuda:2'), covar=tensor([0.0596, 0.0740, 0.0540, 0.0717, 0.0742, 0.0662, 0.1621, 0.0622], device='cuda:2'), in_proj_covar=tensor([0.0356, 0.0379, 0.0376, 0.0359, 0.0423, 0.0403, 0.0496, 0.0318], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 00:19:33,986 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:38,206 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138599.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:42,709 INFO [train.py:904] (2/8) Epoch 14, batch 6650, loss[loss=0.2649, simple_loss=0.3236, pruned_loss=0.1031, over 11319.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3009, pruned_loss=0.06752, over 3058913.25 frames. ], batch size: 247, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:20:19,651 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:20:23,814 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 00:20:58,166 INFO [train.py:904] (2/8) Epoch 14, batch 6700, loss[loss=0.2706, simple_loss=0.3343, pruned_loss=0.1034, over 11505.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2997, pruned_loss=0.06737, over 3063634.37 frames. ], batch size: 248, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:21:39,960 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:44,862 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:53,053 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:03,781 INFO [optim.py:368] (2/8) 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 00:22:09,207 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:15,571 INFO [train.py:904] (2/8) Epoch 14, batch 6750, loss[loss=0.1907, simple_loss=0.278, pruned_loss=0.05171, over 16755.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2986, pruned_loss=0.06716, over 3078196.42 frames. ], batch size: 124, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:23:14,934 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:22,406 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:29,043 INFO [train.py:904] (2/8) Epoch 14, batch 6800, loss[loss=0.1983, simple_loss=0.288, pruned_loss=0.05434, over 17299.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2987, pruned_loss=0.06706, over 3085462.44 frames. ], batch size: 52, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:23:38,436 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:47,374 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9574, 2.6249, 2.6149, 1.8990, 2.4201, 2.6792, 2.5411, 1.8786], device='cuda:2'), covar=tensor([0.0376, 0.0067, 0.0068, 0.0308, 0.0129, 0.0116, 0.0096, 0.0376], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0073, 0.0073, 0.0129, 0.0086, 0.0096, 0.0083, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 00:24:34,694 INFO [optim.py:368] (2/8) 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 00:24:43,022 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1089, 5.1183, 4.9018, 4.4923, 4.5118, 5.0075, 4.9825, 4.6447], device='cuda:2'), covar=tensor([0.0871, 0.1106, 0.0404, 0.0484, 0.1267, 0.0630, 0.0477, 0.1459], device='cuda:2'), in_proj_covar=tensor([0.0251, 0.0347, 0.0303, 0.0281, 0.0319, 0.0329, 0.0206, 0.0358], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 00:24:46,910 INFO [train.py:904] (2/8) Epoch 14, batch 6850, loss[loss=0.2144, simple_loss=0.311, pruned_loss=0.05884, over 16468.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.299, pruned_loss=0.06698, over 3106158.70 frames. ], batch size: 146, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:25:00,750 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9302, 3.9778, 4.3193, 4.2776, 4.2749, 4.0262, 3.9790, 4.0030], device='cuda:2'), covar=tensor([0.0373, 0.0638, 0.0429, 0.0436, 0.0461, 0.0480, 0.1049, 0.0509], device='cuda:2'), in_proj_covar=tensor([0.0352, 0.0373, 0.0372, 0.0355, 0.0419, 0.0398, 0.0489, 0.0316], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 00:26:02,477 INFO [train.py:904] (2/8) Epoch 14, batch 6900, loss[loss=0.2308, simple_loss=0.3169, pruned_loss=0.07232, over 16347.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3018, pruned_loss=0.06653, over 3113025.19 frames. ], batch size: 146, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:10,346 INFO [optim.py:368] (2/8) 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 00:27:22,849 INFO [train.py:904] (2/8) Epoch 14, batch 6950, loss[loss=0.2748, simple_loss=0.3352, pruned_loss=0.1072, over 11201.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3042, pruned_loss=0.0686, over 3097638.17 frames. ], batch size: 246, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:33,764 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:00,006 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:38,216 INFO [train.py:904] (2/8) Epoch 14, batch 7000, loss[loss=0.2097, simple_loss=0.2982, pruned_loss=0.06057, over 16641.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3039, pruned_loss=0.06746, over 3091014.66 frames. ], batch size: 57, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:29:06,850 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:29:12,852 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:18,594 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:42,505 INFO [optim.py:368] (2/8) 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 00:29:55,150 INFO [train.py:904] (2/8) Epoch 14, batch 7050, loss[loss=0.2153, simple_loss=0.31, pruned_loss=0.06031, over 17023.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3041, pruned_loss=0.06666, over 3110099.92 frames. ], batch size: 50, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:30:33,410 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:42,641 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3838, 1.5948, 1.9237, 2.2271, 2.4118, 2.5747, 1.6815, 2.4407], device='cuda:2'), covar=tensor([0.0156, 0.0410, 0.0254, 0.0272, 0.0224, 0.0158, 0.0417, 0.0122], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0173, 0.0159, 0.0164, 0.0172, 0.0131, 0.0176, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 00:30:49,916 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139042.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:31:13,407 INFO [train.py:904] (2/8) Epoch 14, batch 7100, loss[loss=0.1996, simple_loss=0.2874, pruned_loss=0.05595, over 16405.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.302, pruned_loss=0.06595, over 3116489.90 frames. ], batch size: 146, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:31:15,010 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:32:18,348 INFO [optim.py:368] (2/8) 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 00:32:31,526 INFO [train.py:904] (2/8) Epoch 14, batch 7150, loss[loss=0.2091, simple_loss=0.2944, pruned_loss=0.06197, over 16313.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2999, pruned_loss=0.06558, over 3112247.05 frames. ], batch size: 165, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:33:47,876 INFO [train.py:904] (2/8) Epoch 14, batch 7200, loss[loss=0.1886, simple_loss=0.2783, pruned_loss=0.04941, over 16388.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2981, pruned_loss=0.06493, over 3089871.34 frames. ], batch size: 146, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:33:48,892 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 00:34:55,353 INFO [optim.py:368] (2/8) 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 00:35:07,113 INFO [train.py:904] (2/8) Epoch 14, batch 7250, loss[loss=0.1955, simple_loss=0.2743, pruned_loss=0.05831, over 16855.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2955, pruned_loss=0.06346, over 3093054.37 frames. ], batch size: 116, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:36:01,917 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:36:22,182 INFO [train.py:904] (2/8) Epoch 14, batch 7300, loss[loss=0.221, simple_loss=0.3064, pruned_loss=0.06786, over 16845.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.295, pruned_loss=0.06343, over 3098916.55 frames. ], batch size: 116, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:36:42,738 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:37:00,525 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:19,419 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 00:37:20,382 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:29,627 INFO [optim.py:368] (2/8) 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 00:37:37,012 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:40,532 INFO [train.py:904] (2/8) Epoch 14, batch 7350, loss[loss=0.3013, simple_loss=0.3441, pruned_loss=0.1292, over 11370.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2965, pruned_loss=0.06495, over 3065704.41 frames. ], batch size: 248, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:37:57,871 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1918, 4.2336, 4.6416, 4.6166, 4.6125, 4.2732, 4.3089, 4.1636], device='cuda:2'), covar=tensor([0.0339, 0.0544, 0.0375, 0.0430, 0.0495, 0.0431, 0.0957, 0.0542], device='cuda:2'), in_proj_covar=tensor([0.0352, 0.0371, 0.0370, 0.0355, 0.0418, 0.0397, 0.0484, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 00:38:36,753 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:36,861 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:38:44,971 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:56,614 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:59,353 INFO [train.py:904] (2/8) Epoch 14, batch 7400, loss[loss=0.2188, simple_loss=0.3066, pruned_loss=0.0655, over 16194.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2971, pruned_loss=0.06502, over 3079963.60 frames. ], batch size: 165, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:39:01,642 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:39:01,841 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9692, 2.7149, 2.7692, 2.1601, 2.6636, 2.1691, 2.7467, 2.8890], device='cuda:2'), covar=tensor([0.0258, 0.0746, 0.0521, 0.1710, 0.0741, 0.0872, 0.0536, 0.0626], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0139, 0.0127, 0.0139, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:39:31,315 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4658, 2.9575, 2.9381, 1.9486, 2.6836, 2.0581, 3.0525, 3.1294], device='cuda:2'), covar=tensor([0.0278, 0.0663, 0.0617, 0.1833, 0.0804, 0.1005, 0.0608, 0.0688], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0139, 0.0127, 0.0139, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:39:52,860 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:00,699 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:08,330 INFO [optim.py:368] (2/8) 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 00:40:18,576 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:19,292 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 00:40:19,476 INFO [train.py:904] (2/8) Epoch 14, batch 7450, loss[loss=0.2013, simple_loss=0.2819, pruned_loss=0.0603, over 17047.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2984, pruned_loss=0.06658, over 3047138.17 frames. ], batch size: 55, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:41:43,393 INFO [train.py:904] (2/8) Epoch 14, batch 7500, loss[loss=0.2156, simple_loss=0.3045, pruned_loss=0.06336, over 15397.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.299, pruned_loss=0.06608, over 3041797.24 frames. ], batch size: 190, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:42:53,286 INFO [optim.py:368] (2/8) 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 00:43:02,945 INFO [train.py:904] (2/8) Epoch 14, batch 7550, loss[loss=0.2249, simple_loss=0.3069, pruned_loss=0.07142, over 15312.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2975, pruned_loss=0.06589, over 3047170.89 frames. ], batch size: 190, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:43:51,642 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0132, 2.4195, 2.3224, 2.8362, 2.1760, 3.2462, 1.7955, 2.7187], device='cuda:2'), covar=tensor([0.1089, 0.0542, 0.1039, 0.0158, 0.0144, 0.0370, 0.1325, 0.0641], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0162, 0.0203, 0.0209, 0.0187, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:43:53,190 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 00:44:19,041 INFO [train.py:904] (2/8) Epoch 14, batch 7600, loss[loss=0.2669, simple_loss=0.3299, pruned_loss=0.102, over 11589.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2969, pruned_loss=0.06589, over 3046459.81 frames. ], batch size: 247, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:44:39,696 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:44:49,003 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:06,338 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:19,199 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 00:45:23,447 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139594.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:24,242 INFO [optim.py:368] (2/8) 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 00:45:29,396 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-04-30 00:45:34,160 INFO [train.py:904] (2/8) Epoch 14, batch 7650, loss[loss=0.1988, simple_loss=0.2887, pruned_loss=0.05444, over 17175.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.297, pruned_loss=0.06619, over 3046964.75 frames. ], batch size: 46, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:45:50,719 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:19,134 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:46:19,290 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:36,534 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:39,474 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:49,449 INFO [train.py:904] (2/8) Epoch 14, batch 7700, loss[loss=0.2016, simple_loss=0.2883, pruned_loss=0.05747, over 16424.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2974, pruned_loss=0.06691, over 3036275.17 frames. ], batch size: 35, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:47:47,198 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2250, 1.5112, 1.8786, 2.1463, 2.2923, 2.4675, 1.6421, 2.3440], device='cuda:2'), covar=tensor([0.0189, 0.0451, 0.0260, 0.0268, 0.0250, 0.0170, 0.0433, 0.0122], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0176, 0.0159, 0.0165, 0.0174, 0.0131, 0.0177, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 00:47:52,621 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8884, 4.1322, 3.9804, 3.9904, 3.6841, 3.8346, 3.8419, 4.1324], device='cuda:2'), covar=tensor([0.1131, 0.0871, 0.0981, 0.0813, 0.0775, 0.1404, 0.0894, 0.0993], device='cuda:2'), in_proj_covar=tensor([0.0582, 0.0717, 0.0592, 0.0516, 0.0452, 0.0467, 0.0600, 0.0549], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 00:47:57,377 INFO [optim.py:368] (2/8) 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 00:48:06,798 INFO [train.py:904] (2/8) Epoch 14, batch 7750, loss[loss=0.2458, simple_loss=0.3046, pruned_loss=0.09348, over 11273.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2973, pruned_loss=0.06648, over 3046549.81 frames. ], batch size: 246, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:24,608 INFO [train.py:904] (2/8) Epoch 14, batch 7800, loss[loss=0.2114, simple_loss=0.2921, pruned_loss=0.06529, over 16805.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2987, pruned_loss=0.06738, over 3061549.68 frames. ], batch size: 124, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:51,624 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:22,441 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:32,559 INFO [optim.py:368] (2/8) 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 00:50:41,154 INFO [train.py:904] (2/8) Epoch 14, batch 7850, loss[loss=0.2145, simple_loss=0.302, pruned_loss=0.06351, over 16807.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2989, pruned_loss=0.06641, over 3082942.97 frames. ], batch size: 83, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:50:43,902 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9301, 5.2851, 5.5173, 5.2419, 5.2333, 5.8114, 5.2847, 5.0666], device='cuda:2'), covar=tensor([0.0929, 0.1753, 0.2246, 0.1860, 0.2602, 0.1059, 0.1475, 0.2397], device='cuda:2'), in_proj_covar=tensor([0.0374, 0.0523, 0.0578, 0.0446, 0.0599, 0.0600, 0.0453, 0.0607], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 00:50:45,362 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:48,503 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:24,569 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,618 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,632 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:56,191 INFO [train.py:904] (2/8) Epoch 14, batch 7900, loss[loss=0.2535, simple_loss=0.3157, pruned_loss=0.09564, over 11605.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2977, pruned_loss=0.06545, over 3101455.76 frames. ], batch size: 248, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:52:00,361 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-04-30 00:52:15,857 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:18,368 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:33,956 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-30 00:53:01,522 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:03,657 INFO [optim.py:368] (2/8) 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 00:53:12,903 INFO [train.py:904] (2/8) Epoch 14, batch 7950, loss[loss=0.2051, simple_loss=0.2839, pruned_loss=0.06318, over 16782.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2986, pruned_loss=0.06635, over 3079730.70 frames. ], batch size: 39, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:53:27,435 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:51,346 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:59,015 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:54:06,334 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 00:54:08,409 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:18,372 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:28,354 INFO [train.py:904] (2/8) Epoch 14, batch 8000, loss[loss=0.262, simple_loss=0.3206, pruned_loss=0.1017, over 11296.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2996, pruned_loss=0.06697, over 3083513.50 frames. ], batch size: 246, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:55:10,467 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139980.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:30,335 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:34,108 INFO [optim.py:368] (2/8) 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 00:55:46,025 INFO [train.py:904] (2/8) Epoch 14, batch 8050, loss[loss=0.2271, simple_loss=0.3009, pruned_loss=0.07664, over 11962.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3, pruned_loss=0.0674, over 3069070.04 frames. ], batch size: 247, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:56:32,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9983, 3.0681, 1.7390, 3.2125, 2.2023, 3.2827, 1.9712, 2.5071], device='cuda:2'), covar=tensor([0.0270, 0.0403, 0.1620, 0.0209, 0.0850, 0.0560, 0.1410, 0.0717], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0169, 0.0190, 0.0139, 0.0167, 0.0209, 0.0197, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 00:56:33,243 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 00:56:59,039 INFO [train.py:904] (2/8) Epoch 14, batch 8100, loss[loss=0.2144, simple_loss=0.2955, pruned_loss=0.06665, over 16883.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2999, pruned_loss=0.06726, over 3074650.29 frames. ], batch size: 116, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:57:08,150 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-30 00:58:06,569 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 14, batch 8150, loss[loss=0.2009, simple_loss=0.273, pruned_loss=0.06441, over 11396.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2972, pruned_loss=0.06609, over 3073534.26 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:58:18,967 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:29,415 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:52,130 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:18,939 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9961, 2.9521, 2.3788, 2.8837, 3.4002, 3.0631, 3.7679, 3.6622], device='cuda:2'), covar=tensor([0.0062, 0.0311, 0.0445, 0.0305, 0.0190, 0.0261, 0.0171, 0.0178], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0212, 0.0207, 0.0208, 0.0212, 0.0212, 0.0214, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 00:59:23,722 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:34,234 INFO [train.py:904] (2/8) Epoch 14, batch 8200, loss[loss=0.2442, simple_loss=0.313, pruned_loss=0.08775, over 11529.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.295, pruned_loss=0.06584, over 3056879.97 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:59:47,417 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:50,117 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:53,802 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:05,675 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:44,842 INFO [optim.py:368] (2/8) 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 01:00:55,642 INFO [train.py:904] (2/8) Epoch 14, batch 8250, loss[loss=0.1816, simple_loss=0.2688, pruned_loss=0.0472, over 12095.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2936, pruned_loss=0.06297, over 3058873.85 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:01:02,324 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:37,220 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:55,129 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:16,387 INFO [train.py:904] (2/8) Epoch 14, batch 8300, loss[loss=0.1978, simple_loss=0.2748, pruned_loss=0.06043, over 12158.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2903, pruned_loss=0.05968, over 3046555.05 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:02:28,957 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140259.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:54,963 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:12,771 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:27,884 INFO [optim.py:368] (2/8) 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 01:03:36,685 INFO [train.py:904] (2/8) Epoch 14, batch 8350, loss[loss=0.2068, simple_loss=0.2869, pruned_loss=0.06338, over 11882.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2889, pruned_loss=0.05764, over 3040105.49 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:03:42,736 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7423, 4.0839, 4.5081, 2.2975, 3.5369, 2.8253, 4.2742, 4.2350], device='cuda:2'), covar=tensor([0.0203, 0.0637, 0.0310, 0.1877, 0.0648, 0.0843, 0.0481, 0.0798], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0148, 0.0157, 0.0144, 0.0136, 0.0123, 0.0136, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 01:04:07,952 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:04:09,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1457, 2.0704, 2.2142, 3.5668, 2.0599, 2.3543, 2.2105, 2.2152], device='cuda:2'), covar=tensor([0.1071, 0.3433, 0.2620, 0.0535, 0.4356, 0.2471, 0.3486, 0.3560], device='cuda:2'), in_proj_covar=tensor([0.0364, 0.0402, 0.0335, 0.0310, 0.0412, 0.0460, 0.0367, 0.0468], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:04:39,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6815, 1.6853, 2.1602, 2.5419, 2.5044, 2.7941, 1.7947, 2.8626], device='cuda:2'), covar=tensor([0.0152, 0.0453, 0.0281, 0.0245, 0.0251, 0.0196, 0.0439, 0.0134], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0174, 0.0157, 0.0162, 0.0172, 0.0130, 0.0176, 0.0119], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 01:04:57,945 INFO [train.py:904] (2/8) Epoch 14, batch 8400, loss[loss=0.1867, simple_loss=0.2821, pruned_loss=0.04567, over 16830.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.286, pruned_loss=0.05517, over 3046613.49 frames. ], batch size: 116, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:05:25,717 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.63 vs. limit=5.0 2023-04-30 01:05:54,562 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:06:08,144 INFO [optim.py:368] (2/8) 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 01:06:18,701 INFO [train.py:904] (2/8) Epoch 14, batch 8450, loss[loss=0.1675, simple_loss=0.2627, pruned_loss=0.03612, over 15284.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2837, pruned_loss=0.0529, over 3060048.42 frames. ], batch size: 190, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:06:55,919 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:27,249 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:32,364 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:37,955 INFO [train.py:904] (2/8) Epoch 14, batch 8500, loss[loss=0.1802, simple_loss=0.2631, pruned_loss=0.04863, over 15308.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2799, pruned_loss=0.05062, over 3045025.58 frames. ], batch size: 190, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:07:50,011 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:51,373 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:54,723 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:56,326 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 01:08:01,376 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:13,347 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:20,300 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9444, 1.7513, 1.5762, 1.3958, 1.8835, 1.5198, 1.6638, 1.9341], device='cuda:2'), covar=tensor([0.0160, 0.0259, 0.0362, 0.0314, 0.0188, 0.0247, 0.0170, 0.0181], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0207, 0.0201, 0.0202, 0.0206, 0.0207, 0.0208, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:08:38,542 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140489.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:46,067 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:51,328 INFO [optim.py:368] (2/8) 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 01:09:01,494 INFO [train.py:904] (2/8) Epoch 14, batch 8550, loss[loss=0.1888, simple_loss=0.2653, pruned_loss=0.05612, over 12172.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2783, pruned_loss=0.05002, over 3037203.40 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:09:11,968 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:15,278 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:17,241 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:18,513 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:39,009 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:41,516 INFO [train.py:904] (2/8) Epoch 14, batch 8600, loss[loss=0.2144, simple_loss=0.3033, pruned_loss=0.06279, over 16402.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2787, pruned_loss=0.04921, over 3034872.79 frames. ], batch size: 146, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:10:46,487 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:19,627 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:33,306 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8995, 5.4046, 5.5379, 5.3196, 5.3712, 5.8739, 5.4012, 5.0906], device='cuda:2'), covar=tensor([0.0843, 0.1705, 0.1946, 0.1725, 0.2288, 0.0832, 0.1395, 0.2329], device='cuda:2'), in_proj_covar=tensor([0.0362, 0.0507, 0.0559, 0.0430, 0.0577, 0.0586, 0.0443, 0.0583], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 01:12:04,270 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4342, 3.5119, 2.1162, 3.9263, 2.4988, 3.8656, 2.1201, 2.7143], device='cuda:2'), covar=tensor([0.0268, 0.0365, 0.1562, 0.0169, 0.0908, 0.0500, 0.1588, 0.0829], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0163, 0.0184, 0.0134, 0.0163, 0.0201, 0.0192, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 01:12:06,767 INFO [optim.py:368] (2/8) 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 01:12:18,452 INFO [train.py:904] (2/8) Epoch 14, batch 8650, loss[loss=0.1733, simple_loss=0.2721, pruned_loss=0.03725, over 16210.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2763, pruned_loss=0.04728, over 3055839.80 frames. ], batch size: 165, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:12:52,300 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:14:05,319 INFO [train.py:904] (2/8) Epoch 14, batch 8700, loss[loss=0.1834, simple_loss=0.2685, pruned_loss=0.04915, over 12622.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2732, pruned_loss=0.04574, over 3069839.04 frames. ], batch size: 250, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:14:30,943 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7022, 4.9226, 5.0318, 4.8686, 4.9358, 5.4274, 4.9010, 4.5775], device='cuda:2'), covar=tensor([0.0865, 0.1751, 0.1906, 0.1849, 0.2254, 0.0838, 0.1352, 0.2296], device='cuda:2'), in_proj_covar=tensor([0.0357, 0.0500, 0.0550, 0.0424, 0.0569, 0.0578, 0.0437, 0.0576], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 01:14:42,303 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 01:15:24,088 INFO [optim.py:368] (2/8) 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 01:15:25,826 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 01:15:38,747 INFO [train.py:904] (2/8) Epoch 14, batch 8750, loss[loss=0.1932, simple_loss=0.2892, pruned_loss=0.04857, over 16890.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2729, pruned_loss=0.0451, over 3068461.48 frames. ], batch size: 116, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:16:25,374 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 01:16:30,724 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:09,174 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4013, 3.2759, 3.2902, 3.5076, 3.5359, 3.3157, 3.4621, 3.5630], device='cuda:2'), covar=tensor([0.1312, 0.1107, 0.1505, 0.0858, 0.0850, 0.2985, 0.1256, 0.1014], device='cuda:2'), in_proj_covar=tensor([0.0536, 0.0658, 0.0784, 0.0681, 0.0514, 0.0531, 0.0535, 0.0630], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:17:11,077 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:21,843 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 01:17:23,171 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3934, 4.5613, 4.6479, 4.5064, 4.5450, 5.0309, 4.5665, 4.2988], device='cuda:2'), covar=tensor([0.1255, 0.1588, 0.1641, 0.1709, 0.2280, 0.0868, 0.1416, 0.2354], device='cuda:2'), in_proj_covar=tensor([0.0356, 0.0498, 0.0548, 0.0423, 0.0569, 0.0579, 0.0437, 0.0575], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 01:17:28,574 INFO [train.py:904] (2/8) Epoch 14, batch 8800, loss[loss=0.1944, simple_loss=0.2844, pruned_loss=0.05222, over 16200.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2711, pruned_loss=0.04379, over 3062572.94 frames. ], batch size: 165, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:17:37,252 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1888, 4.4138, 4.4529, 4.2778, 4.3337, 4.8295, 4.3954, 4.1757], device='cuda:2'), covar=tensor([0.1598, 0.1542, 0.1641, 0.1932, 0.2477, 0.0948, 0.1452, 0.2464], device='cuda:2'), in_proj_covar=tensor([0.0357, 0.0499, 0.0549, 0.0424, 0.0570, 0.0580, 0.0438, 0.0575], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 01:17:43,128 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:56,978 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:18:11,485 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1636, 3.2957, 1.5663, 3.5557, 2.2031, 3.4687, 1.6109, 2.4512], device='cuda:2'), covar=tensor([0.0316, 0.0347, 0.2043, 0.0236, 0.1019, 0.0520, 0.2029, 0.0836], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0160, 0.0183, 0.0132, 0.0162, 0.0198, 0.0190, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 01:18:14,997 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-30 01:18:35,737 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:00,060 INFO [optim.py:368] (2/8) 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 01:19:09,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7196, 2.3151, 2.1997, 4.5868, 2.2412, 2.8310, 2.3444, 2.5051], device='cuda:2'), covar=tensor([0.0894, 0.3424, 0.2615, 0.0326, 0.3925, 0.2219, 0.3246, 0.3134], device='cuda:2'), in_proj_covar=tensor([0.0364, 0.0404, 0.0337, 0.0312, 0.0413, 0.0460, 0.0368, 0.0468], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:19:12,350 INFO [train.py:904] (2/8) Epoch 14, batch 8850, loss[loss=0.1764, simple_loss=0.2834, pruned_loss=0.03469, over 16892.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2739, pruned_loss=0.04338, over 3059216.21 frames. ], batch size: 102, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:19:23,456 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:27,227 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:37,089 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:50,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4908, 3.5780, 2.1498, 4.0121, 2.5100, 3.9364, 2.0294, 2.7251], device='cuda:2'), covar=tensor([0.0260, 0.0299, 0.1623, 0.0156, 0.0978, 0.0444, 0.1682, 0.0818], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0160, 0.0183, 0.0132, 0.0162, 0.0198, 0.0190, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 01:20:43,590 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:20:58,215 INFO [train.py:904] (2/8) Epoch 14, batch 8900, loss[loss=0.1749, simple_loss=0.2617, pruned_loss=0.04408, over 12591.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2743, pruned_loss=0.04283, over 3058305.26 frames. ], batch size: 247, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:21:19,821 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3100, 4.3547, 4.1731, 3.8880, 3.8832, 4.2669, 4.0154, 3.9765], device='cuda:2'), covar=tensor([0.0555, 0.0552, 0.0309, 0.0292, 0.0845, 0.0489, 0.0567, 0.0639], device='cuda:2'), in_proj_covar=tensor([0.0245, 0.0334, 0.0295, 0.0271, 0.0303, 0.0319, 0.0202, 0.0344], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:21:25,581 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:21:34,667 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:22:37,199 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7804, 2.9105, 2.4476, 4.4548, 3.0885, 4.2799, 1.4838, 3.2174], device='cuda:2'), covar=tensor([0.1348, 0.0661, 0.1211, 0.0127, 0.0147, 0.0294, 0.1614, 0.0638], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0160, 0.0181, 0.0156, 0.0193, 0.0205, 0.0185, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-30 01:22:46,890 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 14, batch 8950, loss[loss=0.1655, simple_loss=0.2594, pruned_loss=0.03579, over 16672.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2737, pruned_loss=0.04281, over 3090724.78 frames. ], batch size: 134, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:23:29,071 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:23:32,803 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:38,381 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:48,129 INFO [train.py:904] (2/8) Epoch 14, batch 9000, loss[loss=0.1817, simple_loss=0.26, pruned_loss=0.05173, over 11775.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2696, pruned_loss=0.04104, over 3092707.57 frames. ], batch size: 246, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:24:48,129 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 01:24:55,472 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4825, 5.4214, 5.3067, 4.9027, 4.9547, 5.2918, 5.4862, 5.0479], device='cuda:2'), covar=tensor([0.0438, 0.0354, 0.0235, 0.0277, 0.0847, 0.0354, 0.0128, 0.0582], device='cuda:2'), in_proj_covar=tensor([0.0245, 0.0335, 0.0295, 0.0272, 0.0303, 0.0320, 0.0201, 0.0343], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:24:58,089 INFO [train.py:938] (2/8) Epoch 14, validation: loss=0.1514, simple_loss=0.2553, pruned_loss=0.0237, over 944034.00 frames. 2023-04-30 01:24:58,090 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 01:25:21,239 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:25:52,461 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:26:29,790 INFO [optim.py:368] (2/8) 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 01:26:39,921 INFO [train.py:904] (2/8) Epoch 14, batch 9050, loss[loss=0.1862, simple_loss=0.2701, pruned_loss=0.05113, over 16349.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2704, pruned_loss=0.04144, over 3092736.10 frames. ], batch size: 146, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:26:54,583 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:27:34,511 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 01:28:07,500 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:20,839 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3157, 3.3587, 2.0592, 3.6596, 2.3851, 3.5767, 2.0793, 2.7150], device='cuda:2'), covar=tensor([0.0254, 0.0310, 0.1423, 0.0162, 0.0851, 0.0478, 0.1525, 0.0694], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0160, 0.0183, 0.0132, 0.0163, 0.0198, 0.0191, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 01:28:25,839 INFO [train.py:904] (2/8) Epoch 14, batch 9100, loss[loss=0.1877, simple_loss=0.2877, pruned_loss=0.04381, over 16209.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2703, pruned_loss=0.04189, over 3083409.71 frames. ], batch size: 165, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:28:53,087 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8310, 3.7000, 3.8687, 3.9821, 4.0741, 3.6673, 4.0413, 4.0973], device='cuda:2'), covar=tensor([0.1330, 0.0976, 0.1200, 0.0594, 0.0555, 0.1677, 0.0577, 0.0572], device='cuda:2'), in_proj_covar=tensor([0.0531, 0.0653, 0.0772, 0.0673, 0.0509, 0.0527, 0.0529, 0.0627], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:29:30,825 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:29:58,703 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:30:13,177 INFO [optim.py:368] (2/8) 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 01:30:22,842 INFO [train.py:904] (2/8) Epoch 14, batch 9150, loss[loss=0.1789, simple_loss=0.2733, pruned_loss=0.04229, over 15490.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2715, pruned_loss=0.04218, over 3079999.45 frames. ], batch size: 192, lr: 4.80e-03, grad_scale: 4.0 2023-04-30 01:30:33,714 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 01:31:01,329 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5667, 5.8878, 5.6211, 5.7169, 5.3568, 5.2712, 5.2708, 6.0011], device='cuda:2'), covar=tensor([0.1162, 0.0783, 0.0841, 0.0655, 0.0701, 0.0512, 0.0978, 0.0766], device='cuda:2'), in_proj_covar=tensor([0.0565, 0.0695, 0.0570, 0.0502, 0.0441, 0.0454, 0.0583, 0.0530], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:31:51,270 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:31:54,437 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:04,830 INFO [train.py:904] (2/8) Epoch 14, batch 9200, loss[loss=0.1886, simple_loss=0.2945, pruned_loss=0.04129, over 16210.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2675, pruned_loss=0.04106, over 3086462.70 frames. ], batch size: 165, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:32:15,279 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-30 01:32:28,726 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:28,762 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:31,525 INFO [optim.py:368] (2/8) 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 01:33:40,990 INFO [train.py:904] (2/8) Epoch 14, batch 9250, loss[loss=0.1492, simple_loss=0.2347, pruned_loss=0.03188, over 12331.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2673, pruned_loss=0.04125, over 3084583.45 frames. ], batch size: 248, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:33:46,067 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141204.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:34:03,406 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:34:34,520 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 01:34:45,864 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 01:34:56,821 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 01:35:30,115 INFO [train.py:904] (2/8) Epoch 14, batch 9300, loss[loss=0.1636, simple_loss=0.2589, pruned_loss=0.03412, over 15533.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2657, pruned_loss=0.04052, over 3069229.47 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:36:11,697 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2846, 3.9442, 3.8511, 2.6208, 3.4891, 3.9032, 3.6566, 2.0941], device='cuda:2'), covar=tensor([0.0473, 0.0033, 0.0038, 0.0358, 0.0085, 0.0070, 0.0057, 0.0452], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0069, 0.0071, 0.0127, 0.0083, 0.0093, 0.0082, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 01:36:12,033 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-30 01:36:20,229 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:37:05,690 INFO [optim.py:368] (2/8) 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 01:37:14,522 INFO [train.py:904] (2/8) Epoch 14, batch 9350, loss[loss=0.1912, simple_loss=0.2834, pruned_loss=0.0495, over 16348.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2658, pruned_loss=0.04076, over 3064529.02 frames. ], batch size: 146, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:37:16,938 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:37:48,519 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4450, 1.9997, 1.7474, 1.7417, 2.2656, 1.9297, 2.0514, 2.3671], device='cuda:2'), covar=tensor([0.0114, 0.0362, 0.0422, 0.0390, 0.0222, 0.0339, 0.0154, 0.0191], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0211, 0.0204, 0.0207, 0.0210, 0.0211, 0.0209, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:38:08,729 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:54,830 INFO [train.py:904] (2/8) Epoch 14, batch 9400, loss[loss=0.1977, simple_loss=0.2989, pruned_loss=0.04825, over 16892.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2658, pruned_loss=0.04061, over 3065298.85 frames. ], batch size: 116, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:39:49,382 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:12,146 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:25,322 INFO [optim.py:368] (2/8) 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 01:40:32,807 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7837, 2.7023, 2.6893, 4.2435, 2.9151, 4.2088, 1.4245, 3.2019], device='cuda:2'), covar=tensor([0.1337, 0.0726, 0.1013, 0.0150, 0.0120, 0.0299, 0.1577, 0.0600], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0160, 0.0181, 0.0156, 0.0190, 0.0205, 0.0186, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-30 01:40:33,410 INFO [train.py:904] (2/8) Epoch 14, batch 9450, loss[loss=0.1556, simple_loss=0.2469, pruned_loss=0.0321, over 16976.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.268, pruned_loss=0.0411, over 3078121.58 frames. ], batch size: 109, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:41:03,061 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 01:41:24,365 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:41:54,158 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:42:14,326 INFO [train.py:904] (2/8) Epoch 14, batch 9500, loss[loss=0.1702, simple_loss=0.2655, pruned_loss=0.03743, over 15535.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2671, pruned_loss=0.04067, over 3094051.18 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:42:43,669 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:43:08,433 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 01:43:34,370 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6526, 3.9485, 2.9606, 2.2979, 2.5521, 2.4387, 4.2787, 3.4887], device='cuda:2'), covar=tensor([0.2748, 0.0688, 0.1665, 0.2535, 0.2546, 0.1818, 0.0382, 0.1141], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0252, 0.0282, 0.0279, 0.0265, 0.0226, 0.0263, 0.0294], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:43:47,717 INFO [optim.py:368] (2/8) 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 01:43:55,178 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141499.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:43:59,690 INFO [train.py:904] (2/8) Epoch 14, batch 9550, loss[loss=0.1627, simple_loss=0.2537, pruned_loss=0.03588, over 16483.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2667, pruned_loss=0.04103, over 3087900.08 frames. ], batch size: 68, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:44:01,102 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:44:23,851 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:44:42,516 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9151, 1.6844, 1.4264, 1.3863, 1.8384, 1.5160, 1.5468, 1.8988], device='cuda:2'), covar=tensor([0.0178, 0.0395, 0.0527, 0.0448, 0.0254, 0.0356, 0.0183, 0.0257], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0211, 0.0204, 0.0206, 0.0210, 0.0212, 0.0209, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:45:14,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5799, 3.8937, 4.1275, 2.2548, 3.2912, 2.8016, 3.8377, 3.9391], device='cuda:2'), covar=tensor([0.0252, 0.0720, 0.0449, 0.1801, 0.0690, 0.0827, 0.0643, 0.1003], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0143, 0.0156, 0.0143, 0.0135, 0.0123, 0.0135, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 01:45:41,072 INFO [train.py:904] (2/8) Epoch 14, batch 9600, loss[loss=0.2061, simple_loss=0.2953, pruned_loss=0.05841, over 16833.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2673, pruned_loss=0.0416, over 3064769.67 frames. ], batch size: 116, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:45:42,848 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9729, 4.2431, 4.1121, 4.0734, 3.7361, 3.8046, 3.8886, 4.2333], device='cuda:2'), covar=tensor([0.1056, 0.0862, 0.0875, 0.0726, 0.0771, 0.1480, 0.0957, 0.0931], device='cuda:2'), in_proj_covar=tensor([0.0559, 0.0688, 0.0564, 0.0497, 0.0439, 0.0450, 0.0579, 0.0526], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:46:22,462 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:46:28,480 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5698, 4.5581, 4.4057, 4.0749, 4.0818, 4.4925, 4.2868, 4.2048], device='cuda:2'), covar=tensor([0.0536, 0.0542, 0.0300, 0.0269, 0.0921, 0.0443, 0.0464, 0.0631], device='cuda:2'), in_proj_covar=tensor([0.0241, 0.0328, 0.0289, 0.0269, 0.0298, 0.0314, 0.0199, 0.0337], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-30 01:47:19,264 INFO [optim.py:368] (2/8) 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 01:47:30,277 INFO [train.py:904] (2/8) Epoch 14, batch 9650, loss[loss=0.1679, simple_loss=0.2699, pruned_loss=0.03296, over 16346.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2686, pruned_loss=0.04184, over 3058495.08 frames. ], batch size: 146, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:47:34,542 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141603.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:48:16,939 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:49:04,813 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5797, 4.7287, 4.8906, 4.7478, 4.7393, 5.2698, 4.8426, 4.5408], device='cuda:2'), covar=tensor([0.1100, 0.1608, 0.1865, 0.1698, 0.2369, 0.0941, 0.1298, 0.2174], device='cuda:2'), in_proj_covar=tensor([0.0348, 0.0492, 0.0544, 0.0417, 0.0562, 0.0572, 0.0430, 0.0564], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 01:49:16,990 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141651.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:49:17,847 INFO [train.py:904] (2/8) Epoch 14, batch 9700, loss[loss=0.1847, simple_loss=0.2763, pruned_loss=0.04658, over 16968.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2677, pruned_loss=0.04168, over 3053852.02 frames. ], batch size: 109, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:50:27,991 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141685.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:50:53,324 INFO [optim.py:368] (2/8) 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 01:51:00,294 INFO [train.py:904] (2/8) Epoch 14, batch 9750, loss[loss=0.1695, simple_loss=0.266, pruned_loss=0.03646, over 16268.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.267, pruned_loss=0.0419, over 3037199.50 frames. ], batch size: 165, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:52:38,197 INFO [train.py:904] (2/8) Epoch 14, batch 9800, loss[loss=0.1664, simple_loss=0.2677, pruned_loss=0.03257, over 16521.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.267, pruned_loss=0.04089, over 3048411.32 frames. ], batch size: 68, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:53:12,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3991, 5.7182, 5.4228, 5.5164, 5.2067, 5.1666, 5.0745, 5.8076], device='cuda:2'), covar=tensor([0.1151, 0.0765, 0.0986, 0.0707, 0.0795, 0.0659, 0.1100, 0.0791], device='cuda:2'), in_proj_covar=tensor([0.0564, 0.0694, 0.0570, 0.0503, 0.0443, 0.0454, 0.0583, 0.0529], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 01:53:18,068 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 01:53:25,947 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8995, 3.8287, 4.0013, 3.7900, 3.9368, 4.3166, 4.0214, 3.7106], device='cuda:2'), covar=tensor([0.1697, 0.2226, 0.2168, 0.2343, 0.2576, 0.1656, 0.1489, 0.2602], device='cuda:2'), in_proj_covar=tensor([0.0349, 0.0493, 0.0547, 0.0418, 0.0563, 0.0574, 0.0430, 0.0566], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 01:54:13,236 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:54:13,940 INFO [optim.py:368] (2/8) 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 01:54:17,324 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141799.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:54:21,447 INFO [train.py:904] (2/8) Epoch 14, batch 9850, loss[loss=0.175, simple_loss=0.2648, pruned_loss=0.0426, over 15310.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.268, pruned_loss=0.04045, over 3049327.40 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:54:44,607 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 01:56:04,787 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:56:14,788 INFO [train.py:904] (2/8) Epoch 14, batch 9900, loss[loss=0.1915, simple_loss=0.2884, pruned_loss=0.04735, over 17012.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2698, pruned_loss=0.04064, over 3071404.66 frames. ], batch size: 109, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:57:39,926 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-30 01:58:03,748 INFO [optim.py:368] (2/8) 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 01:58:13,944 INFO [train.py:904] (2/8) Epoch 14, batch 9950, loss[loss=0.1974, simple_loss=0.2924, pruned_loss=0.05123, over 17165.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2715, pruned_loss=0.04103, over 3069855.04 frames. ], batch size: 48, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:58:48,310 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:21,076 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:31,829 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 01:59:39,958 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2157, 1.4578, 1.8495, 2.0725, 2.1648, 2.2667, 1.7598, 2.3031], device='cuda:2'), covar=tensor([0.0160, 0.0428, 0.0235, 0.0268, 0.0256, 0.0181, 0.0355, 0.0113], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0175, 0.0158, 0.0161, 0.0172, 0.0128, 0.0175, 0.0118], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 02:00:00,574 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2357, 5.5386, 5.3461, 5.3447, 5.0719, 5.0113, 4.9352, 5.6080], device='cuda:2'), covar=tensor([0.1011, 0.0748, 0.0847, 0.0664, 0.0654, 0.0656, 0.1002, 0.0755], device='cuda:2'), in_proj_covar=tensor([0.0560, 0.0691, 0.0566, 0.0500, 0.0440, 0.0451, 0.0579, 0.0527], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:00:16,381 INFO [train.py:904] (2/8) Epoch 14, batch 10000, loss[loss=0.1675, simple_loss=0.2547, pruned_loss=0.04017, over 12807.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2697, pruned_loss=0.04046, over 3074609.47 frames. ], batch size: 246, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:01:07,113 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:25,912 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:34,607 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:50,846 INFO [optim.py:368] (2/8) 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 02:02:01,157 INFO [train.py:904] (2/8) Epoch 14, batch 10050, loss[loss=0.1611, simple_loss=0.2656, pruned_loss=0.02826, over 16885.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2694, pruned_loss=0.03991, over 3081815.45 frames. ], batch size: 96, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:02:37,574 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3944, 2.9622, 2.6275, 2.1790, 2.1388, 2.2313, 2.9996, 2.8510], device='cuda:2'), covar=tensor([0.2647, 0.0850, 0.1569, 0.2434, 0.2437, 0.1889, 0.0528, 0.1402], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0252, 0.0283, 0.0280, 0.0263, 0.0226, 0.0263, 0.0293], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:03:00,954 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:03:34,640 INFO [train.py:904] (2/8) Epoch 14, batch 10100, loss[loss=0.1504, simple_loss=0.2444, pruned_loss=0.02825, over 15212.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2687, pruned_loss=0.03975, over 3067277.43 frames. ], batch size: 190, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:04:19,924 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:04:49,707 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142097.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:04:50,457 INFO [optim.py:368] (2/8) 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 02:05:19,913 INFO [train.py:904] (2/8) Epoch 15, batch 0, loss[loss=0.1884, simple_loss=0.2749, pruned_loss=0.05096, over 17106.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2749, pruned_loss=0.05096, over 17106.00 frames. ], batch size: 47, lr: 4.62e-03, grad_scale: 8.0 2023-04-30 02:05:19,914 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 02:05:27,343 INFO [train.py:938] (2/8) Epoch 15, validation: loss=0.1501, simple_loss=0.2536, pruned_loss=0.02333, over 944034.00 frames. 2023-04-30 02:05:27,344 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 02:05:53,971 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142121.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:06:27,844 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:06:37,637 INFO [train.py:904] (2/8) Epoch 15, batch 50, loss[loss=0.1503, simple_loss=0.2451, pruned_loss=0.0277, over 17186.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.282, pruned_loss=0.05813, over 745268.92 frames. ], batch size: 46, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:07:44,552 INFO [optim.py:368] (2/8) 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 02:07:47,923 INFO [train.py:904] (2/8) Epoch 15, batch 100, loss[loss=0.173, simple_loss=0.2688, pruned_loss=0.03859, over 17256.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2769, pruned_loss=0.05563, over 1315611.24 frames. ], batch size: 52, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:08:56,687 INFO [train.py:904] (2/8) Epoch 15, batch 150, loss[loss=0.196, simple_loss=0.277, pruned_loss=0.05744, over 16414.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2737, pruned_loss=0.05306, over 1759052.61 frames. ], batch size: 75, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:09:11,304 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1711, 4.2239, 2.7751, 4.6932, 3.2299, 4.7298, 2.8015, 3.4141], device='cuda:2'), covar=tensor([0.0230, 0.0318, 0.1300, 0.0272, 0.0745, 0.0398, 0.1335, 0.0650], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0186, 0.0137, 0.0165, 0.0201, 0.0195, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 02:09:25,347 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:09:37,886 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 02:09:42,880 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:10:04,593 INFO [optim.py:368] (2/8) 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 02:10:07,770 INFO [train.py:904] (2/8) Epoch 15, batch 200, loss[loss=0.1731, simple_loss=0.2687, pruned_loss=0.03869, over 17269.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2726, pruned_loss=0.05236, over 2107673.53 frames. ], batch size: 52, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:11:17,109 INFO [train.py:904] (2/8) Epoch 15, batch 250, loss[loss=0.218, simple_loss=0.2795, pruned_loss=0.07821, over 16706.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2722, pruned_loss=0.05336, over 2358443.99 frames. ], batch size: 83, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:12:22,521 INFO [optim.py:368] (2/8) 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 02:12:25,435 INFO [train.py:904] (2/8) Epoch 15, batch 300, loss[loss=0.1598, simple_loss=0.252, pruned_loss=0.03378, over 17129.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2697, pruned_loss=0.05186, over 2570723.03 frames. ], batch size: 48, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:13:01,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7773, 1.7220, 2.2072, 2.6078, 2.6776, 2.5418, 1.8894, 2.7696], device='cuda:2'), covar=tensor([0.0127, 0.0408, 0.0273, 0.0250, 0.0210, 0.0254, 0.0401, 0.0119], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0178, 0.0162, 0.0166, 0.0175, 0.0132, 0.0180, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 02:13:36,228 INFO [train.py:904] (2/8) Epoch 15, batch 350, loss[loss=0.1706, simple_loss=0.2624, pruned_loss=0.03935, over 17121.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2668, pruned_loss=0.05005, over 2736876.67 frames. ], batch size: 47, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:14:03,700 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 02:14:42,900 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 15, batch 400, loss[loss=0.1713, simple_loss=0.2611, pruned_loss=0.04076, over 16657.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2649, pruned_loss=0.04985, over 2865713.67 frames. ], batch size: 57, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:15:02,022 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0686, 3.2196, 3.3436, 2.2023, 2.7979, 2.2847, 3.5253, 3.4857], device='cuda:2'), covar=tensor([0.0229, 0.0828, 0.0625, 0.1698, 0.0838, 0.0977, 0.0510, 0.0842], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0147, 0.0159, 0.0145, 0.0137, 0.0125, 0.0137, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:15:54,199 INFO [train.py:904] (2/8) Epoch 15, batch 450, loss[loss=0.1818, simple_loss=0.2721, pruned_loss=0.04575, over 17034.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2631, pruned_loss=0.04878, over 2958773.41 frames. ], batch size: 53, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:16:23,380 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:24,675 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:41,600 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:03,318 INFO [optim.py:368] (2/8) 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 02:17:05,223 INFO [train.py:904] (2/8) Epoch 15, batch 500, loss[loss=0.1782, simple_loss=0.271, pruned_loss=0.04266, over 17064.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2611, pruned_loss=0.04758, over 3042926.91 frames. ], batch size: 55, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:17:05,559 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2069, 5.9398, 6.0759, 5.6670, 5.8170, 6.3828, 5.9678, 5.6814], device='cuda:2'), covar=tensor([0.0900, 0.1662, 0.1829, 0.1728, 0.2262, 0.0899, 0.1231, 0.1926], device='cuda:2'), in_proj_covar=tensor([0.0376, 0.0532, 0.0584, 0.0451, 0.0608, 0.0610, 0.0462, 0.0601], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:17:14,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0148, 5.4549, 5.6150, 5.3408, 5.4023, 6.0576, 5.5425, 5.2348], device='cuda:2'), covar=tensor([0.0865, 0.1900, 0.2214, 0.1940, 0.2878, 0.1011, 0.1367, 0.2119], device='cuda:2'), in_proj_covar=tensor([0.0377, 0.0532, 0.0584, 0.0452, 0.0609, 0.0610, 0.0462, 0.0601], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:17:28,778 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142620.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:28,822 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7614, 4.9616, 5.1342, 4.9635, 4.9391, 5.5941, 5.1245, 4.7446], device='cuda:2'), covar=tensor([0.1186, 0.1922, 0.2007, 0.1789, 0.2553, 0.1046, 0.1482, 0.2191], device='cuda:2'), in_proj_covar=tensor([0.0376, 0.0533, 0.0584, 0.0452, 0.0608, 0.0611, 0.0461, 0.0601], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:17:45,635 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:49,253 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:18:13,785 INFO [train.py:904] (2/8) Epoch 15, batch 550, loss[loss=0.2003, simple_loss=0.2692, pruned_loss=0.06573, over 16911.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2599, pruned_loss=0.04655, over 3106413.37 frames. ], batch size: 116, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:18:25,044 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5976, 3.7316, 4.2191, 2.0502, 4.3143, 4.3543, 3.1953, 3.1437], device='cuda:2'), covar=tensor([0.0819, 0.0218, 0.0162, 0.1218, 0.0074, 0.0175, 0.0408, 0.0445], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0102, 0.0087, 0.0138, 0.0072, 0.0113, 0.0124, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 02:18:37,392 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.17 vs. limit=5.0 2023-04-30 02:19:02,219 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2256, 4.0536, 4.3072, 4.4530, 4.5259, 4.0865, 4.2763, 4.5142], device='cuda:2'), covar=tensor([0.1558, 0.1201, 0.1210, 0.0648, 0.0596, 0.1266, 0.1791, 0.0777], device='cuda:2'), in_proj_covar=tensor([0.0585, 0.0723, 0.0857, 0.0739, 0.0554, 0.0574, 0.0586, 0.0685], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:19:08,225 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2176, 5.2248, 5.0133, 4.5370, 5.0119, 1.9803, 4.8031, 4.9690], device='cuda:2'), covar=tensor([0.0083, 0.0071, 0.0161, 0.0338, 0.0094, 0.2407, 0.0119, 0.0177], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0132, 0.0177, 0.0160, 0.0150, 0.0193, 0.0165, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:19:22,092 INFO [optim.py:368] (2/8) 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 02:19:23,199 INFO [train.py:904] (2/8) Epoch 15, batch 600, loss[loss=0.1831, simple_loss=0.2554, pruned_loss=0.05537, over 16562.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2596, pruned_loss=0.0475, over 3150268.73 frames. ], batch size: 146, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:26,082 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0656, 3.2309, 3.4937, 2.2255, 2.9022, 2.1710, 3.6952, 3.4900], device='cuda:2'), covar=tensor([0.0205, 0.0794, 0.0509, 0.1715, 0.0784, 0.1046, 0.0376, 0.0763], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0149, 0.0159, 0.0146, 0.0138, 0.0125, 0.0137, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:19:53,087 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:19:54,163 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6457, 4.8027, 4.9719, 4.8423, 4.7523, 5.4302, 4.9812, 4.6729], device='cuda:2'), covar=tensor([0.1473, 0.2139, 0.2336, 0.2136, 0.2909, 0.1181, 0.1520, 0.2569], device='cuda:2'), in_proj_covar=tensor([0.0377, 0.0532, 0.0585, 0.0451, 0.0609, 0.0609, 0.0461, 0.0602], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:20:32,861 INFO [train.py:904] (2/8) Epoch 15, batch 650, loss[loss=0.1713, simple_loss=0.249, pruned_loss=0.04678, over 16864.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2579, pruned_loss=0.04677, over 3194089.28 frames. ], batch size: 90, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:20:40,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6745, 2.9078, 2.6978, 4.9530, 4.0810, 4.5422, 1.4763, 3.2443], device='cuda:2'), covar=tensor([0.1399, 0.0737, 0.1167, 0.0194, 0.0256, 0.0351, 0.1610, 0.0730], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0164, 0.0185, 0.0164, 0.0196, 0.0211, 0.0190, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:20:41,861 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3868, 2.2067, 2.4218, 4.1862, 2.1824, 2.6426, 2.3281, 2.4522], device='cuda:2'), covar=tensor([0.1166, 0.3539, 0.2576, 0.0561, 0.3869, 0.2425, 0.3463, 0.2980], device='cuda:2'), in_proj_covar=tensor([0.0375, 0.0412, 0.0345, 0.0322, 0.0421, 0.0472, 0.0377, 0.0481], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:20:44,277 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4641, 3.3830, 3.6868, 1.7807, 3.8064, 3.7982, 3.0431, 2.7319], device='cuda:2'), covar=tensor([0.0709, 0.0213, 0.0189, 0.1220, 0.0080, 0.0166, 0.0394, 0.0450], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0104, 0.0089, 0.0140, 0.0072, 0.0114, 0.0126, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 02:21:17,843 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:21:40,878 INFO [optim.py:368] (2/8) 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 02:21:42,123 INFO [train.py:904] (2/8) Epoch 15, batch 700, loss[loss=0.1782, simple_loss=0.271, pruned_loss=0.0427, over 17069.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2587, pruned_loss=0.0466, over 3225715.09 frames. ], batch size: 53, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:22:50,283 INFO [train.py:904] (2/8) Epoch 15, batch 750, loss[loss=0.1451, simple_loss=0.2362, pruned_loss=0.027, over 16797.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2584, pruned_loss=0.04617, over 3243703.20 frames. ], batch size: 42, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:23:57,713 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 15, batch 800, loss[loss=0.1653, simple_loss=0.2608, pruned_loss=0.0349, over 17101.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.258, pruned_loss=0.04592, over 3253890.27 frames. ], batch size: 49, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:24:29,850 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-30 02:24:36,442 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:25:08,642 INFO [train.py:904] (2/8) Epoch 15, batch 850, loss[loss=0.1941, simple_loss=0.2619, pruned_loss=0.06316, over 16766.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2573, pruned_loss=0.04582, over 3263771.39 frames. ], batch size: 124, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:26:15,122 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 15, batch 900, loss[loss=0.1652, simple_loss=0.2681, pruned_loss=0.03113, over 17040.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2559, pruned_loss=0.04474, over 3283980.19 frames. ], batch size: 50, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:27:24,767 INFO [train.py:904] (2/8) Epoch 15, batch 950, loss[loss=0.1556, simple_loss=0.2533, pruned_loss=0.0289, over 17109.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2559, pruned_loss=0.04494, over 3281471.11 frames. ], batch size: 47, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:27:46,650 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2729, 5.1910, 5.0531, 4.6395, 4.6573, 5.0995, 4.9857, 4.7341], device='cuda:2'), covar=tensor([0.0482, 0.0487, 0.0262, 0.0281, 0.1015, 0.0417, 0.0363, 0.0683], device='cuda:2'), in_proj_covar=tensor([0.0268, 0.0368, 0.0319, 0.0300, 0.0336, 0.0350, 0.0220, 0.0380], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:28:02,507 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143080.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:28:30,229 INFO [optim.py:368] (2/8) 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 02:28:31,470 INFO [train.py:904] (2/8) Epoch 15, batch 1000, loss[loss=0.1817, simple_loss=0.2495, pruned_loss=0.05701, over 11745.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2559, pruned_loss=0.04516, over 3280096.51 frames. ], batch size: 246, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:28:43,510 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8204, 3.7653, 3.9888, 3.0704, 3.5570, 4.0891, 3.7933, 2.4741], device='cuda:2'), covar=tensor([0.0416, 0.0254, 0.0047, 0.0281, 0.0107, 0.0097, 0.0078, 0.0383], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0075, 0.0075, 0.0131, 0.0088, 0.0098, 0.0085, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:28:55,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3859, 4.2808, 4.2580, 4.0059, 4.0348, 4.3066, 4.0641, 4.1171], device='cuda:2'), covar=tensor([0.0612, 0.0789, 0.0282, 0.0291, 0.0679, 0.0464, 0.0607, 0.0648], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0366, 0.0318, 0.0299, 0.0334, 0.0348, 0.0218, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:29:21,823 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8868, 3.1845, 3.2107, 2.0601, 2.7810, 2.3773, 3.2563, 3.4498], device='cuda:2'), covar=tensor([0.0348, 0.0871, 0.0559, 0.1815, 0.0870, 0.0898, 0.0695, 0.0985], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0152, 0.0162, 0.0148, 0.0139, 0.0126, 0.0139, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:29:41,403 INFO [train.py:904] (2/8) Epoch 15, batch 1050, loss[loss=0.1775, simple_loss=0.2689, pruned_loss=0.04302, over 17071.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2556, pruned_loss=0.04533, over 3285277.10 frames. ], batch size: 53, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:30:46,963 INFO [optim.py:368] (2/8) 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 02:30:49,004 INFO [train.py:904] (2/8) Epoch 15, batch 1100, loss[loss=0.1615, simple_loss=0.2519, pruned_loss=0.03561, over 17130.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2548, pruned_loss=0.04493, over 3289911.40 frames. ], batch size: 47, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:31:26,426 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:31:40,402 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6728, 4.9742, 4.7648, 4.7619, 4.5273, 4.4768, 4.3842, 5.0674], device='cuda:2'), covar=tensor([0.1127, 0.0853, 0.0986, 0.0808, 0.0761, 0.0999, 0.1125, 0.0850], device='cuda:2'), in_proj_covar=tensor([0.0613, 0.0764, 0.0622, 0.0548, 0.0482, 0.0489, 0.0641, 0.0580], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:31:58,384 INFO [train.py:904] (2/8) Epoch 15, batch 1150, loss[loss=0.1698, simple_loss=0.2456, pruned_loss=0.04698, over 16721.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2542, pruned_loss=0.04368, over 3302812.15 frames. ], batch size: 134, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:32:02,114 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7755, 2.7368, 2.7178, 4.9762, 4.0302, 4.5059, 1.6896, 3.3076], device='cuda:2'), covar=tensor([0.1383, 0.0786, 0.1113, 0.0157, 0.0240, 0.0363, 0.1509, 0.0692], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0166, 0.0187, 0.0167, 0.0198, 0.0214, 0.0191, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:32:07,013 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8708, 3.0112, 2.8193, 5.1158, 4.1809, 4.5646, 1.6786, 3.4087], device='cuda:2'), covar=tensor([0.1306, 0.0684, 0.1048, 0.0130, 0.0229, 0.0348, 0.1477, 0.0625], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0166, 0.0187, 0.0167, 0.0198, 0.0214, 0.0191, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:32:34,545 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:33:01,289 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9607, 4.0122, 4.4587, 2.1131, 4.6053, 4.6905, 3.2232, 3.7687], device='cuda:2'), covar=tensor([0.0676, 0.0247, 0.0202, 0.1151, 0.0083, 0.0128, 0.0443, 0.0316], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0105, 0.0090, 0.0141, 0.0073, 0.0116, 0.0126, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 02:33:07,201 INFO [optim.py:368] (2/8) 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 02:33:08,293 INFO [train.py:904] (2/8) Epoch 15, batch 1200, loss[loss=0.1689, simple_loss=0.256, pruned_loss=0.04085, over 15550.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.253, pruned_loss=0.04333, over 3297093.57 frames. ], batch size: 190, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:33:11,152 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3958, 3.4010, 2.0354, 3.5575, 2.5732, 3.5558, 2.2046, 2.8109], device='cuda:2'), covar=tensor([0.0224, 0.0379, 0.1489, 0.0326, 0.0727, 0.0695, 0.1422, 0.0644], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0171, 0.0192, 0.0147, 0.0170, 0.0213, 0.0200, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:33:41,556 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-04-30 02:33:43,179 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5551, 3.5994, 3.9154, 1.9014, 4.0154, 3.9964, 3.1484, 2.9768], device='cuda:2'), covar=tensor([0.0691, 0.0202, 0.0161, 0.1199, 0.0069, 0.0158, 0.0351, 0.0417], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0104, 0.0089, 0.0140, 0.0073, 0.0116, 0.0126, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 02:33:44,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8376, 4.0166, 3.7550, 3.5086, 3.2886, 3.9701, 3.5491, 3.6878], device='cuda:2'), covar=tensor([0.0991, 0.0785, 0.0493, 0.0420, 0.1308, 0.0551, 0.1794, 0.0728], device='cuda:2'), in_proj_covar=tensor([0.0272, 0.0373, 0.0325, 0.0305, 0.0340, 0.0356, 0.0222, 0.0385], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:34:16,291 INFO [train.py:904] (2/8) Epoch 15, batch 1250, loss[loss=0.1708, simple_loss=0.2467, pruned_loss=0.04748, over 16195.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2525, pruned_loss=0.04342, over 3290797.39 frames. ], batch size: 165, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:34:56,618 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143380.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:35:25,431 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143400.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:35:26,163 INFO [optim.py:368] (2/8) 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 02:35:27,819 INFO [train.py:904] (2/8) Epoch 15, batch 1300, loss[loss=0.1704, simple_loss=0.2534, pruned_loss=0.04375, over 17221.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2531, pruned_loss=0.04368, over 3299822.74 frames. ], batch size: 44, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:03,861 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:36:37,209 INFO [train.py:904] (2/8) Epoch 15, batch 1350, loss[loss=0.1466, simple_loss=0.23, pruned_loss=0.03157, over 17000.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2538, pruned_loss=0.04415, over 3310028.57 frames. ], batch size: 41, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:49,930 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:37:06,461 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:37:40,360 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3959, 2.2236, 2.4287, 4.2096, 2.2824, 2.6687, 2.2888, 2.4424], device='cuda:2'), covar=tensor([0.1147, 0.3451, 0.2510, 0.0479, 0.3389, 0.2318, 0.3510, 0.2894], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0415, 0.0348, 0.0326, 0.0423, 0.0478, 0.0381, 0.0487], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:37:45,743 INFO [optim.py:368] (2/8) 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 02:37:47,548 INFO [train.py:904] (2/8) Epoch 15, batch 1400, loss[loss=0.1845, simple_loss=0.257, pruned_loss=0.05598, over 16702.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2539, pruned_loss=0.04394, over 3317908.46 frames. ], batch size: 134, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:38:19,301 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 02:38:31,349 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 02:38:55,992 INFO [train.py:904] (2/8) Epoch 15, batch 1450, loss[loss=0.1415, simple_loss=0.222, pruned_loss=0.0305, over 15511.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2534, pruned_loss=0.04418, over 3305591.45 frames. ], batch size: 190, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:38:57,464 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 02:39:59,436 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 2023-04-30 02:40:05,552 INFO [optim.py:368] (2/8) 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 02:40:06,726 INFO [train.py:904] (2/8) Epoch 15, batch 1500, loss[loss=0.1707, simple_loss=0.2651, pruned_loss=0.03813, over 17046.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2538, pruned_loss=0.04404, over 3312685.68 frames. ], batch size: 55, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:14,536 INFO [train.py:904] (2/8) Epoch 15, batch 1550, loss[loss=0.1874, simple_loss=0.2729, pruned_loss=0.051, over 16690.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2551, pruned_loss=0.04517, over 3315644.89 frames. ], batch size: 62, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:57,479 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-30 02:42:22,872 INFO [optim.py:368] (2/8) 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 02:42:24,070 INFO [train.py:904] (2/8) Epoch 15, batch 1600, loss[loss=0.1625, simple_loss=0.246, pruned_loss=0.03953, over 17162.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2558, pruned_loss=0.04563, over 3308371.07 frames. ], batch size: 46, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:42:26,677 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8754, 3.9490, 4.2743, 4.2594, 4.2982, 4.0017, 4.0363, 3.9215], device='cuda:2'), covar=tensor([0.0373, 0.0590, 0.0397, 0.0423, 0.0466, 0.0443, 0.0760, 0.0566], device='cuda:2'), in_proj_covar=tensor([0.0375, 0.0398, 0.0394, 0.0374, 0.0439, 0.0418, 0.0508, 0.0333], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 02:43:35,422 INFO [train.py:904] (2/8) Epoch 15, batch 1650, loss[loss=0.2053, simple_loss=0.2833, pruned_loss=0.0636, over 16749.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2577, pruned_loss=0.04632, over 3302221.54 frames. ], batch size: 83, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:40,913 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:44:22,713 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.35 vs. limit=5.0 2023-04-30 02:44:46,123 INFO [optim.py:368] (2/8) 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 02:44:46,139 INFO [train.py:904] (2/8) Epoch 15, batch 1700, loss[loss=0.212, simple_loss=0.2929, pruned_loss=0.06555, over 16435.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2593, pruned_loss=0.04636, over 3308320.26 frames. ], batch size: 68, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:44:58,180 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0038, 5.4796, 5.6399, 5.4076, 5.4593, 6.0615, 5.5867, 5.2805], device='cuda:2'), covar=tensor([0.0968, 0.1826, 0.2076, 0.2263, 0.2916, 0.0987, 0.1518, 0.2523], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0552, 0.0608, 0.0467, 0.0631, 0.0633, 0.0481, 0.0622], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:45:22,437 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:45:54,337 INFO [train.py:904] (2/8) Epoch 15, batch 1750, loss[loss=0.1378, simple_loss=0.2225, pruned_loss=0.02655, over 16766.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2607, pruned_loss=0.04613, over 3318299.23 frames. ], batch size: 39, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:46:04,343 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 02:47:05,601 INFO [optim.py:368] (2/8) 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 02:47:05,617 INFO [train.py:904] (2/8) Epoch 15, batch 1800, loss[loss=0.2114, simple_loss=0.2961, pruned_loss=0.06332, over 16747.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2622, pruned_loss=0.04676, over 3314836.77 frames. ], batch size: 89, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:47:16,612 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 02:47:17,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0741, 5.5445, 5.7697, 5.4567, 5.5231, 6.1289, 5.6402, 5.2860], device='cuda:2'), covar=tensor([0.0877, 0.1837, 0.2101, 0.1966, 0.2521, 0.0975, 0.1389, 0.2390], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0551, 0.0604, 0.0465, 0.0628, 0.0631, 0.0478, 0.0620], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:47:59,710 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 02:48:15,535 INFO [train.py:904] (2/8) Epoch 15, batch 1850, loss[loss=0.1756, simple_loss=0.2612, pruned_loss=0.04502, over 16454.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2623, pruned_loss=0.04621, over 3318308.84 frames. ], batch size: 68, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:48:18,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2584, 4.1164, 4.3093, 4.4801, 4.5939, 4.1663, 4.3058, 4.5864], device='cuda:2'), covar=tensor([0.1544, 0.1175, 0.1464, 0.0711, 0.0553, 0.1110, 0.2878, 0.0628], device='cuda:2'), in_proj_covar=tensor([0.0602, 0.0741, 0.0890, 0.0763, 0.0568, 0.0591, 0.0598, 0.0707], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:48:36,056 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0151, 2.4875, 2.6475, 1.8647, 2.7511, 2.7367, 2.4294, 2.3579], device='cuda:2'), covar=tensor([0.0715, 0.0226, 0.0226, 0.0934, 0.0104, 0.0261, 0.0452, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0105, 0.0092, 0.0141, 0.0074, 0.0118, 0.0126, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 02:49:30,763 INFO [train.py:904] (2/8) Epoch 15, batch 1900, loss[loss=0.199, simple_loss=0.2692, pruned_loss=0.0644, over 16187.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.262, pruned_loss=0.04589, over 3316397.62 frames. ], batch size: 165, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:49:31,843 INFO [optim.py:368] (2/8) 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 02:49:40,855 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5558, 2.3705, 1.8461, 2.1781, 2.7244, 2.4915, 2.7508, 2.8585], device='cuda:2'), covar=tensor([0.0165, 0.0326, 0.0446, 0.0358, 0.0176, 0.0279, 0.0199, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0223, 0.0215, 0.0215, 0.0224, 0.0223, 0.0230, 0.0216], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:50:07,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7786, 2.6049, 2.2420, 4.1899, 3.4927, 4.1855, 1.3342, 2.9987], device='cuda:2'), covar=tensor([0.1401, 0.0706, 0.1323, 0.0163, 0.0192, 0.0338, 0.1635, 0.0764], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0164, 0.0184, 0.0166, 0.0198, 0.0212, 0.0188, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 02:50:39,887 INFO [train.py:904] (2/8) Epoch 15, batch 1950, loss[loss=0.1846, simple_loss=0.2556, pruned_loss=0.05677, over 16924.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2621, pruned_loss=0.04564, over 3307773.00 frames. ], batch size: 109, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:50:46,118 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:51:28,995 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2846, 2.2383, 2.7499, 3.1895, 3.0759, 3.5757, 2.5149, 3.5282], device='cuda:2'), covar=tensor([0.0168, 0.0379, 0.0266, 0.0219, 0.0209, 0.0156, 0.0338, 0.0134], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0181, 0.0165, 0.0169, 0.0179, 0.0136, 0.0181, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:51:40,405 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-30 02:51:40,917 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7518, 4.5635, 4.7639, 4.9890, 5.1507, 4.5724, 5.1008, 5.1263], device='cuda:2'), covar=tensor([0.1711, 0.1288, 0.1789, 0.0714, 0.0568, 0.0958, 0.0585, 0.0605], device='cuda:2'), in_proj_covar=tensor([0.0604, 0.0744, 0.0893, 0.0765, 0.0571, 0.0594, 0.0604, 0.0711], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:51:43,179 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 02:51:49,540 INFO [train.py:904] (2/8) Epoch 15, batch 2000, loss[loss=0.1734, simple_loss=0.2712, pruned_loss=0.03781, over 17124.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2619, pruned_loss=0.04576, over 3317501.42 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:51:51,361 INFO [optim.py:368] (2/8) 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 02:51:52,650 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:52:27,309 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:52:58,029 INFO [train.py:904] (2/8) Epoch 15, batch 2050, loss[loss=0.1822, simple_loss=0.2768, pruned_loss=0.04382, over 17102.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2624, pruned_loss=0.04626, over 3313663.49 frames. ], batch size: 47, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:53:04,246 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4918, 5.8649, 5.6380, 5.7061, 5.3033, 5.2237, 5.2537, 6.0136], device='cuda:2'), covar=tensor([0.1419, 0.0975, 0.1025, 0.0716, 0.0877, 0.0694, 0.1077, 0.0839], device='cuda:2'), in_proj_covar=tensor([0.0625, 0.0775, 0.0638, 0.0555, 0.0490, 0.0497, 0.0647, 0.0586], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:53:32,906 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:54:07,933 INFO [train.py:904] (2/8) Epoch 15, batch 2100, loss[loss=0.1615, simple_loss=0.254, pruned_loss=0.0345, over 17130.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2638, pruned_loss=0.04717, over 3309792.85 frames. ], batch size: 48, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:54:08,981 INFO [optim.py:368] (2/8) 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 02:55:17,937 INFO [train.py:904] (2/8) Epoch 15, batch 2150, loss[loss=0.177, simple_loss=0.2737, pruned_loss=0.04012, over 17116.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2641, pruned_loss=0.0471, over 3311261.01 frames. ], batch size: 48, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:55:19,609 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9181, 4.2810, 3.1223, 2.2431, 2.8004, 2.7178, 4.6900, 3.7269], device='cuda:2'), covar=tensor([0.2636, 0.0599, 0.1632, 0.2748, 0.2854, 0.1798, 0.0331, 0.1226], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0264, 0.0294, 0.0292, 0.0286, 0.0239, 0.0277, 0.0317], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 02:55:31,877 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-30 02:56:20,032 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7310, 4.5294, 4.7565, 4.9417, 5.1210, 4.5292, 5.0805, 5.1130], device='cuda:2'), covar=tensor([0.1699, 0.1209, 0.1639, 0.0733, 0.0579, 0.0999, 0.0667, 0.0587], device='cuda:2'), in_proj_covar=tensor([0.0607, 0.0750, 0.0901, 0.0773, 0.0577, 0.0599, 0.0609, 0.0715], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:56:25,351 INFO [train.py:904] (2/8) Epoch 15, batch 2200, loss[loss=0.209, simple_loss=0.2836, pruned_loss=0.0672, over 16751.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.265, pruned_loss=0.04785, over 3308039.35 frames. ], batch size: 83, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:56:27,073 INFO [optim.py:368] (2/8) 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 02:56:42,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6936, 1.7905, 1.5941, 1.4775, 1.8611, 1.5232, 1.6388, 1.8538], device='cuda:2'), covar=tensor([0.0167, 0.0247, 0.0330, 0.0327, 0.0190, 0.0260, 0.0167, 0.0184], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0222, 0.0214, 0.0214, 0.0224, 0.0222, 0.0229, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:57:03,776 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.53 vs. limit=5.0 2023-04-30 02:57:35,996 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-04-30 02:57:36,223 INFO [train.py:904] (2/8) Epoch 15, batch 2250, loss[loss=0.1654, simple_loss=0.2452, pruned_loss=0.04283, over 16818.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2652, pruned_loss=0.04763, over 3309143.54 frames. ], batch size: 39, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:57:48,252 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8002, 1.2685, 1.6475, 1.5928, 1.6572, 1.9560, 1.5269, 1.7839], device='cuda:2'), covar=tensor([0.0209, 0.0360, 0.0191, 0.0242, 0.0242, 0.0178, 0.0324, 0.0117], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0183, 0.0169, 0.0172, 0.0181, 0.0137, 0.0183, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 02:58:46,689 INFO [train.py:904] (2/8) Epoch 15, batch 2300, loss[loss=0.1907, simple_loss=0.2701, pruned_loss=0.05561, over 16514.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2651, pruned_loss=0.04761, over 3308972.86 frames. ], batch size: 146, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:47,877 INFO [optim.py:368] (2/8) 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 02:58:58,064 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:59:53,214 INFO [train.py:904] (2/8) Epoch 15, batch 2350, loss[loss=0.2004, simple_loss=0.2912, pruned_loss=0.05477, over 16645.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2646, pruned_loss=0.04733, over 3318493.16 frames. ], batch size: 62, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 03:00:20,689 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:01:02,749 INFO [train.py:904] (2/8) Epoch 15, batch 2400, loss[loss=0.175, simple_loss=0.2752, pruned_loss=0.03745, over 17097.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2656, pruned_loss=0.04842, over 3316705.94 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:01:04,724 INFO [optim.py:368] (2/8) 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 03:02:10,318 INFO [train.py:904] (2/8) Epoch 15, batch 2450, loss[loss=0.1685, simple_loss=0.2552, pruned_loss=0.04091, over 15910.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2656, pruned_loss=0.04779, over 3317808.22 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:02:17,939 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 03:02:38,570 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 03:03:17,679 INFO [train.py:904] (2/8) Epoch 15, batch 2500, loss[loss=0.1586, simple_loss=0.235, pruned_loss=0.04111, over 16827.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2651, pruned_loss=0.04766, over 3313156.49 frames. ], batch size: 83, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:03:18,671 INFO [optim.py:368] (2/8) 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 03:04:03,054 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7251, 5.0644, 4.8228, 4.8540, 4.5889, 4.4873, 4.5541, 5.1590], device='cuda:2'), covar=tensor([0.1139, 0.0885, 0.1008, 0.0735, 0.0780, 0.1171, 0.1124, 0.0871], device='cuda:2'), in_proj_covar=tensor([0.0626, 0.0777, 0.0633, 0.0556, 0.0489, 0.0494, 0.0647, 0.0588], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:04:09,449 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 03:04:26,856 INFO [train.py:904] (2/8) Epoch 15, batch 2550, loss[loss=0.1733, simple_loss=0.2722, pruned_loss=0.03719, over 17110.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2652, pruned_loss=0.04699, over 3318180.78 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:34,894 INFO [train.py:904] (2/8) Epoch 15, batch 2600, loss[loss=0.1829, simple_loss=0.2849, pruned_loss=0.04048, over 16784.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2653, pruned_loss=0.04684, over 3314838.63 frames. ], batch size: 62, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:36,053 INFO [optim.py:368] (2/8) 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 03:06:43,554 INFO [train.py:904] (2/8) Epoch 15, batch 2650, loss[loss=0.1642, simple_loss=0.2625, pruned_loss=0.033, over 17120.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2655, pruned_loss=0.04671, over 3321478.09 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:07:04,093 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-30 03:07:05,863 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:07:09,482 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0527, 5.0154, 5.5431, 5.5321, 5.5523, 5.1691, 5.1462, 4.8235], device='cuda:2'), covar=tensor([0.0328, 0.0549, 0.0330, 0.0403, 0.0428, 0.0397, 0.0893, 0.0441], device='cuda:2'), in_proj_covar=tensor([0.0374, 0.0398, 0.0394, 0.0377, 0.0441, 0.0418, 0.0510, 0.0334], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 03:07:53,565 INFO [train.py:904] (2/8) Epoch 15, batch 2700, loss[loss=0.1608, simple_loss=0.2505, pruned_loss=0.03557, over 16798.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2651, pruned_loss=0.04614, over 3326331.84 frames. ], batch size: 39, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:07:54,729 INFO [optim.py:368] (2/8) 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 03:09:02,446 INFO [train.py:904] (2/8) Epoch 15, batch 2750, loss[loss=0.1694, simple_loss=0.2478, pruned_loss=0.04553, over 16762.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2656, pruned_loss=0.04621, over 3329911.34 frames. ], batch size: 39, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:09:09,026 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-30 03:10:11,022 INFO [train.py:904] (2/8) Epoch 15, batch 2800, loss[loss=0.1961, simple_loss=0.2863, pruned_loss=0.05292, over 16734.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2656, pruned_loss=0.0458, over 3336509.35 frames. ], batch size: 62, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:10:12,144 INFO [optim.py:368] (2/8) 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 03:10:26,077 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8971, 2.4845, 1.9648, 2.2338, 2.8977, 2.6319, 3.0199, 3.0182], device='cuda:2'), covar=tensor([0.0156, 0.0324, 0.0440, 0.0398, 0.0207, 0.0310, 0.0189, 0.0219], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0224, 0.0213, 0.0215, 0.0225, 0.0223, 0.0230, 0.0218], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:10:46,139 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 03:11:21,054 INFO [train.py:904] (2/8) Epoch 15, batch 2850, loss[loss=0.1776, simple_loss=0.2514, pruned_loss=0.05196, over 16876.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2646, pruned_loss=0.04577, over 3338723.46 frames. ], batch size: 116, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:22,163 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:12:31,820 INFO [train.py:904] (2/8) Epoch 15, batch 2900, loss[loss=0.1533, simple_loss=0.2386, pruned_loss=0.034, over 16798.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2633, pruned_loss=0.04623, over 3335128.99 frames. ], batch size: 39, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:33,009 INFO [optim.py:368] (2/8) 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 03:13:06,277 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:13:40,919 INFO [train.py:904] (2/8) Epoch 15, batch 2950, loss[loss=0.1592, simple_loss=0.2471, pruned_loss=0.03564, over 17239.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2621, pruned_loss=0.04632, over 3328034.76 frames. ], batch size: 45, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:13:47,719 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:14:01,116 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:02,228 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:30,973 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-30 03:14:31,832 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:37,218 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5157, 3.9957, 4.0471, 2.9555, 3.5916, 4.1148, 3.7646, 2.2214], device='cuda:2'), covar=tensor([0.0479, 0.0147, 0.0059, 0.0320, 0.0108, 0.0103, 0.0087, 0.0457], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0078, 0.0076, 0.0133, 0.0090, 0.0101, 0.0088, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:14:49,684 INFO [train.py:904] (2/8) Epoch 15, batch 3000, loss[loss=0.1643, simple_loss=0.2597, pruned_loss=0.03442, over 17287.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2629, pruned_loss=0.04704, over 3324430.35 frames. ], batch size: 52, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:14:49,684 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 03:14:58,797 INFO [train.py:938] (2/8) Epoch 15, validation: loss=0.138, simple_loss=0.2438, pruned_loss=0.01616, over 944034.00 frames. 2023-04-30 03:14:58,797 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 03:15:00,799 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:15:34,505 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145127.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:16:07,542 INFO [train.py:904] (2/8) Epoch 15, batch 3050, loss[loss=0.1644, simple_loss=0.2603, pruned_loss=0.03422, over 17123.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2644, pruned_loss=0.04796, over 3312972.30 frames. ], batch size: 48, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:16:15,634 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5061, 2.3055, 1.8554, 2.1092, 2.6793, 2.4637, 2.7227, 2.7911], device='cuda:2'), covar=tensor([0.0155, 0.0330, 0.0407, 0.0348, 0.0176, 0.0251, 0.0179, 0.0232], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0223, 0.0212, 0.0214, 0.0223, 0.0222, 0.0231, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:16:15,765 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 03:16:29,528 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-04-30 03:17:18,166 INFO [train.py:904] (2/8) Epoch 15, batch 3100, loss[loss=0.1832, simple_loss=0.2731, pruned_loss=0.04663, over 17132.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2647, pruned_loss=0.04803, over 3313276.96 frames. ], batch size: 53, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:17:19,337 INFO [optim.py:368] (2/8) 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 03:18:28,439 INFO [train.py:904] (2/8) Epoch 15, batch 3150, loss[loss=0.1856, simple_loss=0.2601, pruned_loss=0.05552, over 16770.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2638, pruned_loss=0.04807, over 3317509.36 frames. ], batch size: 124, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:18:30,134 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9745, 4.1191, 2.5219, 4.7720, 3.2258, 4.7189, 2.7207, 3.3453], device='cuda:2'), covar=tensor([0.0252, 0.0340, 0.1377, 0.0211, 0.0678, 0.0413, 0.1304, 0.0588], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0151, 0.0170, 0.0216, 0.0199, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 03:18:38,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9714, 5.3752, 4.8304, 5.2327, 4.9113, 4.6901, 4.9019, 5.4154], device='cuda:2'), covar=tensor([0.2306, 0.1712, 0.2903, 0.1339, 0.1641, 0.1699, 0.2276, 0.1882], device='cuda:2'), in_proj_covar=tensor([0.0628, 0.0779, 0.0638, 0.0558, 0.0494, 0.0496, 0.0648, 0.0590], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:19:37,266 INFO [train.py:904] (2/8) Epoch 15, batch 3200, loss[loss=0.1491, simple_loss=0.2314, pruned_loss=0.03342, over 17207.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2616, pruned_loss=0.04688, over 3318704.63 frames. ], batch size: 44, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:19:38,468 INFO [optim.py:368] (2/8) 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 03:20:06,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2097, 2.5247, 2.2344, 2.3277, 2.9606, 2.6354, 3.1228, 3.0967], device='cuda:2'), covar=tensor([0.0147, 0.0350, 0.0384, 0.0364, 0.0203, 0.0307, 0.0225, 0.0223], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0222, 0.0211, 0.0214, 0.0223, 0.0222, 0.0230, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:20:12,948 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4474, 5.4903, 5.2841, 4.6784, 5.3060, 2.4273, 5.1342, 5.2532], device='cuda:2'), covar=tensor([0.0084, 0.0071, 0.0167, 0.0393, 0.0088, 0.2248, 0.0114, 0.0165], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0140, 0.0190, 0.0176, 0.0160, 0.0200, 0.0177, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:20:46,510 INFO [train.py:904] (2/8) Epoch 15, batch 3250, loss[loss=0.1761, simple_loss=0.2531, pruned_loss=0.0496, over 16834.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2609, pruned_loss=0.04626, over 3310046.31 frames. ], batch size: 39, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:20:46,746 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 03:20:54,913 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-30 03:21:13,309 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:57,381 INFO [train.py:904] (2/8) Epoch 15, batch 3300, loss[loss=0.1781, simple_loss=0.2716, pruned_loss=0.04232, over 17005.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2629, pruned_loss=0.04699, over 3312619.25 frames. ], batch size: 50, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:21:58,628 INFO [optim.py:368] (2/8) 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 03:22:25,239 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:22:38,772 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:23:06,148 INFO [train.py:904] (2/8) Epoch 15, batch 3350, loss[loss=0.1879, simple_loss=0.2787, pruned_loss=0.04856, over 16721.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.264, pruned_loss=0.04713, over 3308081.89 frames. ], batch size: 57, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:17,527 INFO [train.py:904] (2/8) Epoch 15, batch 3400, loss[loss=0.1625, simple_loss=0.2413, pruned_loss=0.04187, over 16782.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2634, pruned_loss=0.04677, over 3312161.24 frames. ], batch size: 76, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:18,608 INFO [optim.py:368] (2/8) 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 03:25:28,519 INFO [train.py:904] (2/8) Epoch 15, batch 3450, loss[loss=0.1485, simple_loss=0.2294, pruned_loss=0.0338, over 16855.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2621, pruned_loss=0.0464, over 3301023.78 frames. ], batch size: 39, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:25:55,444 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7357, 2.7524, 2.5438, 4.2195, 3.5526, 4.2065, 1.5594, 2.9528], device='cuda:2'), covar=tensor([0.1326, 0.0670, 0.1072, 0.0184, 0.0159, 0.0348, 0.1448, 0.0752], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0166, 0.0186, 0.0172, 0.0202, 0.0214, 0.0189, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 03:26:27,219 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145594.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:26:38,103 INFO [train.py:904] (2/8) Epoch 15, batch 3500, loss[loss=0.18, simple_loss=0.2577, pruned_loss=0.05116, over 16858.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2613, pruned_loss=0.04619, over 3298597.32 frames. ], batch size: 102, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:26:39,231 INFO [optim.py:368] (2/8) 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 03:27:35,996 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:27:47,267 INFO [train.py:904] (2/8) Epoch 15, batch 3550, loss[loss=0.1962, simple_loss=0.2928, pruned_loss=0.04985, over 17075.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2601, pruned_loss=0.04567, over 3301451.41 frames. ], batch size: 53, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:27:48,347 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:27:52,745 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:28:26,853 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9859, 4.1833, 2.3724, 4.7127, 3.2145, 4.6795, 2.4678, 3.3478], device='cuda:2'), covar=tensor([0.0246, 0.0300, 0.1498, 0.0226, 0.0667, 0.0455, 0.1518, 0.0650], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0152, 0.0170, 0.0216, 0.0200, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 03:28:31,847 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145683.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:54,822 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:57,584 INFO [train.py:904] (2/8) Epoch 15, batch 3600, loss[loss=0.1663, simple_loss=0.266, pruned_loss=0.03326, over 17118.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2582, pruned_loss=0.04457, over 3316833.08 frames. ], batch size: 48, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:28:58,727 INFO [optim.py:368] (2/8) 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 03:29:00,358 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:11,889 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6814, 4.1879, 4.0637, 1.9718, 3.1381, 2.3726, 4.1587, 4.3396], device='cuda:2'), covar=tensor([0.0284, 0.0663, 0.0480, 0.2038, 0.0907, 0.1055, 0.0524, 0.0786], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0147, 0.0140, 0.0127, 0.0141, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 03:29:26,981 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:33,189 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5067, 4.4340, 4.4040, 4.1347, 4.1333, 4.4509, 4.2398, 4.2451], device='cuda:2'), covar=tensor([0.0609, 0.0683, 0.0292, 0.0288, 0.0828, 0.0491, 0.0537, 0.0642], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0393, 0.0338, 0.0322, 0.0355, 0.0368, 0.0229, 0.0401], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:29:34,302 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:39,793 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:30:10,693 INFO [train.py:904] (2/8) Epoch 15, batch 3650, loss[loss=0.1749, simple_loss=0.2492, pruned_loss=0.05029, over 16765.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2576, pruned_loss=0.04514, over 3318212.43 frames. ], batch size: 124, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:30:38,217 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:31:24,489 INFO [train.py:904] (2/8) Epoch 15, batch 3700, loss[loss=0.184, simple_loss=0.258, pruned_loss=0.05501, over 16477.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2566, pruned_loss=0.04669, over 3297719.66 frames. ], batch size: 146, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:31:26,279 INFO [optim.py:368] (2/8) 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 03:31:46,910 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-30 03:31:52,307 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3631, 3.7395, 3.6808, 2.7169, 3.6039, 3.8947, 3.6596, 1.9410], device='cuda:2'), covar=tensor([0.0512, 0.0175, 0.0076, 0.0359, 0.0089, 0.0119, 0.0095, 0.0545], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0075, 0.0075, 0.0130, 0.0088, 0.0099, 0.0086, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:31:53,363 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:31:55,678 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0519, 4.0725, 3.9425, 3.6915, 3.7244, 4.0414, 3.6515, 3.8451], device='cuda:2'), covar=tensor([0.0591, 0.0549, 0.0271, 0.0269, 0.0664, 0.0439, 0.1093, 0.0513], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0384, 0.0330, 0.0315, 0.0348, 0.0359, 0.0223, 0.0392], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:32:38,908 INFO [train.py:904] (2/8) Epoch 15, batch 3750, loss[loss=0.1762, simple_loss=0.2466, pruned_loss=0.05287, over 16892.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.258, pruned_loss=0.04836, over 3270440.50 frames. ], batch size: 116, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:23,407 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:33:26,091 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6714, 4.6904, 4.8690, 4.7293, 4.6379, 5.3017, 4.8530, 4.5286], device='cuda:2'), covar=tensor([0.1360, 0.1850, 0.1894, 0.2138, 0.2880, 0.1095, 0.1632, 0.2696], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0549, 0.0600, 0.0470, 0.0628, 0.0625, 0.0479, 0.0619], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:33:51,793 INFO [train.py:904] (2/8) Epoch 15, batch 3800, loss[loss=0.1812, simple_loss=0.2658, pruned_loss=0.04826, over 15665.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2591, pruned_loss=0.04957, over 3262752.76 frames. ], batch size: 191, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:53,680 INFO [optim.py:368] (2/8) 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 03:34:12,537 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 03:35:00,935 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 03:35:04,375 INFO [train.py:904] (2/8) Epoch 15, batch 3850, loss[loss=0.1753, simple_loss=0.2459, pruned_loss=0.0524, over 16830.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2592, pruned_loss=0.05041, over 3256261.77 frames. ], batch size: 96, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:35:38,158 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5826, 5.9533, 5.6555, 5.7503, 5.4046, 5.2610, 5.3755, 6.0554], device='cuda:2'), covar=tensor([0.1237, 0.0838, 0.1028, 0.0744, 0.0749, 0.0605, 0.1015, 0.0779], device='cuda:2'), in_proj_covar=tensor([0.0625, 0.0775, 0.0635, 0.0558, 0.0491, 0.0493, 0.0645, 0.0591], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:36:13,419 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:36:21,009 INFO [train.py:904] (2/8) Epoch 15, batch 3900, loss[loss=0.1667, simple_loss=0.2406, pruned_loss=0.04638, over 16475.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2592, pruned_loss=0.05078, over 3268203.66 frames. ], batch size: 75, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:22,202 INFO [optim.py:368] (2/8) 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 03:36:57,924 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:37:32,699 INFO [train.py:904] (2/8) Epoch 15, batch 3950, loss[loss=0.176, simple_loss=0.2528, pruned_loss=0.04955, over 16492.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2583, pruned_loss=0.05089, over 3280719.84 frames. ], batch size: 68, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:06,891 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:38:08,311 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3363, 3.5476, 3.6373, 2.0975, 3.0786, 2.4989, 3.8520, 3.8634], device='cuda:2'), covar=tensor([0.0219, 0.0697, 0.0529, 0.1772, 0.0751, 0.0881, 0.0453, 0.0676], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0147, 0.0139, 0.0127, 0.0140, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 03:38:13,221 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4081, 2.2214, 1.6985, 1.8934, 2.5862, 2.3248, 2.6928, 2.6788], device='cuda:2'), covar=tensor([0.0230, 0.0404, 0.0561, 0.0489, 0.0247, 0.0357, 0.0225, 0.0319], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0220, 0.0212, 0.0214, 0.0222, 0.0220, 0.0230, 0.0215], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:38:19,866 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 03:38:46,133 INFO [train.py:904] (2/8) Epoch 15, batch 4000, loss[loss=0.1693, simple_loss=0.2568, pruned_loss=0.04092, over 16624.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2584, pruned_loss=0.05156, over 3286067.20 frames. ], batch size: 62, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:47,409 INFO [optim.py:368] (2/8) 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 03:39:28,957 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7147, 1.8725, 2.2612, 2.6731, 2.6826, 2.9970, 2.0280, 2.8082], device='cuda:2'), covar=tensor([0.0162, 0.0430, 0.0268, 0.0265, 0.0249, 0.0144, 0.0410, 0.0102], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0182, 0.0168, 0.0173, 0.0181, 0.0139, 0.0183, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:39:56,206 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:39:59,983 INFO [train.py:904] (2/8) Epoch 15, batch 4050, loss[loss=0.1671, simple_loss=0.2513, pruned_loss=0.04143, over 16435.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2586, pruned_loss=0.05057, over 3287835.37 frames. ], batch size: 68, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:40:36,969 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:41:13,947 INFO [train.py:904] (2/8) Epoch 15, batch 4100, loss[loss=0.1892, simple_loss=0.2745, pruned_loss=0.05191, over 16478.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2594, pruned_loss=0.0496, over 3293125.92 frames. ], batch size: 75, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:41:15,745 INFO [optim.py:368] (2/8) 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 03:41:26,641 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:30,100 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:32,940 INFO [train.py:904] (2/8) Epoch 15, batch 4150, loss[loss=0.1921, simple_loss=0.2827, pruned_loss=0.05076, over 17007.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2668, pruned_loss=0.05252, over 3248917.30 frames. ], batch size: 55, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:42:49,049 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7483, 5.1052, 5.5189, 5.4339, 5.4785, 5.0471, 4.6887, 4.6688], device='cuda:2'), covar=tensor([0.0509, 0.0610, 0.0389, 0.0505, 0.0590, 0.0575, 0.1487, 0.0567], device='cuda:2'), in_proj_covar=tensor([0.0368, 0.0397, 0.0390, 0.0370, 0.0438, 0.0411, 0.0505, 0.0329], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 03:43:45,064 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:46,699 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:50,403 INFO [train.py:904] (2/8) Epoch 15, batch 4200, loss[loss=0.2372, simple_loss=0.3221, pruned_loss=0.07616, over 15368.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2743, pruned_loss=0.05482, over 3204422.67 frames. ], batch size: 190, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:43:53,465 INFO [optim.py:368] (2/8) 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 03:44:58,658 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:45:06,384 INFO [train.py:904] (2/8) Epoch 15, batch 4250, loss[loss=0.1971, simple_loss=0.2742, pruned_loss=0.06003, over 12016.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2774, pruned_loss=0.05413, over 3191009.67 frames. ], batch size: 246, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:19,357 INFO [train.py:904] (2/8) Epoch 15, batch 4300, loss[loss=0.218, simple_loss=0.3054, pruned_loss=0.06533, over 11358.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2787, pruned_loss=0.05354, over 3184320.95 frames. ], batch size: 246, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:23,348 INFO [optim.py:368] (2/8) 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 03:46:34,217 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-04-30 03:47:31,149 INFO [train.py:904] (2/8) Epoch 15, batch 4350, loss[loss=0.2022, simple_loss=0.2901, pruned_loss=0.05718, over 16601.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2825, pruned_loss=0.05478, over 3187538.09 frames. ], batch size: 57, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:48:08,881 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:48:45,757 INFO [train.py:904] (2/8) Epoch 15, batch 4400, loss[loss=0.1957, simple_loss=0.278, pruned_loss=0.05673, over 16739.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2849, pruned_loss=0.05595, over 3183240.09 frames. ], batch size: 39, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:48:50,393 INFO [optim.py:368] (2/8) 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 03:48:51,339 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:49:58,633 INFO [train.py:904] (2/8) Epoch 15, batch 4450, loss[loss=0.2226, simple_loss=0.3126, pruned_loss=0.06636, over 15491.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2882, pruned_loss=0.05699, over 3201561.97 frames. ], batch size: 190, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:50:10,705 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 03:50:42,124 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5394, 4.8197, 4.6088, 4.6622, 4.3784, 4.3012, 4.3498, 4.8971], device='cuda:2'), covar=tensor([0.1051, 0.0795, 0.0894, 0.0704, 0.0711, 0.1165, 0.0929, 0.0740], device='cuda:2'), in_proj_covar=tensor([0.0597, 0.0744, 0.0614, 0.0537, 0.0468, 0.0478, 0.0618, 0.0564], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:50:48,967 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7941, 3.5249, 3.8562, 3.5398, 3.7196, 4.2075, 3.8869, 3.4863], device='cuda:2'), covar=tensor([0.2200, 0.2561, 0.2026, 0.2778, 0.2978, 0.2038, 0.1312, 0.2605], device='cuda:2'), in_proj_covar=tensor([0.0377, 0.0534, 0.0580, 0.0453, 0.0604, 0.0607, 0.0459, 0.0605], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:50:57,085 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146591.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:51:12,434 INFO [train.py:904] (2/8) Epoch 15, batch 4500, loss[loss=0.2194, simple_loss=0.287, pruned_loss=0.07592, over 11692.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2882, pruned_loss=0.05731, over 3197590.85 frames. ], batch size: 248, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:51:15,348 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6944, 3.6624, 2.0157, 4.3108, 2.8140, 4.2407, 2.5219, 2.9649], device='cuda:2'), covar=tensor([0.0223, 0.0304, 0.1804, 0.0104, 0.0775, 0.0344, 0.1337, 0.0689], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0171, 0.0191, 0.0145, 0.0167, 0.0211, 0.0197, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 03:51:16,066 INFO [optim.py:368] (2/8) 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 03:51:47,939 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:52:25,473 INFO [train.py:904] (2/8) Epoch 15, batch 4550, loss[loss=0.2178, simple_loss=0.2937, pruned_loss=0.07092, over 16711.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.289, pruned_loss=0.05783, over 3210793.52 frames. ], batch size: 62, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:52:25,930 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:53:16,033 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:53:37,390 INFO [train.py:904] (2/8) Epoch 15, batch 4600, loss[loss=0.1802, simple_loss=0.275, pruned_loss=0.04274, over 16705.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2901, pruned_loss=0.05828, over 3212637.49 frames. ], batch size: 62, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:53:41,728 INFO [optim.py:368] (2/8) 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 03:54:12,295 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8069, 5.1407, 5.3873, 5.0575, 5.1974, 5.8412, 5.2670, 4.9390], device='cuda:2'), covar=tensor([0.0972, 0.1725, 0.1713, 0.1978, 0.2471, 0.0833, 0.1332, 0.2248], device='cuda:2'), in_proj_covar=tensor([0.0376, 0.0531, 0.0578, 0.0453, 0.0603, 0.0606, 0.0459, 0.0603], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:54:43,955 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-30 03:54:49,172 INFO [train.py:904] (2/8) Epoch 15, batch 4650, loss[loss=0.1917, simple_loss=0.2817, pruned_loss=0.05088, over 16884.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2887, pruned_loss=0.05807, over 3215365.37 frames. ], batch size: 96, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:55:25,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0536, 2.4114, 2.5106, 1.8881, 2.6963, 2.7389, 2.4202, 2.3485], device='cuda:2'), covar=tensor([0.0619, 0.0217, 0.0195, 0.0858, 0.0097, 0.0208, 0.0401, 0.0411], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0104, 0.0091, 0.0137, 0.0072, 0.0115, 0.0122, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 03:55:48,498 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-30 03:56:03,160 INFO [train.py:904] (2/8) Epoch 15, batch 4700, loss[loss=0.1752, simple_loss=0.2656, pruned_loss=0.04244, over 16432.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2864, pruned_loss=0.05708, over 3208944.53 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:56:07,886 INFO [optim.py:368] (2/8) 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 03:56:08,861 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:56:35,040 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 03:56:51,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5953, 3.5860, 2.0160, 4.2049, 2.6834, 4.1657, 2.2641, 2.7333], device='cuda:2'), covar=tensor([0.0283, 0.0375, 0.1868, 0.0090, 0.0900, 0.0363, 0.1675, 0.0883], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0145, 0.0169, 0.0211, 0.0198, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 03:57:07,557 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:18,584 INFO [train.py:904] (2/8) Epoch 15, batch 4750, loss[loss=0.1786, simple_loss=0.2608, pruned_loss=0.04816, over 16894.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2823, pruned_loss=0.05517, over 3206952.33 frames. ], batch size: 116, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:57:20,529 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:58:09,323 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7407, 1.7443, 1.5775, 1.5245, 1.8948, 1.5873, 1.6167, 1.9220], device='cuda:2'), covar=tensor([0.0167, 0.0251, 0.0351, 0.0350, 0.0202, 0.0255, 0.0173, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0216, 0.0210, 0.0209, 0.0216, 0.0216, 0.0222, 0.0212], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:58:15,971 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3939, 2.4592, 2.0252, 2.2727, 2.8726, 2.4624, 3.0349, 3.1492], device='cuda:2'), covar=tensor([0.0089, 0.0384, 0.0500, 0.0394, 0.0220, 0.0354, 0.0180, 0.0212], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0216, 0.0210, 0.0209, 0.0216, 0.0216, 0.0221, 0.0212], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:58:31,078 INFO [train.py:904] (2/8) Epoch 15, batch 4800, loss[loss=0.1879, simple_loss=0.2874, pruned_loss=0.04423, over 16375.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2786, pruned_loss=0.05296, over 3201281.46 frames. ], batch size: 165, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:58:36,176 INFO [optim.py:368] (2/8) 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 03:58:36,798 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:58:44,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7495, 1.3965, 1.7092, 1.7085, 1.8201, 1.9765, 1.5638, 1.7855], device='cuda:2'), covar=tensor([0.0196, 0.0344, 0.0171, 0.0236, 0.0220, 0.0167, 0.0341, 0.0120], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0183, 0.0168, 0.0173, 0.0182, 0.0138, 0.0184, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 03:59:20,512 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0059, 3.2411, 3.1796, 2.1950, 2.9052, 3.2020, 3.0118, 1.7584], device='cuda:2'), covar=tensor([0.0459, 0.0047, 0.0045, 0.0323, 0.0089, 0.0108, 0.0093, 0.0438], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0073, 0.0073, 0.0129, 0.0088, 0.0098, 0.0086, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 03:59:40,650 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:59:47,611 INFO [train.py:904] (2/8) Epoch 15, batch 4850, loss[loss=0.1954, simple_loss=0.2923, pruned_loss=0.04927, over 16419.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2791, pruned_loss=0.05243, over 3188487.84 frames. ], batch size: 146, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:00:10,659 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7465, 2.6390, 2.1988, 3.6929, 2.4423, 3.7287, 1.4831, 2.6145], device='cuda:2'), covar=tensor([0.1212, 0.0646, 0.1223, 0.0120, 0.0133, 0.0463, 0.1502, 0.0864], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0165, 0.0185, 0.0166, 0.0199, 0.0209, 0.0188, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:00:18,884 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4757, 2.9633, 3.0612, 1.8946, 2.7307, 2.2025, 3.0987, 3.1486], device='cuda:2'), covar=tensor([0.0252, 0.0674, 0.0555, 0.1815, 0.0796, 0.0889, 0.0577, 0.0670], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0153, 0.0160, 0.0146, 0.0138, 0.0125, 0.0138, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:00:30,963 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5053, 4.1236, 4.1374, 2.7627, 3.5832, 4.0806, 3.6742, 2.1486], device='cuda:2'), covar=tensor([0.0455, 0.0031, 0.0028, 0.0313, 0.0090, 0.0093, 0.0090, 0.0427], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0073, 0.0073, 0.0130, 0.0088, 0.0098, 0.0086, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 04:00:34,143 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 04:01:03,669 INFO [train.py:904] (2/8) Epoch 15, batch 4900, loss[loss=0.1854, simple_loss=0.271, pruned_loss=0.04989, over 11915.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2777, pruned_loss=0.05076, over 3192519.90 frames. ], batch size: 248, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:01:08,001 INFO [optim.py:368] (2/8) 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 04:02:16,313 INFO [train.py:904] (2/8) Epoch 15, batch 4950, loss[loss=0.1901, simple_loss=0.2834, pruned_loss=0.04837, over 16173.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2774, pruned_loss=0.05003, over 3215796.92 frames. ], batch size: 165, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:02:41,848 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7896, 3.8646, 2.2546, 4.5359, 2.9396, 4.3560, 2.4577, 2.8855], device='cuda:2'), covar=tensor([0.0254, 0.0311, 0.1590, 0.0090, 0.0735, 0.0387, 0.1423, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0170, 0.0190, 0.0143, 0.0168, 0.0210, 0.0197, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:03:25,563 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9025, 5.1645, 4.8106, 4.5661, 4.0389, 5.0673, 4.9668, 4.6312], device='cuda:2'), covar=tensor([0.0917, 0.0477, 0.0448, 0.0378, 0.2004, 0.0461, 0.0325, 0.0624], device='cuda:2'), in_proj_covar=tensor([0.0260, 0.0358, 0.0312, 0.0293, 0.0322, 0.0338, 0.0209, 0.0366], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:03:28,736 INFO [train.py:904] (2/8) Epoch 15, batch 5000, loss[loss=0.174, simple_loss=0.2739, pruned_loss=0.037, over 16674.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.279, pruned_loss=0.05009, over 3216461.32 frames. ], batch size: 89, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:03:32,273 INFO [optim.py:368] (2/8) 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 04:04:12,242 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3410, 2.2754, 2.2895, 4.0795, 2.1290, 2.6339, 2.2981, 2.4018], device='cuda:2'), covar=tensor([0.1131, 0.3176, 0.2559, 0.0432, 0.3695, 0.2133, 0.3162, 0.3177], device='cuda:2'), in_proj_covar=tensor([0.0378, 0.0418, 0.0345, 0.0320, 0.0421, 0.0479, 0.0382, 0.0488], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:04:23,376 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 04:04:39,304 INFO [train.py:904] (2/8) Epoch 15, batch 5050, loss[loss=0.1778, simple_loss=0.2681, pruned_loss=0.04373, over 16509.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2795, pruned_loss=0.04999, over 3225035.24 frames. ], batch size: 75, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:44,394 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9730, 2.7173, 2.7867, 2.0781, 2.5987, 2.1511, 2.7620, 2.8957], device='cuda:2'), covar=tensor([0.0253, 0.0692, 0.0543, 0.1621, 0.0777, 0.0842, 0.0559, 0.0648], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0154, 0.0161, 0.0146, 0.0139, 0.0126, 0.0139, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:05:46,854 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:05:48,950 INFO [train.py:904] (2/8) Epoch 15, batch 5100, loss[loss=0.1712, simple_loss=0.2566, pruned_loss=0.04291, over 16577.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2783, pruned_loss=0.04972, over 3224419.40 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:52,960 INFO [optim.py:368] (2/8) 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 04:06:45,610 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:52,662 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147247.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:59,911 INFO [train.py:904] (2/8) Epoch 15, batch 5150, loss[loss=0.1961, simple_loss=0.2889, pruned_loss=0.05165, over 16501.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2787, pruned_loss=0.04932, over 3215783.48 frames. ], batch size: 75, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:07:44,998 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:03,058 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:14,125 INFO [train.py:904] (2/8) Epoch 15, batch 5200, loss[loss=0.1725, simple_loss=0.2623, pruned_loss=0.04139, over 16887.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2773, pruned_loss=0.04892, over 3223130.35 frames. ], batch size: 96, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:08:14,613 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147302.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:17,787 INFO [optim.py:368] (2/8) 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 04:08:25,229 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:55,067 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147330.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:09:27,024 INFO [train.py:904] (2/8) Epoch 15, batch 5250, loss[loss=0.1885, simple_loss=0.2764, pruned_loss=0.05033, over 16337.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2753, pruned_loss=0.0487, over 3218982.10 frames. ], batch size: 165, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:09:54,272 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:10:02,467 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5719, 2.0327, 3.0047, 3.4121, 3.1825, 3.9736, 2.1746, 3.8062], device='cuda:2'), covar=tensor([0.0110, 0.0494, 0.0235, 0.0186, 0.0208, 0.0080, 0.0562, 0.0067], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0182, 0.0168, 0.0173, 0.0181, 0.0138, 0.0185, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:10:37,510 INFO [train.py:904] (2/8) Epoch 15, batch 5300, loss[loss=0.17, simple_loss=0.2563, pruned_loss=0.04189, over 16301.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.272, pruned_loss=0.04787, over 3197787.47 frames. ], batch size: 165, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:10:40,967 INFO [optim.py:368] (2/8) 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 04:10:45,270 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:14,153 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6566, 4.8048, 4.9988, 4.7574, 4.8230, 5.3755, 4.8538, 4.5005], device='cuda:2'), covar=tensor([0.1014, 0.1691, 0.1779, 0.1749, 0.2277, 0.0881, 0.1377, 0.2146], device='cuda:2'), in_proj_covar=tensor([0.0376, 0.0526, 0.0569, 0.0448, 0.0598, 0.0606, 0.0454, 0.0597], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 04:11:23,785 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:48,187 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3565, 3.3296, 3.3905, 3.4947, 3.5365, 3.2807, 3.5055, 3.5968], device='cuda:2'), covar=tensor([0.1160, 0.0942, 0.1108, 0.0602, 0.0603, 0.2605, 0.0927, 0.0670], device='cuda:2'), in_proj_covar=tensor([0.0577, 0.0713, 0.0858, 0.0729, 0.0545, 0.0574, 0.0578, 0.0672], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:11:49,963 INFO [train.py:904] (2/8) Epoch 15, batch 5350, loss[loss=0.1873, simple_loss=0.273, pruned_loss=0.0508, over 16710.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2708, pruned_loss=0.04749, over 3192608.22 frames. ], batch size: 134, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:12:14,751 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:12:53,069 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:01,310 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:03,282 INFO [train.py:904] (2/8) Epoch 15, batch 5400, loss[loss=0.192, simple_loss=0.2829, pruned_loss=0.05056, over 17051.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2732, pruned_loss=0.04866, over 3185011.25 frames. ], batch size: 50, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:13:07,672 INFO [optim.py:368] (2/8) 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 04:13:23,052 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 04:14:13,938 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 15, batch 5450, loss[loss=0.1869, simple_loss=0.2776, pruned_loss=0.04804, over 16728.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2756, pruned_loss=0.0496, over 3181358.81 frames. ], batch size: 89, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:15:32,460 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147597.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:15:40,035 INFO [train.py:904] (2/8) Epoch 15, batch 5500, loss[loss=0.1965, simple_loss=0.2905, pruned_loss=0.05126, over 16787.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2824, pruned_loss=0.05378, over 3150696.31 frames. ], batch size: 102, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:15:45,683 INFO [optim.py:368] (2/8) 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 04:16:58,368 INFO [train.py:904] (2/8) Epoch 15, batch 5550, loss[loss=0.2238, simple_loss=0.3094, pruned_loss=0.06911, over 16752.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2902, pruned_loss=0.05946, over 3126558.77 frames. ], batch size: 124, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:17:10,693 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:22,922 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:40,332 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 04:18:21,669 INFO [train.py:904] (2/8) Epoch 15, batch 5600, loss[loss=0.1999, simple_loss=0.2877, pruned_loss=0.05606, over 16745.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2953, pruned_loss=0.06415, over 3087114.11 frames. ], batch size: 76, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:18:28,273 INFO [optim.py:368] (2/8) 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 04:18:52,995 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:19:11,826 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-30 04:19:46,190 INFO [train.py:904] (2/8) Epoch 15, batch 5650, loss[loss=0.321, simple_loss=0.3732, pruned_loss=0.1344, over 11348.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3, pruned_loss=0.06737, over 3088456.68 frames. ], batch size: 248, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:20:04,641 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:20:10,322 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3956, 2.8859, 2.6515, 2.2604, 2.3456, 2.2767, 2.9288, 2.9538], device='cuda:2'), covar=tensor([0.2012, 0.0780, 0.1303, 0.1947, 0.1851, 0.1687, 0.0520, 0.0970], device='cuda:2'), in_proj_covar=tensor([0.0310, 0.0261, 0.0292, 0.0292, 0.0284, 0.0234, 0.0277, 0.0310], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:20:47,700 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:21:05,458 INFO [train.py:904] (2/8) Epoch 15, batch 5700, loss[loss=0.239, simple_loss=0.3234, pruned_loss=0.07734, over 16272.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3023, pruned_loss=0.06969, over 3079627.14 frames. ], batch size: 165, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:21:11,567 INFO [optim.py:368] (2/8) 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 04:21:26,390 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:22:23,096 INFO [train.py:904] (2/8) Epoch 15, batch 5750, loss[loss=0.2145, simple_loss=0.3001, pruned_loss=0.06443, over 16701.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3046, pruned_loss=0.07133, over 3044712.24 frames. ], batch size: 134, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:22:32,712 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:01,481 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1808, 1.5486, 1.9322, 2.0928, 2.2159, 2.2890, 1.8118, 2.2162], device='cuda:2'), covar=tensor([0.0173, 0.0403, 0.0222, 0.0258, 0.0242, 0.0182, 0.0353, 0.0115], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0165, 0.0169, 0.0177, 0.0136, 0.0182, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:23:03,168 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:36,681 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147897.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:44,667 INFO [train.py:904] (2/8) Epoch 15, batch 5800, loss[loss=0.1875, simple_loss=0.281, pruned_loss=0.04696, over 17077.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3048, pruned_loss=0.0704, over 3054406.62 frames. ], batch size: 47, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:23:50,699 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6425, 2.6194, 1.8686, 2.7156, 2.1286, 2.7353, 2.0942, 2.3884], device='cuda:2'), covar=tensor([0.0300, 0.0356, 0.1287, 0.0338, 0.0665, 0.0516, 0.1262, 0.0597], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0172, 0.0190, 0.0144, 0.0169, 0.0211, 0.0198, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:23:51,419 INFO [optim.py:368] (2/8) 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 04:24:00,091 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:12,261 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:27,909 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6697, 3.7964, 4.0526, 4.0257, 4.0300, 3.8315, 3.6933, 3.7965], device='cuda:2'), covar=tensor([0.0495, 0.0745, 0.0518, 0.0571, 0.0661, 0.0551, 0.1331, 0.0641], device='cuda:2'), in_proj_covar=tensor([0.0362, 0.0387, 0.0383, 0.0364, 0.0432, 0.0404, 0.0499, 0.0325], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 04:24:55,226 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:05,474 INFO [train.py:904] (2/8) Epoch 15, batch 5850, loss[loss=0.1991, simple_loss=0.2968, pruned_loss=0.0507, over 16861.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3022, pruned_loss=0.06813, over 3079420.21 frames. ], batch size: 96, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:25:28,533 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:38,350 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:26:30,239 INFO [train.py:904] (2/8) Epoch 15, batch 5900, loss[loss=0.2085, simple_loss=0.2925, pruned_loss=0.06227, over 16384.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3027, pruned_loss=0.06879, over 3056677.99 frames. ], batch size: 146, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:26:34,478 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 04:26:39,372 INFO [optim.py:368] (2/8) 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 04:26:52,641 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:26:54,010 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:27:08,209 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4525, 2.9186, 2.7119, 2.2967, 2.3058, 2.2601, 2.8972, 2.9285], device='cuda:2'), covar=tensor([0.2145, 0.0687, 0.1273, 0.2127, 0.2107, 0.1753, 0.0533, 0.1006], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0262, 0.0292, 0.0292, 0.0285, 0.0235, 0.0277, 0.0313], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 04:27:29,339 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8625, 4.9679, 5.3301, 5.2772, 5.2863, 4.9738, 4.8906, 4.6859], device='cuda:2'), covar=tensor([0.0301, 0.0491, 0.0424, 0.0437, 0.0442, 0.0403, 0.0961, 0.0498], device='cuda:2'), in_proj_covar=tensor([0.0362, 0.0388, 0.0382, 0.0365, 0.0433, 0.0404, 0.0498, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 04:27:49,429 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0989, 1.9582, 2.1381, 3.6471, 1.9849, 2.2963, 2.1270, 2.0783], device='cuda:2'), covar=tensor([0.1163, 0.3621, 0.2563, 0.0507, 0.4246, 0.2500, 0.3430, 0.3517], device='cuda:2'), in_proj_covar=tensor([0.0372, 0.0410, 0.0339, 0.0315, 0.0418, 0.0471, 0.0377, 0.0478], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:27:50,061 INFO [train.py:904] (2/8) Epoch 15, batch 5950, loss[loss=0.2328, simple_loss=0.3185, pruned_loss=0.07353, over 16399.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3035, pruned_loss=0.06734, over 3065348.37 frames. ], batch size: 146, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:28:08,640 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:28:26,421 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 04:28:49,078 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:29:08,854 INFO [train.py:904] (2/8) Epoch 15, batch 6000, loss[loss=0.2073, simple_loss=0.291, pruned_loss=0.06176, over 16936.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3019, pruned_loss=0.06644, over 3091167.13 frames. ], batch size: 116, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:29:08,854 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 04:29:19,431 INFO [train.py:938] (2/8) Epoch 15, validation: loss=0.1559, simple_loss=0.2691, pruned_loss=0.0214, over 944034.00 frames. 2023-04-30 04:29:19,431 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 04:29:26,129 INFO [optim.py:368] (2/8) 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 04:29:33,078 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:30:36,856 INFO [train.py:904] (2/8) Epoch 15, batch 6050, loss[loss=0.2265, simple_loss=0.2956, pruned_loss=0.07869, over 12042.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.3003, pruned_loss=0.06549, over 3108265.15 frames. ], batch size: 246, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:31:07,983 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:31:59,320 INFO [train.py:904] (2/8) Epoch 15, batch 6100, loss[loss=0.2218, simple_loss=0.3015, pruned_loss=0.0711, over 15291.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.3003, pruned_loss=0.06519, over 3090486.70 frames. ], batch size: 190, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:32:08,601 INFO [optim.py:368] (2/8) 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 04:32:17,949 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:33:18,245 INFO [train.py:904] (2/8) Epoch 15, batch 6150, loss[loss=0.2206, simple_loss=0.3028, pruned_loss=0.06919, over 15253.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2979, pruned_loss=0.06398, over 3094034.46 frames. ], batch size: 190, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:33:42,602 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:15,495 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:37,596 INFO [train.py:904] (2/8) Epoch 15, batch 6200, loss[loss=0.1767, simple_loss=0.2574, pruned_loss=0.04797, over 17025.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.296, pruned_loss=0.06368, over 3082323.94 frames. ], batch size: 55, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:34:46,166 INFO [optim.py:368] (2/8) 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 04:34:58,577 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:58,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9466, 3.2155, 3.1474, 2.0883, 2.9574, 3.1906, 3.1083, 1.9143], device='cuda:2'), covar=tensor([0.0512, 0.0049, 0.0058, 0.0405, 0.0100, 0.0106, 0.0081, 0.0425], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0075, 0.0074, 0.0132, 0.0089, 0.0099, 0.0086, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 04:35:48,351 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:54,858 INFO [train.py:904] (2/8) Epoch 15, batch 6250, loss[loss=0.1876, simple_loss=0.2823, pruned_loss=0.04644, over 16769.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2955, pruned_loss=0.06375, over 3082893.04 frames. ], batch size: 83, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:36:11,758 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:36:11,940 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6751, 3.8617, 4.3976, 2.1477, 4.6445, 4.6022, 3.1971, 3.3410], device='cuda:2'), covar=tensor([0.0806, 0.0215, 0.0145, 0.1120, 0.0040, 0.0104, 0.0355, 0.0422], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0104, 0.0090, 0.0138, 0.0072, 0.0115, 0.0122, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 04:37:11,932 INFO [train.py:904] (2/8) Epoch 15, batch 6300, loss[loss=0.2384, simple_loss=0.2991, pruned_loss=0.08881, over 11545.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2951, pruned_loss=0.06261, over 3109212.12 frames. ], batch size: 248, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:37:21,867 INFO [optim.py:368] (2/8) 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 04:37:26,964 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7804, 5.1722, 5.3108, 5.0645, 5.1551, 5.7036, 5.1498, 4.9644], device='cuda:2'), covar=tensor([0.1014, 0.1788, 0.2309, 0.1991, 0.2368, 0.0951, 0.1691, 0.2580], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0545, 0.0591, 0.0461, 0.0615, 0.0622, 0.0472, 0.0619], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 04:37:57,319 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:38:33,525 INFO [train.py:904] (2/8) Epoch 15, batch 6350, loss[loss=0.1876, simple_loss=0.2726, pruned_loss=0.05129, over 16630.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2962, pruned_loss=0.06414, over 3101218.26 frames. ], batch size: 62, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:39:03,885 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:34,197 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:45,742 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:51,469 INFO [train.py:904] (2/8) Epoch 15, batch 6400, loss[loss=0.2859, simple_loss=0.3497, pruned_loss=0.111, over 11302.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2957, pruned_loss=0.06493, over 3098934.15 frames. ], batch size: 247, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:39:59,983 INFO [optim.py:368] (2/8) 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 04:40:08,876 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:40:18,145 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:40:53,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 04:41:07,997 INFO [train.py:904] (2/8) Epoch 15, batch 6450, loss[loss=0.1958, simple_loss=0.2785, pruned_loss=0.05657, over 16241.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2957, pruned_loss=0.06415, over 3097361.83 frames. ], batch size: 165, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:41:19,889 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:22,166 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:31,971 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:42:27,007 INFO [train.py:904] (2/8) Epoch 15, batch 6500, loss[loss=0.2008, simple_loss=0.2868, pruned_loss=0.05737, over 16630.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2935, pruned_loss=0.06335, over 3094040.02 frames. ], batch size: 62, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:42:36,999 INFO [optim.py:368] (2/8) 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 04:42:47,762 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:25,098 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4798, 3.4813, 3.4160, 2.8009, 3.3192, 2.0951, 3.1332, 2.7858], device='cuda:2'), covar=tensor([0.0138, 0.0111, 0.0158, 0.0239, 0.0096, 0.2130, 0.0136, 0.0226], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0130, 0.0178, 0.0166, 0.0150, 0.0189, 0.0167, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:43:33,196 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:46,685 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148651.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:47,985 INFO [train.py:904] (2/8) Epoch 15, batch 6550, loss[loss=0.2099, simple_loss=0.3011, pruned_loss=0.05941, over 16671.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2966, pruned_loss=0.06428, over 3102905.08 frames. ], batch size: 62, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:45:05,102 INFO [train.py:904] (2/8) Epoch 15, batch 6600, loss[loss=0.234, simple_loss=0.3045, pruned_loss=0.08176, over 11427.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2988, pruned_loss=0.06478, over 3095243.80 frames. ], batch size: 248, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:45:13,947 INFO [optim.py:368] (2/8) 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 04:45:19,952 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:26,336 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:58,846 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1297, 3.2179, 1.8225, 3.4510, 2.3747, 3.5083, 2.0583, 2.6500], device='cuda:2'), covar=tensor([0.0255, 0.0371, 0.1552, 0.0209, 0.0806, 0.0487, 0.1449, 0.0671], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0170, 0.0188, 0.0142, 0.0168, 0.0209, 0.0195, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:46:22,113 INFO [train.py:904] (2/8) Epoch 15, batch 6650, loss[loss=0.2405, simple_loss=0.3233, pruned_loss=0.07883, over 16691.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3004, pruned_loss=0.06671, over 3074650.93 frames. ], batch size: 134, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:46:32,251 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5872, 3.7870, 2.3075, 4.3546, 2.8926, 4.3030, 2.5162, 2.9820], device='cuda:2'), covar=tensor([0.0273, 0.0323, 0.1586, 0.0146, 0.0798, 0.0460, 0.1448, 0.0722], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0170, 0.0188, 0.0142, 0.0168, 0.0209, 0.0195, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:46:32,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2847, 3.5159, 3.6174, 2.0497, 3.0567, 2.3077, 3.6462, 3.6560], device='cuda:2'), covar=tensor([0.0219, 0.0720, 0.0520, 0.1923, 0.0804, 0.0911, 0.0607, 0.0922], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0154, 0.0162, 0.0146, 0.0139, 0.0127, 0.0140, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:46:50,525 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5198, 4.3557, 4.5519, 4.7592, 4.8960, 4.4211, 4.8585, 4.8918], device='cuda:2'), covar=tensor([0.1666, 0.1185, 0.1480, 0.0639, 0.0544, 0.1052, 0.0569, 0.0599], device='cuda:2'), in_proj_covar=tensor([0.0569, 0.0703, 0.0844, 0.0714, 0.0537, 0.0565, 0.0572, 0.0666], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:47:00,946 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:13,453 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:39,116 INFO [train.py:904] (2/8) Epoch 15, batch 6700, loss[loss=0.219, simple_loss=0.298, pruned_loss=0.07002, over 15384.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2994, pruned_loss=0.0668, over 3089813.55 frames. ], batch size: 190, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:47:47,145 INFO [optim.py:368] (2/8) 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 04:47:57,628 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0456, 2.3277, 2.3430, 2.8850, 2.0996, 3.2074, 1.7342, 2.7035], device='cuda:2'), covar=tensor([0.1135, 0.0602, 0.1045, 0.0167, 0.0161, 0.0383, 0.1428, 0.0682], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0165, 0.0186, 0.0167, 0.0201, 0.0209, 0.0190, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:48:54,931 INFO [train.py:904] (2/8) Epoch 15, batch 6750, loss[loss=0.1974, simple_loss=0.2836, pruned_loss=0.05556, over 16845.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2981, pruned_loss=0.06688, over 3067365.33 frames. ], batch size: 116, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:48:58,422 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:48:58,793 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 04:49:35,150 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 04:50:07,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0442, 2.2854, 2.3933, 2.7382, 1.9667, 3.1880, 1.8402, 2.6502], device='cuda:2'), covar=tensor([0.1091, 0.0603, 0.0991, 0.0168, 0.0146, 0.0380, 0.1325, 0.0668], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0166, 0.0201, 0.0209, 0.0190, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:50:09,773 INFO [train.py:904] (2/8) Epoch 15, batch 6800, loss[loss=0.2452, simple_loss=0.3149, pruned_loss=0.08782, over 11527.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2976, pruned_loss=0.06612, over 3075134.04 frames. ], batch size: 248, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:50:21,236 INFO [optim.py:368] (2/8) 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 04:50:43,506 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-30 04:51:01,883 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 04:51:15,807 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:51:28,462 INFO [train.py:904] (2/8) Epoch 15, batch 6850, loss[loss=0.2117, simple_loss=0.31, pruned_loss=0.05671, over 17297.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2983, pruned_loss=0.06639, over 3071604.71 frames. ], batch size: 52, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:51:32,579 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1643, 4.0043, 4.2144, 4.3907, 4.4998, 4.0476, 4.3865, 4.4837], device='cuda:2'), covar=tensor([0.1538, 0.1062, 0.1332, 0.0656, 0.0536, 0.1366, 0.0836, 0.0592], device='cuda:2'), in_proj_covar=tensor([0.0565, 0.0698, 0.0835, 0.0707, 0.0533, 0.0560, 0.0569, 0.0661], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:51:41,423 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 04:52:26,496 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:43,265 INFO [train.py:904] (2/8) Epoch 15, batch 6900, loss[loss=0.2156, simple_loss=0.3034, pruned_loss=0.06386, over 16785.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.3006, pruned_loss=0.06556, over 3093894.58 frames. ], batch size: 83, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:43,858 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9821, 1.8597, 2.5662, 2.8620, 2.7055, 3.2243, 2.0826, 3.3127], device='cuda:2'), covar=tensor([0.0153, 0.0428, 0.0255, 0.0224, 0.0258, 0.0142, 0.0433, 0.0096], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0179, 0.0164, 0.0168, 0.0178, 0.0135, 0.0182, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:52:52,141 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149007.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:54,700 INFO [optim.py:368] (2/8) 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 04:53:04,299 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6467, 2.5998, 1.8319, 2.7289, 2.1275, 2.7457, 2.0572, 2.3296], device='cuda:2'), covar=tensor([0.0242, 0.0346, 0.1213, 0.0205, 0.0547, 0.0453, 0.1093, 0.0532], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0170, 0.0189, 0.0143, 0.0168, 0.0209, 0.0196, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:53:58,308 INFO [train.py:904] (2/8) Epoch 15, batch 6950, loss[loss=0.189, simple_loss=0.2789, pruned_loss=0.04953, over 16783.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3026, pruned_loss=0.06748, over 3073537.22 frames. ], batch size: 83, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:53:59,966 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1998, 4.0806, 4.2512, 4.4370, 4.5449, 4.1239, 4.4500, 4.5290], device='cuda:2'), covar=tensor([0.1564, 0.1135, 0.1354, 0.0606, 0.0553, 0.1219, 0.0772, 0.0634], device='cuda:2'), in_proj_covar=tensor([0.0564, 0.0699, 0.0836, 0.0706, 0.0535, 0.0561, 0.0569, 0.0662], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:54:27,425 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:45,897 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:58,921 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8772, 4.7017, 4.9113, 5.1762, 5.3489, 4.7371, 5.3258, 5.2964], device='cuda:2'), covar=tensor([0.1861, 0.1312, 0.1718, 0.0723, 0.0566, 0.0888, 0.0660, 0.0664], device='cuda:2'), in_proj_covar=tensor([0.0564, 0.0699, 0.0837, 0.0707, 0.0536, 0.0562, 0.0570, 0.0663], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:55:00,331 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 04:55:10,879 INFO [train.py:904] (2/8) Epoch 15, batch 7000, loss[loss=0.2218, simple_loss=0.3154, pruned_loss=0.06412, over 16361.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3021, pruned_loss=0.06646, over 3076595.58 frames. ], batch size: 35, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:55:23,351 INFO [optim.py:368] (2/8) 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 04:55:57,574 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:56:25,563 INFO [train.py:904] (2/8) Epoch 15, batch 7050, loss[loss=0.2188, simple_loss=0.307, pruned_loss=0.0653, over 16673.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3027, pruned_loss=0.06579, over 3094003.80 frames. ], batch size: 124, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:56:28,520 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:21,655 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5063, 5.5199, 5.3135, 4.9299, 4.9136, 5.3526, 5.2883, 5.0648], device='cuda:2'), covar=tensor([0.0609, 0.0413, 0.0281, 0.0301, 0.1141, 0.0422, 0.0282, 0.0666], device='cuda:2'), in_proj_covar=tensor([0.0258, 0.0358, 0.0306, 0.0289, 0.0321, 0.0337, 0.0209, 0.0363], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 04:57:25,166 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1372, 3.1966, 1.8738, 3.4885, 2.3466, 3.4628, 2.1164, 2.5624], device='cuda:2'), covar=tensor([0.0281, 0.0373, 0.1665, 0.0178, 0.0861, 0.0605, 0.1433, 0.0761], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0171, 0.0189, 0.0143, 0.0169, 0.0210, 0.0197, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 04:57:40,343 INFO [train.py:904] (2/8) Epoch 15, batch 7100, loss[loss=0.2423, simple_loss=0.3046, pruned_loss=0.08997, over 11353.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.3011, pruned_loss=0.06574, over 3077138.61 frames. ], batch size: 248, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:57:40,682 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:53,981 INFO [optim.py:368] (2/8) 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 04:58:55,136 INFO [train.py:904] (2/8) Epoch 15, batch 7150, loss[loss=0.1983, simple_loss=0.2847, pruned_loss=0.05597, over 16893.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2995, pruned_loss=0.06565, over 3072518.47 frames. ], batch size: 116, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 05:00:05,782 INFO [train.py:904] (2/8) Epoch 15, batch 7200, loss[loss=0.1806, simple_loss=0.2716, pruned_loss=0.0448, over 16271.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.297, pruned_loss=0.06433, over 3048989.37 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:00:13,246 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:00:17,916 INFO [optim.py:368] (2/8) 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 05:01:26,136 INFO [train.py:904] (2/8) Epoch 15, batch 7250, loss[loss=0.2187, simple_loss=0.2991, pruned_loss=0.06915, over 16205.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2939, pruned_loss=0.0622, over 3058621.44 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:01:31,290 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:01:51,760 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 05:01:57,263 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:02:13,589 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1110, 3.9747, 4.1797, 4.3275, 4.4471, 4.0192, 4.3851, 4.4358], device='cuda:2'), covar=tensor([0.1603, 0.1116, 0.1398, 0.0663, 0.0557, 0.1406, 0.0700, 0.0664], device='cuda:2'), in_proj_covar=tensor([0.0560, 0.0695, 0.0833, 0.0704, 0.0535, 0.0562, 0.0566, 0.0659], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:02:32,226 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7765, 1.2868, 1.6739, 1.6086, 1.7377, 1.9646, 1.5423, 1.7835], device='cuda:2'), covar=tensor([0.0179, 0.0340, 0.0157, 0.0226, 0.0224, 0.0141, 0.0338, 0.0101], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0180, 0.0164, 0.0169, 0.0178, 0.0135, 0.0181, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:02:38,180 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7672, 1.7506, 1.6272, 1.5741, 1.9237, 1.5798, 1.6625, 1.9086], device='cuda:2'), covar=tensor([0.0140, 0.0214, 0.0293, 0.0258, 0.0163, 0.0196, 0.0130, 0.0155], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0216, 0.0211, 0.0212, 0.0216, 0.0217, 0.0218, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:02:42,297 INFO [train.py:904] (2/8) Epoch 15, batch 7300, loss[loss=0.2328, simple_loss=0.3025, pruned_loss=0.08159, over 11890.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2932, pruned_loss=0.06199, over 3059274.42 frames. ], batch size: 247, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:02:55,637 INFO [optim.py:368] (2/8) 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 05:03:10,460 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:03:46,155 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0315, 1.4581, 1.8259, 2.0793, 2.1647, 2.3943, 1.6136, 2.2375], device='cuda:2'), covar=tensor([0.0215, 0.0447, 0.0254, 0.0257, 0.0270, 0.0147, 0.0453, 0.0104], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0180, 0.0164, 0.0169, 0.0178, 0.0135, 0.0182, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:03:58,395 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-30 05:03:58,890 INFO [train.py:904] (2/8) Epoch 15, batch 7350, loss[loss=0.2173, simple_loss=0.3011, pruned_loss=0.0668, over 16218.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2939, pruned_loss=0.06254, over 3053928.18 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:04:28,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2914, 5.1022, 5.2743, 5.4893, 5.6410, 4.9803, 5.6493, 5.6473], device='cuda:2'), covar=tensor([0.1568, 0.1096, 0.1663, 0.0628, 0.0541, 0.0778, 0.0415, 0.0512], device='cuda:2'), in_proj_covar=tensor([0.0551, 0.0685, 0.0821, 0.0692, 0.0528, 0.0552, 0.0557, 0.0647], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:04:45,221 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 05:05:17,915 INFO [train.py:904] (2/8) Epoch 15, batch 7400, loss[loss=0.2568, simple_loss=0.333, pruned_loss=0.09027, over 15323.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2953, pruned_loss=0.06346, over 3052602.88 frames. ], batch size: 190, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:05:32,172 INFO [optim.py:368] (2/8) 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 05:05:37,411 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:06:37,031 INFO [train.py:904] (2/8) Epoch 15, batch 7450, loss[loss=0.2027, simple_loss=0.2968, pruned_loss=0.05433, over 15399.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2965, pruned_loss=0.06451, over 3078821.22 frames. ], batch size: 191, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:07:00,553 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:07:16,492 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:08:00,166 INFO [train.py:904] (2/8) Epoch 15, batch 7500, loss[loss=0.218, simple_loss=0.2998, pruned_loss=0.06814, over 16154.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2964, pruned_loss=0.06349, over 3082632.49 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:08:16,081 INFO [optim.py:368] (2/8) 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 05:08:39,458 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:08:58,820 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0724, 3.6000, 3.6125, 2.1895, 3.3717, 3.6500, 3.3789, 1.9382], device='cuda:2'), covar=tensor([0.0522, 0.0047, 0.0046, 0.0445, 0.0089, 0.0115, 0.0081, 0.0439], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0074, 0.0073, 0.0132, 0.0087, 0.0097, 0.0084, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 05:09:18,805 INFO [train.py:904] (2/8) Epoch 15, batch 7550, loss[loss=0.1911, simple_loss=0.2826, pruned_loss=0.04984, over 16864.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2948, pruned_loss=0.06322, over 3081806.71 frames. ], batch size: 83, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:10:35,809 INFO [train.py:904] (2/8) Epoch 15, batch 7600, loss[loss=0.2102, simple_loss=0.2962, pruned_loss=0.06212, over 15360.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2947, pruned_loss=0.06382, over 3076496.90 frames. ], batch size: 190, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:10:50,665 INFO [optim.py:368] (2/8) 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 05:11:19,631 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9313, 2.3858, 1.9741, 2.2344, 2.7186, 2.4729, 2.7202, 2.9541], device='cuda:2'), covar=tensor([0.0151, 0.0381, 0.0469, 0.0394, 0.0246, 0.0328, 0.0198, 0.0236], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0217, 0.0211, 0.0213, 0.0216, 0.0217, 0.0218, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:11:45,912 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149746.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:11:55,062 INFO [train.py:904] (2/8) Epoch 15, batch 7650, loss[loss=0.2043, simple_loss=0.2848, pruned_loss=0.06194, over 17160.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2955, pruned_loss=0.06439, over 3086469.08 frames. ], batch size: 46, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:11:56,479 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 05:13:13,692 INFO [train.py:904] (2/8) Epoch 15, batch 7700, loss[loss=0.2094, simple_loss=0.2888, pruned_loss=0.06498, over 16867.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2955, pruned_loss=0.06493, over 3082074.70 frames. ], batch size: 116, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:13:22,701 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149807.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:13:29,265 INFO [optim.py:368] (2/8) 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 05:14:33,190 INFO [train.py:904] (2/8) Epoch 15, batch 7750, loss[loss=0.2552, simple_loss=0.3189, pruned_loss=0.09571, over 11479.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.296, pruned_loss=0.06482, over 3079671.08 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:15:02,326 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:15:28,848 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4069, 4.4846, 4.6296, 4.4846, 4.4793, 5.0159, 4.5650, 4.3106], device='cuda:2'), covar=tensor([0.1348, 0.1966, 0.2247, 0.1937, 0.2416, 0.1027, 0.1615, 0.2432], device='cuda:2'), in_proj_covar=tensor([0.0384, 0.0541, 0.0590, 0.0457, 0.0606, 0.0617, 0.0469, 0.0609], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 05:15:51,918 INFO [train.py:904] (2/8) Epoch 15, batch 7800, loss[loss=0.1892, simple_loss=0.2772, pruned_loss=0.05056, over 16876.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2972, pruned_loss=0.06582, over 3085447.95 frames. ], batch size: 109, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:16:07,341 INFO [optim.py:368] (2/8) 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 05:16:20,884 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:16:47,251 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-04-30 05:17:06,632 INFO [train.py:904] (2/8) Epoch 15, batch 7850, loss[loss=0.2084, simple_loss=0.2945, pruned_loss=0.06118, over 15366.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.298, pruned_loss=0.06576, over 3080050.93 frames. ], batch size: 190, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:18:25,130 INFO [train.py:904] (2/8) Epoch 15, batch 7900, loss[loss=0.2505, simple_loss=0.3104, pruned_loss=0.09532, over 11432.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2964, pruned_loss=0.065, over 3075930.09 frames. ], batch size: 247, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:18:40,305 INFO [optim.py:368] (2/8) 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 05:18:52,489 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8503, 2.6737, 2.6264, 1.9349, 2.5014, 2.7123, 2.6137, 1.8747], device='cuda:2'), covar=tensor([0.0420, 0.0075, 0.0070, 0.0337, 0.0126, 0.0119, 0.0101, 0.0373], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0075, 0.0074, 0.0133, 0.0089, 0.0098, 0.0086, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 05:19:43,699 INFO [train.py:904] (2/8) Epoch 15, batch 7950, loss[loss=0.2021, simple_loss=0.2879, pruned_loss=0.05814, over 15237.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2969, pruned_loss=0.0657, over 3055864.61 frames. ], batch size: 190, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:19:48,588 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7134, 2.9123, 2.7165, 4.7684, 3.6522, 4.2507, 1.7030, 3.1009], device='cuda:2'), covar=tensor([0.1388, 0.0725, 0.1182, 0.0180, 0.0384, 0.0365, 0.1614, 0.0804], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0166, 0.0188, 0.0167, 0.0203, 0.0211, 0.0192, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 05:20:00,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1580, 2.0076, 1.6983, 1.7790, 2.2270, 1.9068, 1.9649, 2.3560], device='cuda:2'), covar=tensor([0.0173, 0.0304, 0.0416, 0.0361, 0.0214, 0.0306, 0.0181, 0.0217], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0217, 0.0211, 0.0213, 0.0216, 0.0218, 0.0219, 0.0210], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:20:25,859 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:20:58,686 INFO [train.py:904] (2/8) Epoch 15, batch 8000, loss[loss=0.2128, simple_loss=0.2983, pruned_loss=0.06372, over 17105.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2974, pruned_loss=0.06579, over 3075465.12 frames. ], batch size: 49, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:20:59,784 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 05:21:55,891 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:22:14,245 INFO [train.py:904] (2/8) Epoch 15, batch 8050, loss[loss=0.2378, simple_loss=0.2987, pruned_loss=0.0885, over 11636.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2981, pruned_loss=0.06645, over 3050404.75 frames. ], batch size: 247, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:22:18,395 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 05:22:20,178 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 05:22:41,260 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:22:51,858 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4755, 3.3153, 2.6801, 2.1495, 2.1902, 2.1593, 3.3457, 3.0494], device='cuda:2'), covar=tensor([0.2691, 0.0662, 0.1590, 0.2506, 0.2599, 0.2097, 0.0468, 0.1184], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0263, 0.0295, 0.0296, 0.0289, 0.0238, 0.0279, 0.0313], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 05:23:30,589 INFO [train.py:904] (2/8) Epoch 15, batch 8100, loss[loss=0.2061, simple_loss=0.2954, pruned_loss=0.05836, over 16930.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2984, pruned_loss=0.06646, over 3049260.10 frames. ], batch size: 109, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:23:45,513 INFO [optim.py:368] (2/8) 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 05:23:55,519 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:23:59,896 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:24:40,907 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7949, 2.7427, 2.6741, 4.3615, 3.2115, 4.1033, 1.6828, 2.8290], device='cuda:2'), covar=tensor([0.1315, 0.0722, 0.1091, 0.0160, 0.0307, 0.0418, 0.1495, 0.0852], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0166, 0.0188, 0.0167, 0.0204, 0.0211, 0.0191, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 05:24:45,710 INFO [train.py:904] (2/8) Epoch 15, batch 8150, loss[loss=0.2279, simple_loss=0.303, pruned_loss=0.07641, over 16839.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2956, pruned_loss=0.06482, over 3081201.62 frames. ], batch size: 102, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:25:11,109 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150269.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:19,018 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:58,487 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0962, 2.4384, 2.6246, 1.9195, 2.7323, 2.8022, 2.4465, 2.3694], device='cuda:2'), covar=tensor([0.0658, 0.0232, 0.0212, 0.0916, 0.0080, 0.0215, 0.0402, 0.0439], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0105, 0.0091, 0.0138, 0.0073, 0.0114, 0.0123, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 05:25:58,654 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 05:26:00,375 INFO [train.py:904] (2/8) Epoch 15, batch 8200, loss[loss=0.2099, simple_loss=0.3016, pruned_loss=0.05912, over 16724.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2928, pruned_loss=0.06406, over 3085317.97 frames. ], batch size: 89, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:26:16,345 INFO [optim.py:368] (2/8) 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 05:26:31,588 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2023-04-30 05:26:54,546 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:27:22,317 INFO [train.py:904] (2/8) Epoch 15, batch 8250, loss[loss=0.1859, simple_loss=0.2829, pruned_loss=0.04445, over 16558.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2915, pruned_loss=0.06171, over 3046672.16 frames. ], batch size: 62, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:21,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0487, 1.8817, 2.1648, 3.4803, 1.8956, 2.1943, 2.0860, 2.0026], device='cuda:2'), covar=tensor([0.1312, 0.4409, 0.2919, 0.0659, 0.5507, 0.3122, 0.4040, 0.4379], device='cuda:2'), in_proj_covar=tensor([0.0371, 0.0410, 0.0341, 0.0314, 0.0419, 0.0470, 0.0378, 0.0479], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:28:45,006 INFO [train.py:904] (2/8) Epoch 15, batch 8300, loss[loss=0.1864, simple_loss=0.2674, pruned_loss=0.05266, over 12330.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.288, pruned_loss=0.05825, over 3047406.85 frames. ], batch size: 247, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:45,699 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:28:55,968 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 05:29:01,274 INFO [optim.py:368] (2/8) 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 05:29:39,972 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:30:07,756 INFO [train.py:904] (2/8) Epoch 15, batch 8350, loss[loss=0.1784, simple_loss=0.2745, pruned_loss=0.0412, over 16746.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2871, pruned_loss=0.05581, over 3061715.32 frames. ], batch size: 83, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:29,421 INFO [train.py:904] (2/8) Epoch 15, batch 8400, loss[loss=0.1798, simple_loss=0.2748, pruned_loss=0.04239, over 16533.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2845, pruned_loss=0.05357, over 3060117.92 frames. ], batch size: 68, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:46,288 INFO [optim.py:368] (2/8) 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 05:32:24,594 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 05:32:27,358 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150537.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:32:49,402 INFO [train.py:904] (2/8) Epoch 15, batch 8450, loss[loss=0.1958, simple_loss=0.2964, pruned_loss=0.04753, over 16240.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2825, pruned_loss=0.05196, over 3057160.81 frames. ], batch size: 165, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:34:01,038 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1228, 3.4377, 3.3634, 2.3913, 3.1180, 3.4143, 3.2670, 2.1067], device='cuda:2'), covar=tensor([0.0466, 0.0041, 0.0040, 0.0349, 0.0083, 0.0073, 0.0069, 0.0399], device='cuda:2'), in_proj_covar=tensor([0.0128, 0.0073, 0.0072, 0.0130, 0.0086, 0.0095, 0.0083, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 05:34:02,215 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:34:08,623 INFO [train.py:904] (2/8) Epoch 15, batch 8500, loss[loss=0.1708, simple_loss=0.2505, pruned_loss=0.04548, over 12243.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2787, pruned_loss=0.04968, over 3027055.42 frames. ], batch size: 248, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:34:25,004 INFO [optim.py:368] (2/8) 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 05:34:43,453 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3756, 3.7534, 3.6761, 2.7686, 3.3482, 3.7000, 3.4941, 2.1973], device='cuda:2'), covar=tensor([0.0407, 0.0046, 0.0043, 0.0286, 0.0089, 0.0089, 0.0075, 0.0438], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0073, 0.0072, 0.0130, 0.0086, 0.0095, 0.0084, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 05:34:52,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9171, 2.7221, 2.5841, 2.0902, 2.4164, 2.6911, 2.6085, 1.9015], device='cuda:2'), covar=tensor([0.0370, 0.0067, 0.0061, 0.0297, 0.0107, 0.0095, 0.0095, 0.0407], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0073, 0.0072, 0.0131, 0.0086, 0.0095, 0.0084, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 05:34:54,216 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:35:30,232 INFO [train.py:904] (2/8) Epoch 15, batch 8550, loss[loss=0.187, simple_loss=0.2852, pruned_loss=0.04436, over 16726.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2763, pruned_loss=0.04857, over 3022876.80 frames. ], batch size: 83, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:37:12,304 INFO [train.py:904] (2/8) Epoch 15, batch 8600, loss[loss=0.1611, simple_loss=0.2552, pruned_loss=0.03347, over 17082.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2767, pruned_loss=0.0473, over 3034612.19 frames. ], batch size: 53, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:37:32,421 INFO [optim.py:368] (2/8) 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 05:38:14,982 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:20,224 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:51,966 INFO [train.py:904] (2/8) Epoch 15, batch 8650, loss[loss=0.1748, simple_loss=0.2635, pruned_loss=0.0431, over 12325.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2753, pruned_loss=0.04615, over 3044502.02 frames. ], batch size: 248, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:39:51,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6857, 4.0563, 3.9863, 2.1470, 3.3169, 2.6735, 4.0031, 4.1703], device='cuda:2'), covar=tensor([0.0197, 0.0633, 0.0505, 0.1949, 0.0722, 0.0897, 0.0557, 0.0788], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0150, 0.0159, 0.0146, 0.0138, 0.0125, 0.0139, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 05:40:03,477 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:40:08,300 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8619, 3.7417, 3.9393, 4.0156, 4.0865, 3.6682, 4.0605, 4.1168], device='cuda:2'), covar=tensor([0.1295, 0.0978, 0.1028, 0.0578, 0.0532, 0.1751, 0.0648, 0.0572], device='cuda:2'), in_proj_covar=tensor([0.0542, 0.0676, 0.0797, 0.0687, 0.0520, 0.0542, 0.0551, 0.0639], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:40:23,178 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:40:39,983 INFO [train.py:904] (2/8) Epoch 15, batch 8700, loss[loss=0.1765, simple_loss=0.27, pruned_loss=0.04148, over 15406.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2732, pruned_loss=0.04495, over 3075854.54 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:41:01,899 INFO [optim.py:368] (2/8) 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 05:41:16,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0773, 2.2102, 1.8596, 2.0403, 2.6081, 2.2985, 2.7713, 2.8548], device='cuda:2'), covar=tensor([0.0141, 0.0451, 0.0528, 0.0470, 0.0295, 0.0369, 0.0174, 0.0240], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0212, 0.0206, 0.0206, 0.0211, 0.0211, 0.0210, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:41:49,589 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1571, 2.4933, 2.6777, 1.9217, 2.7520, 2.8753, 2.4645, 2.5011], device='cuda:2'), covar=tensor([0.0639, 0.0263, 0.0221, 0.0937, 0.0072, 0.0179, 0.0422, 0.0388], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0101, 0.0087, 0.0134, 0.0070, 0.0110, 0.0119, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 05:42:16,602 INFO [train.py:904] (2/8) Epoch 15, batch 8750, loss[loss=0.205, simple_loss=0.2992, pruned_loss=0.05538, over 16704.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2721, pruned_loss=0.04387, over 3081555.06 frames. ], batch size: 134, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:43:29,791 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6922, 2.3379, 2.2827, 4.6362, 2.1714, 2.8057, 2.4300, 2.4521], device='cuda:2'), covar=tensor([0.0917, 0.3574, 0.2795, 0.0303, 0.4134, 0.2354, 0.3234, 0.3454], device='cuda:2'), in_proj_covar=tensor([0.0368, 0.0407, 0.0341, 0.0310, 0.0417, 0.0466, 0.0375, 0.0474], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:43:52,657 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:06,457 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:08,758 INFO [train.py:904] (2/8) Epoch 15, batch 8800, loss[loss=0.2066, simple_loss=0.2866, pruned_loss=0.06324, over 12582.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2708, pruned_loss=0.04277, over 3089378.72 frames. ], batch size: 250, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:44:28,833 INFO [optim.py:368] (2/8) 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 05:45:06,313 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150930.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:45:42,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8175, 2.6017, 2.7922, 1.9498, 2.6152, 2.0171, 2.5288, 2.6770], device='cuda:2'), covar=tensor([0.0272, 0.0841, 0.0492, 0.1880, 0.0796, 0.1027, 0.0577, 0.0765], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0148, 0.0158, 0.0144, 0.0137, 0.0124, 0.0137, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 05:45:50,511 INFO [train.py:904] (2/8) Epoch 15, batch 8850, loss[loss=0.1663, simple_loss=0.2566, pruned_loss=0.03796, over 12546.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2732, pruned_loss=0.04243, over 3067746.45 frames. ], batch size: 248, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:46:10,225 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:46:44,392 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5761, 4.3557, 4.6327, 4.7521, 4.9001, 4.3626, 4.9056, 4.9232], device='cuda:2'), covar=tensor([0.1544, 0.1089, 0.1404, 0.0641, 0.0493, 0.0958, 0.0475, 0.0494], device='cuda:2'), in_proj_covar=tensor([0.0540, 0.0671, 0.0792, 0.0685, 0.0516, 0.0540, 0.0548, 0.0634], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:46:46,307 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:47:09,394 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1423, 1.4907, 1.8561, 2.1129, 2.1414, 2.2667, 1.6907, 2.3233], device='cuda:2'), covar=tensor([0.0212, 0.0479, 0.0270, 0.0268, 0.0291, 0.0183, 0.0430, 0.0139], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0177, 0.0161, 0.0165, 0.0175, 0.0133, 0.0178, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 05:47:36,865 INFO [train.py:904] (2/8) Epoch 15, batch 8900, loss[loss=0.1915, simple_loss=0.2983, pruned_loss=0.04239, over 16691.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2739, pruned_loss=0.04181, over 3084687.40 frames. ], batch size: 134, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:47:59,490 INFO [optim.py:368] (2/8) 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 05:49:02,428 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3490, 2.5089, 2.1129, 2.3320, 2.9105, 2.6555, 3.0367, 3.1075], device='cuda:2'), covar=tensor([0.0106, 0.0358, 0.0448, 0.0383, 0.0226, 0.0300, 0.0210, 0.0210], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0211, 0.0205, 0.0206, 0.0210, 0.0210, 0.0209, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:49:42,111 INFO [train.py:904] (2/8) Epoch 15, batch 8950, loss[loss=0.1609, simple_loss=0.2552, pruned_loss=0.03334, over 16768.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2733, pruned_loss=0.04202, over 3099967.79 frames. ], batch size: 76, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:01,604 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:51:01,920 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-30 05:51:30,995 INFO [train.py:904] (2/8) Epoch 15, batch 9000, loss[loss=0.1721, simple_loss=0.268, pruned_loss=0.03813, over 16191.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2697, pruned_loss=0.04081, over 3084345.15 frames. ], batch size: 166, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:30,995 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 05:51:40,821 INFO [train.py:938] (2/8) Epoch 15, validation: loss=0.15, simple_loss=0.2539, pruned_loss=0.02307, over 944034.00 frames. 2023-04-30 05:51:40,821 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 05:52:03,718 INFO [optim.py:368] (2/8) 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 05:52:39,778 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9285, 4.2208, 4.0940, 4.0826, 3.7443, 3.8082, 3.9156, 4.2283], device='cuda:2'), covar=tensor([0.1170, 0.0825, 0.0813, 0.0671, 0.0803, 0.1572, 0.0866, 0.0935], device='cuda:2'), in_proj_covar=tensor([0.0577, 0.0714, 0.0583, 0.0512, 0.0447, 0.0462, 0.0588, 0.0542], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:53:23,276 INFO [train.py:904] (2/8) Epoch 15, batch 9050, loss[loss=0.1777, simple_loss=0.2606, pruned_loss=0.0474, over 12469.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2703, pruned_loss=0.04117, over 3080812.52 frames. ], batch size: 247, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:54:27,818 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:54:51,812 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:55:09,971 INFO [train.py:904] (2/8) Epoch 15, batch 9100, loss[loss=0.1954, simple_loss=0.3041, pruned_loss=0.04331, over 15449.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2701, pruned_loss=0.04138, over 3080258.81 frames. ], batch size: 192, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:55:31,070 INFO [optim.py:368] (2/8) 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 05:55:46,825 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3723, 5.7346, 5.5161, 5.5128, 5.2032, 5.1255, 5.1918, 5.8206], device='cuda:2'), covar=tensor([0.1215, 0.0815, 0.0827, 0.0729, 0.0796, 0.0640, 0.0971, 0.0818], device='cuda:2'), in_proj_covar=tensor([0.0579, 0.0716, 0.0586, 0.0514, 0.0448, 0.0465, 0.0592, 0.0544], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 05:56:44,481 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:56:54,402 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:57:08,972 INFO [train.py:904] (2/8) Epoch 15, batch 9150, loss[loss=0.1667, simple_loss=0.2603, pruned_loss=0.03656, over 15338.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2705, pruned_loss=0.04105, over 3074021.22 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:57:18,559 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:00,935 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 05:58:12,615 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:14,781 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0732, 2.7569, 2.8547, 2.1070, 2.6474, 2.1141, 2.7289, 2.8950], device='cuda:2'), covar=tensor([0.0309, 0.0785, 0.0468, 0.1674, 0.0737, 0.0905, 0.0606, 0.0804], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0145, 0.0156, 0.0143, 0.0135, 0.0123, 0.0135, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 05:58:41,076 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 05:58:46,533 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:51,677 INFO [train.py:904] (2/8) Epoch 15, batch 9200, loss[loss=0.1691, simple_loss=0.2606, pruned_loss=0.0388, over 16645.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.266, pruned_loss=0.04028, over 3070029.78 frames. ], batch size: 62, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:59:12,236 INFO [optim.py:368] (2/8) 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 06:00:03,935 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6499, 2.0770, 1.7421, 1.9085, 2.4245, 2.1043, 2.2714, 2.5629], device='cuda:2'), covar=tensor([0.0142, 0.0378, 0.0495, 0.0407, 0.0240, 0.0357, 0.0172, 0.0241], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0215, 0.0209, 0.0208, 0.0213, 0.0214, 0.0212, 0.0205], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:00:09,252 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3489, 4.6632, 4.4842, 4.4904, 4.1721, 4.2070, 4.2024, 4.7142], device='cuda:2'), covar=tensor([0.1136, 0.0869, 0.0919, 0.0755, 0.0903, 0.1323, 0.1089, 0.0839], device='cuda:2'), in_proj_covar=tensor([0.0577, 0.0715, 0.0582, 0.0513, 0.0449, 0.0464, 0.0592, 0.0543], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:00:09,343 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151343.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:00:24,878 INFO [train.py:904] (2/8) Epoch 15, batch 9250, loss[loss=0.187, simple_loss=0.2763, pruned_loss=0.04885, over 16422.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2658, pruned_loss=0.04018, over 3075676.95 frames. ], batch size: 146, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:00:40,441 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:01:19,124 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2545, 4.2063, 4.1019, 3.5699, 4.1591, 1.7193, 3.9353, 3.8803], device='cuda:2'), covar=tensor([0.0074, 0.0073, 0.0152, 0.0206, 0.0079, 0.2474, 0.0118, 0.0184], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0129, 0.0175, 0.0158, 0.0149, 0.0189, 0.0163, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:01:43,237 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:02:14,982 INFO [train.py:904] (2/8) Epoch 15, batch 9300, loss[loss=0.169, simple_loss=0.258, pruned_loss=0.03998, over 15149.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2649, pruned_loss=0.03999, over 3075713.70 frames. ], batch size: 191, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:02:37,909 INFO [optim.py:368] (2/8) 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 06:02:51,196 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6597, 2.7067, 2.4536, 4.2139, 2.8380, 4.0561, 1.4236, 3.0193], device='cuda:2'), covar=tensor([0.1369, 0.0727, 0.1168, 0.0163, 0.0132, 0.0339, 0.1620, 0.0654], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0159, 0.0181, 0.0158, 0.0188, 0.0201, 0.0184, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:2') 2023-04-30 06:03:08,665 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2023-04-30 06:03:29,231 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:03:38,222 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1982, 4.1274, 4.0158, 3.3773, 4.1108, 1.6398, 3.8575, 3.7835], device='cuda:2'), covar=tensor([0.0105, 0.0102, 0.0176, 0.0303, 0.0107, 0.2825, 0.0156, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0128, 0.0173, 0.0157, 0.0148, 0.0188, 0.0162, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:03:59,110 INFO [train.py:904] (2/8) Epoch 15, batch 9350, loss[loss=0.1641, simple_loss=0.2596, pruned_loss=0.03425, over 16754.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2642, pruned_loss=0.03983, over 3085741.97 frames. ], batch size: 89, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:04:17,538 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-30 06:05:04,291 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8574, 3.1660, 3.4186, 1.8752, 2.7949, 2.1459, 3.3385, 3.2716], device='cuda:2'), covar=tensor([0.0294, 0.0823, 0.0468, 0.2078, 0.0826, 0.1005, 0.0705, 0.0999], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0145, 0.0157, 0.0144, 0.0136, 0.0123, 0.0136, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 06:05:36,965 INFO [train.py:904] (2/8) Epoch 15, batch 9400, loss[loss=0.187, simple_loss=0.2897, pruned_loss=0.04216, over 16617.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2649, pruned_loss=0.04008, over 3077072.97 frames. ], batch size: 134, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:05:59,180 INFO [optim.py:368] (2/8) 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 06:06:55,071 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:06:57,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3452, 4.6639, 4.5110, 4.4819, 4.1659, 4.2152, 4.2045, 4.7071], device='cuda:2'), covar=tensor([0.1143, 0.0929, 0.0921, 0.0732, 0.0893, 0.1269, 0.1104, 0.0830], device='cuda:2'), in_proj_covar=tensor([0.0571, 0.0710, 0.0575, 0.0510, 0.0445, 0.0460, 0.0586, 0.0538], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:07:17,458 INFO [train.py:904] (2/8) Epoch 15, batch 9450, loss[loss=0.149, simple_loss=0.2441, pruned_loss=0.02692, over 16391.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2664, pruned_loss=0.0404, over 3062329.76 frames. ], batch size: 68, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:07:24,089 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:07:57,643 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8882, 4.8095, 4.6486, 4.2052, 4.7236, 1.9190, 4.4957, 4.5943], device='cuda:2'), covar=tensor([0.0058, 0.0077, 0.0149, 0.0262, 0.0082, 0.2181, 0.0097, 0.0162], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0127, 0.0172, 0.0156, 0.0147, 0.0187, 0.0162, 0.0152], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:08:58,267 INFO [train.py:904] (2/8) Epoch 15, batch 9500, loss[loss=0.1747, simple_loss=0.2673, pruned_loss=0.04103, over 16227.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2658, pruned_loss=0.04015, over 3060243.03 frames. ], batch size: 165, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:09:04,124 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:09:22,316 INFO [optim.py:368] (2/8) 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 06:10:10,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1080, 3.5069, 3.6393, 2.0640, 2.9007, 2.5536, 3.3795, 3.6112], device='cuda:2'), covar=tensor([0.0354, 0.0803, 0.0542, 0.1980, 0.0853, 0.0878, 0.0972, 0.1087], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0144, 0.0156, 0.0143, 0.0135, 0.0122, 0.0135, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 06:10:11,840 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:10:35,980 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8700, 3.7229, 3.8808, 3.7261, 3.8738, 4.2825, 3.9833, 3.6068], device='cuda:2'), covar=tensor([0.1814, 0.2596, 0.2483, 0.2427, 0.2746, 0.1536, 0.1611, 0.2453], device='cuda:2'), in_proj_covar=tensor([0.0358, 0.0509, 0.0556, 0.0427, 0.0569, 0.0587, 0.0440, 0.0567], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 06:10:44,450 INFO [train.py:904] (2/8) Epoch 15, batch 9550, loss[loss=0.1702, simple_loss=0.273, pruned_loss=0.03372, over 15337.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.265, pruned_loss=0.04002, over 3069341.71 frames. ], batch size: 191, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:10:49,179 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:11:58,456 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 15, batch 9600, loss[loss=0.1783, simple_loss=0.2609, pruned_loss=0.04783, over 12158.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2664, pruned_loss=0.04095, over 3042305.05 frames. ], batch size: 250, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:12:44,296 INFO [optim.py:368] (2/8) 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 06:13:39,250 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3982, 1.9244, 1.6431, 1.6164, 2.2410, 1.8691, 2.0091, 2.3262], device='cuda:2'), covar=tensor([0.0161, 0.0381, 0.0465, 0.0472, 0.0252, 0.0367, 0.0182, 0.0238], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0214, 0.0207, 0.0207, 0.0212, 0.0214, 0.0209, 0.0203], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:13:53,529 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-30 06:14:04,280 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:14:11,685 INFO [train.py:904] (2/8) Epoch 15, batch 9650, loss[loss=0.1865, simple_loss=0.2819, pruned_loss=0.04559, over 16277.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2685, pruned_loss=0.04122, over 3040890.33 frames. ], batch size: 165, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:14:43,758 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-30 06:15:57,530 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:15:58,769 INFO [train.py:904] (2/8) Epoch 15, batch 9700, loss[loss=0.1714, simple_loss=0.2771, pruned_loss=0.03287, over 16862.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2668, pruned_loss=0.04081, over 3027387.57 frames. ], batch size: 102, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:16:19,905 INFO [optim.py:368] (2/8) 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 06:17:18,829 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151840.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:17:41,644 INFO [train.py:904] (2/8) Epoch 15, batch 9750, loss[loss=0.1373, simple_loss=0.2162, pruned_loss=0.02923, over 16229.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2661, pruned_loss=0.04098, over 3032591.50 frames. ], batch size: 35, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:18:01,483 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:18:55,492 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:19:18,821 INFO [train.py:904] (2/8) Epoch 15, batch 9800, loss[loss=0.1759, simple_loss=0.2824, pruned_loss=0.03474, over 16775.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2656, pruned_loss=0.0395, over 3037355.82 frames. ], batch size: 83, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:19:40,658 INFO [optim.py:368] (2/8) 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 06:20:30,617 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:21:03,084 INFO [train.py:904] (2/8) Epoch 15, batch 9850, loss[loss=0.182, simple_loss=0.2672, pruned_loss=0.04837, over 12231.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2665, pruned_loss=0.03887, over 3041983.67 frames. ], batch size: 247, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:21:08,470 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:21:21,935 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1623, 3.2246, 1.9080, 3.5270, 2.2931, 3.4831, 2.0567, 2.6142], device='cuda:2'), covar=tensor([0.0272, 0.0348, 0.1488, 0.0162, 0.0946, 0.0468, 0.1550, 0.0701], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0160, 0.0181, 0.0135, 0.0165, 0.0197, 0.0192, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 06:22:05,461 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2023-04-30 06:22:17,651 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151986.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:22:57,937 INFO [train.py:904] (2/8) Epoch 15, batch 9900, loss[loss=0.1752, simple_loss=0.2722, pruned_loss=0.03909, over 15306.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2664, pruned_loss=0.0389, over 3018628.50 frames. ], batch size: 192, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:22:58,683 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152002.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:23:24,819 INFO [optim.py:368] (2/8) 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 06:23:54,863 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7323, 1.8257, 2.2608, 2.6658, 2.4907, 3.0131, 1.9948, 2.9703], device='cuda:2'), covar=tensor([0.0177, 0.0409, 0.0281, 0.0251, 0.0269, 0.0152, 0.0410, 0.0128], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0175, 0.0160, 0.0164, 0.0173, 0.0131, 0.0176, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 06:23:58,599 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 06:24:37,870 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:24:55,576 INFO [train.py:904] (2/8) Epoch 15, batch 9950, loss[loss=0.1816, simple_loss=0.2742, pruned_loss=0.04454, over 16846.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2682, pruned_loss=0.03954, over 3022603.37 frames. ], batch size: 124, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:26:56,976 INFO [train.py:904] (2/8) Epoch 15, batch 10000, loss[loss=0.1766, simple_loss=0.276, pruned_loss=0.0386, over 16748.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2674, pruned_loss=0.03929, over 3056027.42 frames. ], batch size: 134, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:27:18,765 INFO [optim.py:368] (2/8) 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 06:27:38,614 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1463, 1.5663, 1.8105, 2.0538, 2.1076, 2.3263, 1.6993, 2.2516], device='cuda:2'), covar=tensor([0.0201, 0.0411, 0.0275, 0.0278, 0.0294, 0.0184, 0.0439, 0.0134], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0177, 0.0162, 0.0165, 0.0176, 0.0132, 0.0179, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 06:28:35,912 INFO [train.py:904] (2/8) Epoch 15, batch 10050, loss[loss=0.1717, simple_loss=0.2638, pruned_loss=0.03978, over 12223.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2674, pruned_loss=0.039, over 3047329.43 frames. ], batch size: 248, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:28:45,745 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:30:08,501 INFO [train.py:904] (2/8) Epoch 15, batch 10100, loss[loss=0.1794, simple_loss=0.268, pruned_loss=0.04535, over 12067.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2677, pruned_loss=0.03896, over 3050049.35 frames. ], batch size: 246, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:30:13,156 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8535, 1.4132, 1.7161, 1.7163, 1.7615, 1.9347, 1.6679, 1.8335], device='cuda:2'), covar=tensor([0.0211, 0.0324, 0.0172, 0.0243, 0.0270, 0.0182, 0.0320, 0.0111], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0177, 0.0162, 0.0164, 0.0175, 0.0131, 0.0178, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:2') 2023-04-30 06:30:28,184 INFO [optim.py:368] (2/8) 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 06:30:55,303 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:31:24,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9733, 2.7003, 2.9138, 2.0372, 2.6624, 2.1778, 2.6409, 2.8476], device='cuda:2'), covar=tensor([0.0346, 0.0805, 0.0464, 0.1780, 0.0805, 0.0876, 0.0742, 0.0790], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0143, 0.0156, 0.0144, 0.0135, 0.0123, 0.0136, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 06:31:53,397 INFO [train.py:904] (2/8) Epoch 16, batch 0, loss[loss=0.2008, simple_loss=0.2888, pruned_loss=0.05643, over 17043.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2888, pruned_loss=0.05643, over 17043.00 frames. ], batch size: 55, lr: 4.32e-03, grad_scale: 8.0 2023-04-30 06:31:53,397 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 06:32:00,884 INFO [train.py:938] (2/8) Epoch 16, validation: loss=0.1492, simple_loss=0.2525, pruned_loss=0.02288, over 944034.00 frames. 2023-04-30 06:32:00,885 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 06:32:10,244 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 06:32:29,742 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:32:29,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4296, 3.7084, 4.2490, 2.2302, 3.1343, 2.5765, 3.8546, 3.8706], device='cuda:2'), covar=tensor([0.0329, 0.0808, 0.0407, 0.1839, 0.0802, 0.0922, 0.0672, 0.1054], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0145, 0.0157, 0.0145, 0.0137, 0.0124, 0.0137, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 06:32:48,511 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:33:09,963 INFO [train.py:904] (2/8) Epoch 16, batch 50, loss[loss=0.2073, simple_loss=0.2712, pruned_loss=0.07164, over 16737.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2781, pruned_loss=0.05692, over 750995.96 frames. ], batch size: 83, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:33:12,412 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 06:33:29,876 INFO [optim.py:368] (2/8) 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 06:33:53,551 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:07,570 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 100, loss[loss=0.2093, simple_loss=0.2825, pruned_loss=0.06804, over 16682.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2739, pruned_loss=0.05458, over 1311009.41 frames. ], batch size: 134, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:35:14,870 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:35:26,676 INFO [train.py:904] (2/8) Epoch 16, batch 150, loss[loss=0.1497, simple_loss=0.2394, pruned_loss=0.02999, over 16863.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2711, pruned_loss=0.05213, over 1762404.02 frames. ], batch size: 42, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:35:48,055 INFO [optim.py:368] (2/8) 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 06:35:53,416 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 06:36:24,959 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:36:35,176 INFO [train.py:904] (2/8) Epoch 16, batch 200, loss[loss=0.1822, simple_loss=0.2762, pruned_loss=0.04411, over 17082.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2701, pruned_loss=0.05085, over 2110253.58 frames. ], batch size: 55, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:36:42,915 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:44,243 INFO [train.py:904] (2/8) Epoch 16, batch 250, loss[loss=0.1661, simple_loss=0.2631, pruned_loss=0.03454, over 17225.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2672, pruned_loss=0.04997, over 2376634.79 frames. ], batch size: 52, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:37:48,057 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:49,558 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:38:05,708 INFO [optim.py:368] (2/8) 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 06:38:53,027 INFO [train.py:904] (2/8) Epoch 16, batch 300, loss[loss=0.1797, simple_loss=0.2572, pruned_loss=0.05104, over 16329.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2647, pruned_loss=0.049, over 2590234.84 frames. ], batch size: 165, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:39:07,982 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 06:39:33,772 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:35,039 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:43,037 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:54,920 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4068, 2.9367, 2.6553, 2.1987, 2.2528, 2.2934, 2.9254, 2.7692], device='cuda:2'), covar=tensor([0.2334, 0.0843, 0.1540, 0.2215, 0.2254, 0.1868, 0.0500, 0.1191], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0258, 0.0290, 0.0288, 0.0275, 0.0234, 0.0274, 0.0307], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:40:01,537 INFO [train.py:904] (2/8) Epoch 16, batch 350, loss[loss=0.182, simple_loss=0.2634, pruned_loss=0.0503, over 12269.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2618, pruned_loss=0.04787, over 2743456.75 frames. ], batch size: 247, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:40:20,731 INFO [optim.py:368] (2/8) 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 06:40:36,241 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0450, 3.1599, 2.9270, 5.2428, 4.4928, 4.6626, 1.7542, 3.3627], device='cuda:2'), covar=tensor([0.1198, 0.0676, 0.1083, 0.0162, 0.0227, 0.0335, 0.1507, 0.0698], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0164, 0.0192, 0.0208, 0.0189, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 06:40:38,375 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:40:43,877 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1626, 3.9623, 4.2329, 4.3567, 4.4286, 3.9921, 4.2659, 4.4221], device='cuda:2'), covar=tensor([0.1700, 0.1230, 0.1379, 0.0733, 0.0626, 0.1447, 0.2367, 0.0709], device='cuda:2'), in_proj_covar=tensor([0.0577, 0.0718, 0.0842, 0.0728, 0.0548, 0.0572, 0.0588, 0.0673], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:40:58,699 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:06,967 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:08,822 INFO [train.py:904] (2/8) Epoch 16, batch 400, loss[loss=0.1635, simple_loss=0.2538, pruned_loss=0.03657, over 17035.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2606, pruned_loss=0.04703, over 2870357.80 frames. ], batch size: 50, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:42:16,089 INFO [train.py:904] (2/8) Epoch 16, batch 450, loss[loss=0.1538, simple_loss=0.2361, pruned_loss=0.03578, over 17060.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.259, pruned_loss=0.04616, over 2970951.91 frames. ], batch size: 41, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:42:36,231 INFO [optim.py:368] (2/8) 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 06:42:50,506 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:43:11,986 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-30 06:43:25,831 INFO [train.py:904] (2/8) Epoch 16, batch 500, loss[loss=0.1727, simple_loss=0.2703, pruned_loss=0.03759, over 16966.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.257, pruned_loss=0.04467, over 3052971.82 frames. ], batch size: 55, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:14,745 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:33,459 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:34,219 INFO [train.py:904] (2/8) Epoch 16, batch 550, loss[loss=0.1708, simple_loss=0.253, pruned_loss=0.04428, over 16826.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2565, pruned_loss=0.04409, over 3120126.83 frames. ], batch size: 42, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:55,542 INFO [optim.py:368] (2/8) 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 06:45:14,122 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5093, 3.2776, 2.6955, 2.1442, 2.2366, 2.2804, 3.3494, 3.0012], device='cuda:2'), covar=tensor([0.2691, 0.0759, 0.1707, 0.2599, 0.2605, 0.2039, 0.0527, 0.1523], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0262, 0.0293, 0.0292, 0.0279, 0.0237, 0.0277, 0.0313], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 06:45:31,124 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 06:45:46,473 INFO [train.py:904] (2/8) Epoch 16, batch 600, loss[loss=0.1561, simple_loss=0.2427, pruned_loss=0.03471, over 16247.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2559, pruned_loss=0.04376, over 3164090.32 frames. ], batch size: 36, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:46:24,478 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:46:38,782 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8926, 4.2927, 3.1802, 2.3332, 2.8273, 2.7135, 4.6589, 3.7496], device='cuda:2'), covar=tensor([0.2702, 0.0643, 0.1677, 0.2608, 0.2608, 0.1866, 0.0395, 0.1250], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0262, 0.0294, 0.0292, 0.0280, 0.0238, 0.0278, 0.0314], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 06:46:41,469 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 06:46:53,568 INFO [train.py:904] (2/8) Epoch 16, batch 650, loss[loss=0.1864, simple_loss=0.2763, pruned_loss=0.04828, over 17100.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2549, pruned_loss=0.04365, over 3198090.55 frames. ], batch size: 47, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:47:14,337 INFO [optim.py:368] (2/8) 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 06:47:30,632 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:30,704 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:43,485 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:52,079 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:48:00,889 INFO [train.py:904] (2/8) Epoch 16, batch 700, loss[loss=0.1984, simple_loss=0.2663, pruned_loss=0.06526, over 16885.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2557, pruned_loss=0.04393, over 3222137.98 frames. ], batch size: 109, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:48:12,806 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 06:48:37,710 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:49:10,743 INFO [train.py:904] (2/8) Epoch 16, batch 750, loss[loss=0.1766, simple_loss=0.2797, pruned_loss=0.03672, over 17053.00 frames. ], tot_loss[loss=0.172, simple_loss=0.256, pruned_loss=0.04396, over 3234341.75 frames. ], batch size: 50, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:49:31,099 INFO [optim.py:368] (2/8) 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 06:49:38,934 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 06:50:17,791 INFO [train.py:904] (2/8) Epoch 16, batch 800, loss[loss=0.1419, simple_loss=0.2278, pruned_loss=0.02806, over 16939.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2558, pruned_loss=0.04374, over 3255261.57 frames. ], batch size: 41, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:00,661 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:25,488 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153101.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:26,412 INFO [train.py:904] (2/8) Epoch 16, batch 850, loss[loss=0.1765, simple_loss=0.2494, pruned_loss=0.05182, over 16721.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.255, pruned_loss=0.04317, over 3264664.83 frames. ], batch size: 134, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:46,528 INFO [optim.py:368] (2/8) 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 06:51:46,975 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:49,964 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0173, 3.8992, 4.4401, 2.2476, 4.6447, 4.7694, 3.3338, 3.5659], device='cuda:2'), covar=tensor([0.0639, 0.0226, 0.0202, 0.1035, 0.0068, 0.0114, 0.0398, 0.0345], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0106, 0.0092, 0.0139, 0.0073, 0.0117, 0.0125, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 06:52:11,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1141, 4.7895, 5.1828, 5.3553, 5.5149, 4.8599, 5.5171, 5.4883], device='cuda:2'), covar=tensor([0.1701, 0.1377, 0.1579, 0.0645, 0.0497, 0.0835, 0.0489, 0.0558], device='cuda:2'), in_proj_covar=tensor([0.0596, 0.0740, 0.0877, 0.0749, 0.0565, 0.0587, 0.0606, 0.0700], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:52:32,207 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:35,391 INFO [train.py:904] (2/8) Epoch 16, batch 900, loss[loss=0.1663, simple_loss=0.2606, pruned_loss=0.036, over 17138.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2546, pruned_loss=0.04315, over 3279319.20 frames. ], batch size: 48, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:53:10,988 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:53:14,262 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3067, 4.2653, 4.6448, 4.6380, 4.6951, 4.3744, 4.3889, 4.2477], device='cuda:2'), covar=tensor([0.0342, 0.0659, 0.0424, 0.0469, 0.0474, 0.0415, 0.0816, 0.0573], device='cuda:2'), in_proj_covar=tensor([0.0374, 0.0406, 0.0395, 0.0378, 0.0445, 0.0419, 0.0510, 0.0333], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 06:53:43,175 INFO [train.py:904] (2/8) Epoch 16, batch 950, loss[loss=0.199, simple_loss=0.2655, pruned_loss=0.06623, over 16909.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2544, pruned_loss=0.04335, over 3295615.97 frames. ], batch size: 116, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:54:04,602 INFO [optim.py:368] (2/8) 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 06:54:33,125 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:42,589 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:53,048 INFO [train.py:904] (2/8) Epoch 16, batch 1000, loss[loss=0.1687, simple_loss=0.2627, pruned_loss=0.03739, over 16840.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2541, pruned_loss=0.04362, over 3300272.19 frames. ], batch size: 42, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:55:29,835 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153278.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:40,234 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:50,994 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:02,860 INFO [train.py:904] (2/8) Epoch 16, batch 1050, loss[loss=0.1934, simple_loss=0.2821, pruned_loss=0.0523, over 17042.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2546, pruned_loss=0.04413, over 3306104.95 frames. ], batch size: 53, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:56:24,059 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2639, 1.5785, 2.0370, 2.0793, 2.2511, 2.2565, 1.6837, 2.3225], device='cuda:2'), covar=tensor([0.0175, 0.0401, 0.0224, 0.0289, 0.0245, 0.0264, 0.0438, 0.0147], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0173, 0.0182, 0.0140, 0.0185, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 06:56:24,688 INFO [optim.py:368] (2/8) 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 06:56:55,490 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:55,577 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7522, 4.1499, 3.0121, 2.2696, 2.6471, 2.5337, 4.4240, 3.5578], device='cuda:2'), covar=tensor([0.2638, 0.0504, 0.1675, 0.2565, 0.2545, 0.1854, 0.0331, 0.1181], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0263, 0.0294, 0.0291, 0.0282, 0.0237, 0.0278, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 06:57:12,821 INFO [train.py:904] (2/8) Epoch 16, batch 1100, loss[loss=0.1574, simple_loss=0.2437, pruned_loss=0.03552, over 15533.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2539, pruned_loss=0.04402, over 3310045.75 frames. ], batch size: 190, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:57:53,702 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:21,563 INFO [train.py:904] (2/8) Epoch 16, batch 1150, loss[loss=0.1953, simple_loss=0.2694, pruned_loss=0.06057, over 16681.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.253, pruned_loss=0.04379, over 3304991.98 frames. ], batch size: 134, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:58:42,013 INFO [optim.py:368] (2/8) 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 06:58:46,822 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:58,552 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153430.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:59:27,882 INFO [train.py:904] (2/8) Epoch 16, batch 1200, loss[loss=0.1893, simple_loss=0.2497, pruned_loss=0.06447, over 16740.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2521, pruned_loss=0.04382, over 3309606.61 frames. ], batch size: 83, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 06:59:57,455 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:10,696 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:34,392 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9105, 4.5788, 4.7196, 5.1068, 5.2638, 4.7284, 5.3520, 5.2866], device='cuda:2'), covar=tensor([0.1802, 0.1609, 0.2586, 0.1056, 0.0807, 0.0955, 0.0710, 0.0830], device='cuda:2'), in_proj_covar=tensor([0.0611, 0.0759, 0.0899, 0.0770, 0.0578, 0.0603, 0.0618, 0.0712], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:00:37,161 INFO [train.py:904] (2/8) Epoch 16, batch 1250, loss[loss=0.2104, simple_loss=0.277, pruned_loss=0.07186, over 16464.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2529, pruned_loss=0.04419, over 3310671.33 frames. ], batch size: 75, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:00:57,397 INFO [optim.py:368] (2/8) 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 07:01:43,332 INFO [train.py:904] (2/8) Epoch 16, batch 1300, loss[loss=0.1735, simple_loss=0.2627, pruned_loss=0.0422, over 16704.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2528, pruned_loss=0.04384, over 3320511.18 frames. ], batch size: 57, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:02:24,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 07:02:44,874 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7912, 3.8278, 4.4214, 2.0994, 4.5055, 4.6639, 3.2742, 3.5800], device='cuda:2'), covar=tensor([0.0742, 0.0230, 0.0185, 0.1193, 0.0088, 0.0143, 0.0394, 0.0364], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 07:02:52,302 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 07:02:52,642 INFO [train.py:904] (2/8) Epoch 16, batch 1350, loss[loss=0.1608, simple_loss=0.2562, pruned_loss=0.03274, over 16665.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2533, pruned_loss=0.04343, over 3324786.67 frames. ], batch size: 57, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:03:12,844 INFO [optim.py:368] (2/8) 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 07:03:37,014 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:04:02,080 INFO [train.py:904] (2/8) Epoch 16, batch 1400, loss[loss=0.1536, simple_loss=0.2385, pruned_loss=0.03434, over 16812.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.253, pruned_loss=0.04298, over 3321764.01 frames. ], batch size: 102, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:05:12,109 INFO [train.py:904] (2/8) Epoch 16, batch 1450, loss[loss=0.1796, simple_loss=0.2422, pruned_loss=0.05846, over 16824.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2521, pruned_loss=0.04257, over 3327490.78 frames. ], batch size: 83, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:05:34,100 INFO [optim.py:368] (2/8) 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 07:06:11,389 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8733, 3.9966, 2.6854, 4.6484, 3.1243, 4.5803, 2.6243, 3.2610], device='cuda:2'), covar=tensor([0.0248, 0.0316, 0.1369, 0.0193, 0.0713, 0.0421, 0.1373, 0.0629], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0193, 0.0152, 0.0174, 0.0214, 0.0202, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:06:22,460 INFO [train.py:904] (2/8) Epoch 16, batch 1500, loss[loss=0.1889, simple_loss=0.2619, pruned_loss=0.05794, over 16891.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2519, pruned_loss=0.04306, over 3324632.73 frames. ], batch size: 116, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:06:50,917 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:06:55,365 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:06:59,021 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0451, 4.0542, 4.4623, 4.4516, 4.4743, 4.1232, 4.1620, 4.1261], device='cuda:2'), covar=tensor([0.0352, 0.0697, 0.0430, 0.0445, 0.0490, 0.0463, 0.0905, 0.0564], device='cuda:2'), in_proj_covar=tensor([0.0384, 0.0415, 0.0403, 0.0385, 0.0455, 0.0430, 0.0524, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 07:07:01,442 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0805, 4.2786, 3.3841, 2.2837, 2.8695, 2.6467, 4.7198, 3.7263], device='cuda:2'), covar=tensor([0.2403, 0.0740, 0.1473, 0.2692, 0.2740, 0.1836, 0.0350, 0.1276], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0264, 0.0293, 0.0292, 0.0283, 0.0237, 0.0279, 0.0316], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:07:09,854 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8581, 3.9035, 2.3066, 4.2855, 2.9868, 4.2249, 2.4617, 3.1781], device='cuda:2'), covar=tensor([0.0214, 0.0302, 0.1455, 0.0263, 0.0682, 0.0509, 0.1283, 0.0597], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0172, 0.0192, 0.0152, 0.0173, 0.0214, 0.0201, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:07:24,871 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 07:07:30,670 INFO [train.py:904] (2/8) Epoch 16, batch 1550, loss[loss=0.2043, simple_loss=0.2982, pruned_loss=0.05524, over 17074.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.254, pruned_loss=0.04415, over 3334437.27 frames. ], batch size: 55, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:07:52,823 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 07:07:53,743 INFO [optim.py:368] (2/8) 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 07:07:58,215 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:08:39,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0517, 4.3723, 3.2319, 2.3210, 2.8473, 2.6041, 4.7755, 3.7939], device='cuda:2'), covar=tensor([0.2434, 0.0593, 0.1595, 0.2671, 0.2620, 0.1851, 0.0335, 0.1157], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0263, 0.0293, 0.0291, 0.0282, 0.0237, 0.0279, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:08:40,385 INFO [train.py:904] (2/8) Epoch 16, batch 1600, loss[loss=0.1564, simple_loss=0.2463, pruned_loss=0.03318, over 17198.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.256, pruned_loss=0.0453, over 3325293.90 frames. ], batch size: 44, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:09:47,766 INFO [train.py:904] (2/8) Epoch 16, batch 1650, loss[loss=0.2, simple_loss=0.3066, pruned_loss=0.04669, over 17068.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2583, pruned_loss=0.04624, over 3313834.85 frames. ], batch size: 55, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:10:09,057 INFO [optim.py:368] (2/8) 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 07:10:09,529 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4272, 5.8680, 5.6044, 5.6296, 5.2097, 5.2211, 5.2172, 5.9762], device='cuda:2'), covar=tensor([0.1514, 0.1075, 0.1210, 0.0814, 0.1067, 0.0774, 0.1234, 0.0969], device='cuda:2'), in_proj_covar=tensor([0.0632, 0.0778, 0.0634, 0.0560, 0.0491, 0.0497, 0.0650, 0.0594], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:10:16,039 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:18,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9444, 4.8581, 4.7900, 4.2230, 4.8629, 1.9188, 4.6433, 4.6627], device='cuda:2'), covar=tensor([0.0116, 0.0096, 0.0187, 0.0405, 0.0097, 0.2501, 0.0148, 0.0181], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0138, 0.0186, 0.0170, 0.0159, 0.0198, 0.0174, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:10:32,340 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:46,554 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 07:10:56,044 INFO [train.py:904] (2/8) Epoch 16, batch 1700, loss[loss=0.1509, simple_loss=0.2426, pruned_loss=0.02961, over 17231.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2589, pruned_loss=0.04648, over 3324470.65 frames. ], batch size: 45, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:11:38,410 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:11:40,828 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:11:42,090 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5496, 3.5948, 3.2957, 3.0260, 3.2010, 3.5167, 3.3306, 3.2866], device='cuda:2'), covar=tensor([0.0568, 0.0562, 0.0282, 0.0262, 0.0508, 0.0411, 0.1088, 0.0448], device='cuda:2'), in_proj_covar=tensor([0.0283, 0.0391, 0.0332, 0.0321, 0.0349, 0.0369, 0.0228, 0.0399], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:12:09,295 INFO [train.py:904] (2/8) Epoch 16, batch 1750, loss[loss=0.1774, simple_loss=0.2659, pruned_loss=0.04444, over 16759.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2608, pruned_loss=0.04711, over 3301999.34 frames. ], batch size: 62, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:12:21,055 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5215, 4.5648, 4.7675, 4.5781, 4.5522, 5.2149, 4.7453, 4.3863], device='cuda:2'), covar=tensor([0.1718, 0.2085, 0.2325, 0.2233, 0.3084, 0.1274, 0.1551, 0.2518], device='cuda:2'), in_proj_covar=tensor([0.0396, 0.0568, 0.0617, 0.0474, 0.0637, 0.0649, 0.0483, 0.0629], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:12:21,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3244, 3.2042, 3.5304, 2.4095, 3.2129, 3.6561, 3.4477, 1.9661], device='cuda:2'), covar=tensor([0.0446, 0.0157, 0.0048, 0.0333, 0.0097, 0.0066, 0.0075, 0.0429], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0076, 0.0075, 0.0132, 0.0089, 0.0098, 0.0087, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:12:33,137 INFO [optim.py:368] (2/8) 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 07:13:15,087 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:13:18,924 INFO [train.py:904] (2/8) Epoch 16, batch 1800, loss[loss=0.2105, simple_loss=0.2815, pruned_loss=0.06974, over 16755.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2625, pruned_loss=0.04723, over 3306146.35 frames. ], batch size: 124, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:13:52,577 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:13:54,189 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:03,063 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:06,330 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3958, 3.3386, 3.4390, 3.4879, 3.5508, 3.3075, 3.4720, 3.6077], device='cuda:2'), covar=tensor([0.1078, 0.0839, 0.0933, 0.0551, 0.0583, 0.2067, 0.1266, 0.0702], device='cuda:2'), in_proj_covar=tensor([0.0613, 0.0761, 0.0902, 0.0770, 0.0577, 0.0603, 0.0615, 0.0714], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:14:28,054 INFO [train.py:904] (2/8) Epoch 16, batch 1850, loss[loss=0.1661, simple_loss=0.2628, pruned_loss=0.03469, over 17142.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2624, pruned_loss=0.04655, over 3307871.35 frames. ], batch size: 47, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:14:38,685 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:14:39,029 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-30 07:14:50,218 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:15,176 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:15:21,213 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2635, 3.2295, 1.9920, 3.4400, 2.5101, 3.4355, 2.0993, 2.6044], device='cuda:2'), covar=tensor([0.0248, 0.0377, 0.1469, 0.0255, 0.0788, 0.0659, 0.1352, 0.0699], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0174, 0.0195, 0.0154, 0.0174, 0.0216, 0.0203, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:15:25,826 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:33,361 INFO [train.py:904] (2/8) Epoch 16, batch 1900, loss[loss=0.1825, simple_loss=0.2553, pruned_loss=0.0549, over 16933.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2613, pruned_loss=0.04551, over 3315574.91 frames. ], batch size: 109, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:16:08,895 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-04-30 07:16:44,657 INFO [train.py:904] (2/8) Epoch 16, batch 1950, loss[loss=0.1629, simple_loss=0.2496, pruned_loss=0.03814, over 16493.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2611, pruned_loss=0.04498, over 3322266.97 frames. ], batch size: 68, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:16:49,566 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1033, 5.1248, 5.6254, 5.5928, 5.6216, 5.2190, 5.1552, 4.9126], device='cuda:2'), covar=tensor([0.0326, 0.0454, 0.0301, 0.0408, 0.0447, 0.0355, 0.0931, 0.0424], device='cuda:2'), in_proj_covar=tensor([0.0383, 0.0412, 0.0400, 0.0382, 0.0451, 0.0429, 0.0523, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 07:16:50,686 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7111, 2.9465, 3.0532, 2.0210, 2.6629, 2.1405, 3.2507, 3.2557], device='cuda:2'), covar=tensor([0.0256, 0.0839, 0.0563, 0.1809, 0.0848, 0.0976, 0.0576, 0.0783], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0157, 0.0163, 0.0150, 0.0141, 0.0127, 0.0142, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:17:04,523 INFO [optim.py:368] (2/8) 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 07:17:10,546 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:17:51,482 INFO [train.py:904] (2/8) Epoch 16, batch 2000, loss[loss=0.1638, simple_loss=0.2638, pruned_loss=0.03192, over 17278.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2599, pruned_loss=0.04436, over 3327767.73 frames. ], batch size: 52, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:18:27,771 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:33,382 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:18:38,737 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:44,511 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:59,383 INFO [train.py:904] (2/8) Epoch 16, batch 2050, loss[loss=0.2007, simple_loss=0.2881, pruned_loss=0.05665, over 16721.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.26, pruned_loss=0.0452, over 3318111.24 frames. ], batch size: 83, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:19:19,612 INFO [optim.py:368] (2/8) 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 07:19:46,267 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:19:50,663 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 07:20:01,076 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:04,571 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8529, 2.4729, 1.9645, 2.2873, 2.8413, 2.6585, 2.9767, 2.9679], device='cuda:2'), covar=tensor([0.0170, 0.0325, 0.0457, 0.0432, 0.0197, 0.0292, 0.0194, 0.0255], device='cuda:2'), in_proj_covar=tensor([0.0187, 0.0225, 0.0218, 0.0218, 0.0228, 0.0229, 0.0232, 0.0223], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:20:06,433 INFO [train.py:904] (2/8) Epoch 16, batch 2100, loss[loss=0.1523, simple_loss=0.2409, pruned_loss=0.0318, over 16857.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2597, pruned_loss=0.04531, over 3322955.32 frames. ], batch size: 42, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:20:06,904 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:57,938 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2241, 2.1076, 2.2084, 3.9560, 2.1792, 2.5002, 2.1866, 2.3138], device='cuda:2'), covar=tensor([0.1275, 0.3405, 0.2714, 0.0537, 0.3581, 0.2372, 0.3302, 0.3175], device='cuda:2'), in_proj_covar=tensor([0.0383, 0.0420, 0.0352, 0.0323, 0.0424, 0.0483, 0.0389, 0.0493], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:21:09,595 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154398.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:21:14,694 INFO [train.py:904] (2/8) Epoch 16, batch 2150, loss[loss=0.19, simple_loss=0.2736, pruned_loss=0.05313, over 16547.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2611, pruned_loss=0.04564, over 3325700.48 frames. ], batch size: 75, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:21:18,544 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:21:36,523 INFO [optim.py:368] (2/8) 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 07:21:57,835 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:22:07,247 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:22:25,079 INFO [train.py:904] (2/8) Epoch 16, batch 2200, loss[loss=0.1727, simple_loss=0.2698, pruned_loss=0.03783, over 16776.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2611, pruned_loss=0.04547, over 3316953.95 frames. ], batch size: 57, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:22:41,608 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-30 07:22:55,698 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:23:00,727 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-30 07:23:32,242 INFO [train.py:904] (2/8) Epoch 16, batch 2250, loss[loss=0.1839, simple_loss=0.2549, pruned_loss=0.05641, over 16412.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2618, pruned_loss=0.04604, over 3316019.35 frames. ], batch size: 75, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:23:38,513 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 07:23:55,133 INFO [optim.py:368] (2/8) 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 07:24:18,308 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:24:40,021 INFO [train.py:904] (2/8) Epoch 16, batch 2300, loss[loss=0.153, simple_loss=0.2439, pruned_loss=0.03111, over 17229.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2619, pruned_loss=0.04585, over 3327718.05 frames. ], batch size: 45, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:24:55,071 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-30 07:25:16,140 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:25:16,288 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2855, 5.1915, 5.0838, 4.6359, 4.6775, 5.1624, 5.1672, 4.7750], device='cuda:2'), covar=tensor([0.0573, 0.0469, 0.0296, 0.0322, 0.1109, 0.0413, 0.0287, 0.0770], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0393, 0.0336, 0.0323, 0.0352, 0.0372, 0.0230, 0.0403], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:25:20,035 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:25:51,248 INFO [train.py:904] (2/8) Epoch 16, batch 2350, loss[loss=0.1838, simple_loss=0.2664, pruned_loss=0.05058, over 16505.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2637, pruned_loss=0.04709, over 3319812.40 frames. ], batch size: 68, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:26:05,957 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0332, 4.3468, 3.0782, 2.3159, 2.7672, 2.5826, 4.7770, 3.8354], device='cuda:2'), covar=tensor([0.2533, 0.0616, 0.1687, 0.2715, 0.2969, 0.1953, 0.0325, 0.1164], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0267, 0.0295, 0.0293, 0.0286, 0.0239, 0.0281, 0.0320], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:26:11,866 INFO [optim.py:368] (2/8) 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 07:26:25,329 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:26,686 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:45,778 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:52,355 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 2400, loss[loss=0.1483, simple_loss=0.241, pruned_loss=0.02777, over 16851.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2639, pruned_loss=0.04697, over 3309381.53 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:26:59,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7913, 4.5394, 4.4645, 4.9312, 5.1421, 4.5661, 5.0925, 5.1414], device='cuda:2'), covar=tensor([0.1561, 0.1267, 0.2300, 0.1020, 0.0843, 0.1188, 0.1058, 0.0884], device='cuda:2'), in_proj_covar=tensor([0.0613, 0.0760, 0.0902, 0.0769, 0.0579, 0.0603, 0.0615, 0.0714], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:27:07,045 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4854, 5.9067, 5.6942, 5.7419, 5.3581, 5.2552, 5.3225, 6.0479], device='cuda:2'), covar=tensor([0.1390, 0.0947, 0.0905, 0.0735, 0.0925, 0.0701, 0.1007, 0.0808], device='cuda:2'), in_proj_covar=tensor([0.0639, 0.0789, 0.0643, 0.0568, 0.0497, 0.0505, 0.0658, 0.0603], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:27:50,388 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:27:55,502 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:27:55,933 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-04-30 07:28:06,455 INFO [train.py:904] (2/8) Epoch 16, batch 2450, loss[loss=0.1855, simple_loss=0.2736, pruned_loss=0.04877, over 16432.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2647, pruned_loss=0.04673, over 3317747.69 frames. ], batch size: 68, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:28:11,573 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:28:28,918 INFO [optim.py:368] (2/8) 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 07:28:31,839 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:49,586 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:59,950 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:15,868 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 07:29:16,067 INFO [train.py:904] (2/8) Epoch 16, batch 2500, loss[loss=0.1691, simple_loss=0.2678, pruned_loss=0.03518, over 17134.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2649, pruned_loss=0.04672, over 3322893.36 frames. ], batch size: 48, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:29:18,086 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:29:53,787 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:55,019 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:24,146 INFO [train.py:904] (2/8) Epoch 16, batch 2550, loss[loss=0.1756, simple_loss=0.2667, pruned_loss=0.04222, over 17201.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2637, pruned_loss=0.04618, over 3328115.18 frames. ], batch size: 44, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:30:47,016 INFO [optim.py:368] (2/8) 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 07:31:02,557 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:31:06,900 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 07:31:32,684 INFO [train.py:904] (2/8) Epoch 16, batch 2600, loss[loss=0.1641, simple_loss=0.247, pruned_loss=0.04065, over 16728.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2638, pruned_loss=0.04555, over 3331292.54 frames. ], batch size: 89, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:32:07,859 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:32:40,410 INFO [train.py:904] (2/8) Epoch 16, batch 2650, loss[loss=0.1347, simple_loss=0.2255, pruned_loss=0.02194, over 16815.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2643, pruned_loss=0.04564, over 3335184.67 frames. ], batch size: 39, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:32:44,954 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5753, 2.6402, 2.4486, 4.2957, 3.1720, 4.0476, 1.6856, 2.7399], device='cuda:2'), covar=tensor([0.1729, 0.0938, 0.1419, 0.0219, 0.0362, 0.0451, 0.1891, 0.1111], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0167, 0.0188, 0.0173, 0.0201, 0.0215, 0.0191, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:32:49,772 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1863, 4.1562, 4.1821, 3.1805, 4.1953, 1.5921, 3.8647, 3.6128], device='cuda:2'), covar=tensor([0.0223, 0.0178, 0.0219, 0.0645, 0.0146, 0.3489, 0.0224, 0.0435], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0142, 0.0191, 0.0175, 0.0162, 0.0202, 0.0179, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:33:01,394 INFO [optim.py:368] (2/8) 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 07:33:12,794 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:14,055 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6347, 4.4534, 4.6825, 4.8289, 5.0005, 4.4935, 4.9773, 4.9992], device='cuda:2'), covar=tensor([0.1594, 0.1216, 0.1520, 0.0725, 0.0546, 0.0996, 0.0716, 0.0583], device='cuda:2'), in_proj_covar=tensor([0.0619, 0.0769, 0.0913, 0.0777, 0.0584, 0.0615, 0.0619, 0.0724], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:33:35,906 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:43,497 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 2700, loss[loss=0.1658, simple_loss=0.2574, pruned_loss=0.03716, over 17226.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.264, pruned_loss=0.0454, over 3337734.99 frames. ], batch size: 45, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:33:50,415 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2568, 4.5513, 4.8516, 4.7994, 4.8364, 4.5187, 4.1591, 4.2693], device='cuda:2'), covar=tensor([0.0656, 0.0754, 0.0571, 0.0686, 0.0807, 0.0639, 0.1616, 0.0740], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0416, 0.0405, 0.0383, 0.0457, 0.0432, 0.0527, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 07:34:21,624 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:32,700 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:41,399 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:45,803 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:34:47,944 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154995.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:56,797 INFO [train.py:904] (2/8) Epoch 16, batch 2750, loss[loss=0.1956, simple_loss=0.2674, pruned_loss=0.06188, over 16711.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2638, pruned_loss=0.04499, over 3338711.89 frames. ], batch size: 124, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:34:59,910 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2267, 5.8453, 5.9058, 5.6071, 5.7204, 6.2020, 5.7630, 5.4597], device='cuda:2'), covar=tensor([0.0872, 0.1581, 0.1946, 0.2035, 0.2368, 0.1004, 0.1352, 0.2186], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0568, 0.0616, 0.0480, 0.0637, 0.0648, 0.0483, 0.0633], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:35:18,259 INFO [optim.py:368] (2/8) 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 07:35:46,096 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:35:50,510 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:36:04,755 INFO [train.py:904] (2/8) Epoch 16, batch 2800, loss[loss=0.1846, simple_loss=0.2799, pruned_loss=0.04462, over 17056.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2644, pruned_loss=0.04549, over 3339398.04 frames. ], batch size: 53, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:36:28,422 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:57,806 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0235, 3.2253, 3.1872, 2.1699, 2.7727, 2.2367, 3.5487, 3.5011], device='cuda:2'), covar=tensor([0.0253, 0.0852, 0.0655, 0.1733, 0.0886, 0.1023, 0.0521, 0.0850], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0155, 0.0161, 0.0147, 0.0139, 0.0125, 0.0140, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:37:15,725 INFO [train.py:904] (2/8) Epoch 16, batch 2850, loss[loss=0.2183, simple_loss=0.2965, pruned_loss=0.07001, over 15261.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2635, pruned_loss=0.04552, over 3329523.37 frames. ], batch size: 190, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:37:16,029 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7431, 5.0240, 4.7877, 4.8065, 4.5803, 4.5234, 4.4820, 5.0964], device='cuda:2'), covar=tensor([0.1106, 0.0888, 0.1055, 0.0784, 0.0843, 0.1089, 0.1091, 0.0867], device='cuda:2'), in_proj_covar=tensor([0.0644, 0.0793, 0.0647, 0.0574, 0.0501, 0.0508, 0.0661, 0.0611], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:37:26,148 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 07:37:36,422 INFO [optim.py:368] (2/8) 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 07:37:54,357 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:37:54,506 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:38:23,947 INFO [train.py:904] (2/8) Epoch 16, batch 2900, loss[loss=0.1994, simple_loss=0.2738, pruned_loss=0.0625, over 15645.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2625, pruned_loss=0.04555, over 3329479.46 frames. ], batch size: 190, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:00,496 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:39:33,220 INFO [train.py:904] (2/8) Epoch 16, batch 2950, loss[loss=0.2399, simple_loss=0.3074, pruned_loss=0.08621, over 11485.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.262, pruned_loss=0.04625, over 3318105.99 frames. ], batch size: 248, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:54,115 INFO [optim.py:368] (2/8) 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 07:40:40,768 INFO [train.py:904] (2/8) Epoch 16, batch 3000, loss[loss=0.1901, simple_loss=0.2777, pruned_loss=0.05125, over 16715.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2629, pruned_loss=0.04663, over 3325916.17 frames. ], batch size: 62, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:40:40,768 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 07:40:49,849 INFO [train.py:938] (2/8) Epoch 16, validation: loss=0.1368, simple_loss=0.2429, pruned_loss=0.01541, over 944034.00 frames. 2023-04-30 07:40:49,850 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 07:41:02,233 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6981, 2.7024, 2.4485, 2.4978, 2.9756, 2.7945, 3.4236, 3.2494], device='cuda:2'), covar=tensor([0.0138, 0.0331, 0.0405, 0.0415, 0.0252, 0.0349, 0.0207, 0.0258], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0227, 0.0219, 0.0221, 0.0230, 0.0231, 0.0236, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:41:34,181 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:41:59,649 INFO [train.py:904] (2/8) Epoch 16, batch 3050, loss[loss=0.1882, simple_loss=0.2599, pruned_loss=0.05826, over 16449.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2615, pruned_loss=0.04626, over 3319243.89 frames. ], batch size: 75, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:42:21,040 INFO [optim.py:368] (2/8) 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 07:42:23,344 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2301, 3.4136, 3.4776, 2.1660, 3.0190, 2.4127, 3.5728, 3.6722], device='cuda:2'), covar=tensor([0.0227, 0.0813, 0.0561, 0.1846, 0.0795, 0.0956, 0.0559, 0.0876], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0156, 0.0161, 0.0148, 0.0139, 0.0126, 0.0141, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:42:42,356 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:42,370 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:43:10,099 INFO [train.py:904] (2/8) Epoch 16, batch 3100, loss[loss=0.1744, simple_loss=0.2487, pruned_loss=0.05004, over 16829.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2608, pruned_loss=0.04633, over 3318543.47 frames. ], batch size: 96, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:43:42,803 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:43:59,322 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8636, 4.0257, 3.1749, 2.3510, 2.7847, 2.6483, 4.3403, 3.6560], device='cuda:2'), covar=tensor([0.2446, 0.0603, 0.1497, 0.2411, 0.2300, 0.1744, 0.0371, 0.1073], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0264, 0.0293, 0.0293, 0.0287, 0.0239, 0.0279, 0.0320], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:44:17,694 INFO [train.py:904] (2/8) Epoch 16, batch 3150, loss[loss=0.192, simple_loss=0.2641, pruned_loss=0.05996, over 16808.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2599, pruned_loss=0.0459, over 3309004.65 frames. ], batch size: 124, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:44:28,033 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1578, 5.5692, 5.7387, 5.4845, 5.5283, 6.1061, 5.6169, 5.3346], device='cuda:2'), covar=tensor([0.0780, 0.1857, 0.1735, 0.1969, 0.2430, 0.0903, 0.1312, 0.2151], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0572, 0.0618, 0.0483, 0.0641, 0.0649, 0.0485, 0.0636], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:44:39,887 INFO [optim.py:368] (2/8) 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 07:44:40,800 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 07:44:47,892 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:49,113 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:45:27,253 INFO [train.py:904] (2/8) Epoch 16, batch 3200, loss[loss=0.2151, simple_loss=0.2887, pruned_loss=0.07071, over 12316.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2583, pruned_loss=0.04496, over 3304699.11 frames. ], batch size: 246, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:45:33,605 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:46:31,709 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9506, 2.0477, 2.4953, 2.8798, 2.6846, 3.3864, 2.1357, 3.2653], device='cuda:2'), covar=tensor([0.0208, 0.0434, 0.0294, 0.0273, 0.0299, 0.0154, 0.0441, 0.0181], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0186, 0.0172, 0.0175, 0.0187, 0.0143, 0.0185, 0.0136], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:46:36,135 INFO [train.py:904] (2/8) Epoch 16, batch 3250, loss[loss=0.1629, simple_loss=0.2419, pruned_loss=0.042, over 16768.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2588, pruned_loss=0.04519, over 3300828.87 frames. ], batch size: 83, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:46:52,822 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1042, 2.0809, 2.2846, 3.8096, 2.1327, 2.4234, 2.2006, 2.2541], device='cuda:2'), covar=tensor([0.1357, 0.3501, 0.2634, 0.0577, 0.3782, 0.2462, 0.3458, 0.2986], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0425, 0.0355, 0.0326, 0.0428, 0.0490, 0.0393, 0.0497], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:46:58,461 INFO [optim.py:368] (2/8) 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 07:46:58,986 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:47:33,708 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9507, 4.3695, 2.8389, 2.3923, 2.8015, 2.3040, 4.5324, 3.6675], device='cuda:2'), covar=tensor([0.2807, 0.0659, 0.2019, 0.2833, 0.2900, 0.2225, 0.0520, 0.1330], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0267, 0.0295, 0.0296, 0.0290, 0.0241, 0.0282, 0.0323], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:47:45,922 INFO [train.py:904] (2/8) Epoch 16, batch 3300, loss[loss=0.1717, simple_loss=0.2565, pruned_loss=0.0435, over 16503.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2601, pruned_loss=0.04543, over 3305384.40 frames. ], batch size: 75, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:47:57,480 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8236, 2.6602, 2.6434, 1.9667, 2.5827, 2.7846, 2.5821, 1.9138], device='cuda:2'), covar=tensor([0.0431, 0.0084, 0.0069, 0.0339, 0.0121, 0.0094, 0.0107, 0.0367], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0077, 0.0076, 0.0131, 0.0090, 0.0099, 0.0088, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:48:56,952 INFO [train.py:904] (2/8) Epoch 16, batch 3350, loss[loss=0.1707, simple_loss=0.2448, pruned_loss=0.04828, over 16716.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2611, pruned_loss=0.04571, over 3307680.88 frames. ], batch size: 124, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:49:01,210 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:49:19,483 INFO [optim.py:368] (2/8) 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 07:49:39,840 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:08,446 INFO [train.py:904] (2/8) Epoch 16, batch 3400, loss[loss=0.1513, simple_loss=0.2516, pruned_loss=0.0255, over 17237.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2615, pruned_loss=0.0459, over 3309288.20 frames. ], batch size: 52, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:50:16,701 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:23,361 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2034, 3.9605, 4.0488, 4.3812, 4.4360, 4.0844, 4.2611, 4.4837], device='cuda:2'), covar=tensor([0.1483, 0.1411, 0.1880, 0.0832, 0.0767, 0.1412, 0.3068, 0.0840], device='cuda:2'), in_proj_covar=tensor([0.0631, 0.0782, 0.0923, 0.0795, 0.0593, 0.0622, 0.0627, 0.0735], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:50:27,976 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:40,676 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7794, 3.1293, 3.1123, 2.0609, 2.7749, 2.2734, 3.2957, 3.2753], device='cuda:2'), covar=tensor([0.0266, 0.0819, 0.0578, 0.1667, 0.0834, 0.0898, 0.0602, 0.0842], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0157, 0.0162, 0.0149, 0.0140, 0.0126, 0.0142, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:50:49,342 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:51:10,652 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.40 vs. limit=5.0 2023-04-30 07:51:19,270 INFO [train.py:904] (2/8) Epoch 16, batch 3450, loss[loss=0.1488, simple_loss=0.2347, pruned_loss=0.03148, over 16749.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2594, pruned_loss=0.04507, over 3315922.02 frames. ], batch size: 39, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:51:41,374 INFO [optim.py:368] (2/8) 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 07:51:41,767 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:51:52,550 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:52:05,943 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5422, 2.1462, 2.2499, 4.3202, 2.1201, 2.4987, 2.2994, 2.3315], device='cuda:2'), covar=tensor([0.1148, 0.3779, 0.2911, 0.0482, 0.4398, 0.2789, 0.3431, 0.3989], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0425, 0.0355, 0.0327, 0.0428, 0.0490, 0.0393, 0.0497], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:52:28,785 INFO [train.py:904] (2/8) Epoch 16, batch 3500, loss[loss=0.1445, simple_loss=0.2239, pruned_loss=0.03253, over 16750.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2592, pruned_loss=0.04471, over 3314896.39 frames. ], batch size: 39, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:52:58,254 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:14,798 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:40,576 INFO [train.py:904] (2/8) Epoch 16, batch 3550, loss[loss=0.1876, simple_loss=0.2667, pruned_loss=0.05424, over 15359.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2587, pruned_loss=0.04489, over 3304498.36 frames. ], batch size: 190, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:53:56,148 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:54:04,791 INFO [optim.py:368] (2/8) 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 07:54:15,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0924, 3.2775, 3.3076, 2.1150, 2.8308, 2.2955, 3.5935, 3.4999], device='cuda:2'), covar=tensor([0.0227, 0.0856, 0.0562, 0.1760, 0.0811, 0.0979, 0.0503, 0.0853], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0159, 0.0164, 0.0150, 0.0141, 0.0127, 0.0143, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:54:42,515 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:54:52,021 INFO [train.py:904] (2/8) Epoch 16, batch 3600, loss[loss=0.1719, simple_loss=0.2449, pruned_loss=0.04948, over 16798.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2579, pruned_loss=0.04469, over 3308239.67 frames. ], batch size: 96, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:03,881 INFO [train.py:904] (2/8) Epoch 16, batch 3650, loss[loss=0.1831, simple_loss=0.2523, pruned_loss=0.05696, over 16881.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2573, pruned_loss=0.04531, over 3308529.46 frames. ], batch size: 109, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:28,564 INFO [optim.py:368] (2/8) 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 07:56:29,319 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 07:57:17,739 INFO [train.py:904] (2/8) Epoch 16, batch 3700, loss[loss=0.1649, simple_loss=0.2449, pruned_loss=0.04245, over 16682.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2566, pruned_loss=0.04685, over 3293546.34 frames. ], batch size: 89, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:57:31,973 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:57:38,324 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3422, 4.2686, 4.2762, 4.0793, 4.1097, 4.3444, 4.0424, 4.1372], device='cuda:2'), covar=tensor([0.0551, 0.0606, 0.0250, 0.0237, 0.0584, 0.0449, 0.0719, 0.0501], device='cuda:2'), in_proj_covar=tensor([0.0287, 0.0397, 0.0337, 0.0326, 0.0353, 0.0377, 0.0230, 0.0406], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 07:58:12,333 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7667, 2.5538, 2.3860, 3.4338, 2.8264, 3.6465, 1.5152, 2.6728], device='cuda:2'), covar=tensor([0.1312, 0.0703, 0.1105, 0.0172, 0.0158, 0.0359, 0.1552, 0.0832], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0168, 0.0187, 0.0174, 0.0202, 0.0214, 0.0189, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:58:24,908 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2422, 3.3072, 1.9874, 3.4165, 2.5874, 3.4764, 2.0772, 2.6698], device='cuda:2'), covar=tensor([0.0297, 0.0403, 0.1596, 0.0300, 0.0789, 0.0626, 0.1464, 0.0703], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0175, 0.0193, 0.0156, 0.0173, 0.0217, 0.0202, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:58:29,322 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5488, 4.6455, 4.9957, 4.9863, 4.9975, 4.6270, 4.6647, 4.3703], device='cuda:2'), covar=tensor([0.0337, 0.0721, 0.0421, 0.0408, 0.0448, 0.0423, 0.0855, 0.0600], device='cuda:2'), in_proj_covar=tensor([0.0393, 0.0424, 0.0411, 0.0389, 0.0466, 0.0439, 0.0533, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 07:58:36,928 INFO [train.py:904] (2/8) Epoch 16, batch 3750, loss[loss=0.1658, simple_loss=0.2407, pruned_loss=0.0454, over 16837.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2568, pruned_loss=0.04822, over 3269320.51 frames. ], batch size: 102, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:58:37,292 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5628, 4.3350, 4.4311, 4.7465, 4.8419, 4.4393, 4.8269, 4.8613], device='cuda:2'), covar=tensor([0.1721, 0.1603, 0.2084, 0.1078, 0.1057, 0.1133, 0.1716, 0.1450], device='cuda:2'), in_proj_covar=tensor([0.0620, 0.0770, 0.0907, 0.0781, 0.0582, 0.0612, 0.0615, 0.0724], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 07:58:47,507 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9466, 2.9712, 2.5847, 4.4347, 3.7053, 4.2411, 1.6990, 3.0127], device='cuda:2'), covar=tensor([0.1231, 0.0628, 0.1090, 0.0167, 0.0230, 0.0344, 0.1449, 0.0813], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0167, 0.0187, 0.0174, 0.0201, 0.0213, 0.0189, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 07:58:53,073 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:59:02,683 INFO [optim.py:368] (2/8) 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 07:59:31,684 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 07:59:51,246 INFO [train.py:904] (2/8) Epoch 16, batch 3800, loss[loss=0.1673, simple_loss=0.2477, pruned_loss=0.04347, over 16907.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2575, pruned_loss=0.04899, over 3275815.31 frames. ], batch size: 96, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 08:00:17,831 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:03,112 INFO [train.py:904] (2/8) Epoch 16, batch 3850, loss[loss=0.2189, simple_loss=0.2899, pruned_loss=0.07397, over 12552.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2578, pruned_loss=0.04994, over 3273990.39 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:01:20,107 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156113.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:20,221 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9837, 3.1813, 3.3550, 2.2750, 3.0041, 3.4606, 3.1640, 1.9493], device='cuda:2'), covar=tensor([0.0463, 0.0104, 0.0054, 0.0346, 0.0107, 0.0087, 0.0079, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0076, 0.0076, 0.0130, 0.0089, 0.0099, 0.0087, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 08:01:28,704 INFO [optim.py:368] (2/8) 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 08:01:47,405 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:01:59,095 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 3900, loss[loss=0.1546, simple_loss=0.2324, pruned_loss=0.03841, over 16358.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2575, pruned_loss=0.05043, over 3272280.73 frames. ], batch size: 75, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:02:29,442 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:09,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6645, 2.6794, 1.8848, 2.7312, 2.2347, 2.8244, 2.0066, 2.3148], device='cuda:2'), covar=tensor([0.0279, 0.0362, 0.1362, 0.0209, 0.0640, 0.0330, 0.1229, 0.0662], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0175, 0.0192, 0.0154, 0.0173, 0.0216, 0.0201, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:03:26,070 INFO [train.py:904] (2/8) Epoch 16, batch 3950, loss[loss=0.2303, simple_loss=0.2952, pruned_loss=0.08269, over 12788.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2573, pruned_loss=0.05108, over 3274663.92 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:03:47,391 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:50,337 INFO [optim.py:368] (2/8) 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 08:04:09,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8791, 2.8516, 2.6916, 5.0572, 3.8723, 4.2683, 2.1816, 3.2509], device='cuda:2'), covar=tensor([0.1388, 0.0891, 0.1285, 0.0133, 0.0419, 0.0386, 0.1435, 0.0869], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0169, 0.0187, 0.0174, 0.0202, 0.0214, 0.0189, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:04:37,101 INFO [train.py:904] (2/8) Epoch 16, batch 4000, loss[loss=0.2012, simple_loss=0.2799, pruned_loss=0.06123, over 12518.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2572, pruned_loss=0.05114, over 3270737.12 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:04:49,882 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:12,202 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:31,064 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:49,244 INFO [train.py:904] (2/8) Epoch 16, batch 4050, loss[loss=0.1715, simple_loss=0.2573, pruned_loss=0.04285, over 16313.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2575, pruned_loss=0.05023, over 3260243.76 frames. ], batch size: 165, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:05:59,640 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:04,455 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:14,228 INFO [optim.py:368] (2/8) 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 08:07:01,325 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:01,987 INFO [train.py:904] (2/8) Epoch 16, batch 4100, loss[loss=0.1751, simple_loss=0.2619, pruned_loss=0.04415, over 16423.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2591, pruned_loss=0.04965, over 3264777.68 frames. ], batch size: 68, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:07:15,154 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:36,949 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 08:08:15,973 INFO [train.py:904] (2/8) Epoch 16, batch 4150, loss[loss=0.2196, simple_loss=0.3077, pruned_loss=0.06575, over 16831.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2662, pruned_loss=0.05217, over 3237744.16 frames. ], batch size: 116, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:08:40,341 INFO [optim.py:368] (2/8) 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 08:08:52,135 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:09:13,256 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156440.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:09:30,684 INFO [train.py:904] (2/8) Epoch 16, batch 4200, loss[loss=0.23, simple_loss=0.3095, pruned_loss=0.07526, over 11150.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2732, pruned_loss=0.05385, over 3195923.96 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:10:24,398 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:10:43,961 INFO [train.py:904] (2/8) Epoch 16, batch 4250, loss[loss=0.1616, simple_loss=0.2562, pruned_loss=0.03351, over 16787.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2767, pruned_loss=0.05335, over 3204268.01 frames. ], batch size: 124, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:11:09,162 INFO [optim.py:368] (2/8) 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 08:11:10,702 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4988, 4.8331, 4.3442, 4.7139, 4.3581, 4.3062, 4.3943, 4.8873], device='cuda:2'), covar=tensor([0.2104, 0.1401, 0.2084, 0.1195, 0.1508, 0.1817, 0.2101, 0.1554], device='cuda:2'), in_proj_covar=tensor([0.0617, 0.0764, 0.0623, 0.0552, 0.0479, 0.0491, 0.0635, 0.0586], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:11:49,386 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-30 08:11:53,965 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-04-30 08:11:56,451 INFO [train.py:904] (2/8) Epoch 16, batch 4300, loss[loss=0.1866, simple_loss=0.2741, pruned_loss=0.0496, over 16353.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2784, pruned_loss=0.05276, over 3207316.46 frames. ], batch size: 35, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:12:27,168 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:13:07,781 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 08:13:09,402 INFO [train.py:904] (2/8) Epoch 16, batch 4350, loss[loss=0.1998, simple_loss=0.287, pruned_loss=0.05631, over 16719.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2812, pruned_loss=0.05377, over 3208657.98 frames. ], batch size: 124, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:13:34,596 INFO [optim.py:368] (2/8) 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 08:14:14,560 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 4400, loss[loss=0.2102, simple_loss=0.296, pruned_loss=0.06225, over 15460.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.283, pruned_loss=0.05465, over 3204279.71 frames. ], batch size: 190, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:06,762 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4215, 1.5553, 2.0755, 2.3553, 2.3352, 2.6950, 1.7309, 2.6261], device='cuda:2'), covar=tensor([0.0176, 0.0472, 0.0255, 0.0267, 0.0261, 0.0136, 0.0436, 0.0122], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0174, 0.0185, 0.0140, 0.0183, 0.0134], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:15:32,099 INFO [train.py:904] (2/8) Epoch 16, batch 4450, loss[loss=0.21, simple_loss=0.3002, pruned_loss=0.05983, over 16888.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2869, pruned_loss=0.056, over 3214897.56 frames. ], batch size: 109, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:57,564 INFO [optim.py:368] (2/8) 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 08:16:04,243 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5732, 3.5903, 2.1572, 4.1720, 2.7236, 4.1456, 2.3401, 2.8291], device='cuda:2'), covar=tensor([0.0276, 0.0405, 0.1698, 0.0124, 0.0906, 0.0425, 0.1495, 0.0783], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0190, 0.0149, 0.0172, 0.0212, 0.0199, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:16:09,709 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:16:41,428 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:16:45,639 INFO [train.py:904] (2/8) Epoch 16, batch 4500, loss[loss=0.2019, simple_loss=0.2889, pruned_loss=0.05749, over 16634.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2875, pruned_loss=0.0565, over 3228926.26 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:16:47,451 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6634, 2.2185, 1.8240, 1.9625, 2.5368, 2.1304, 2.5544, 2.7018], device='cuda:2'), covar=tensor([0.0136, 0.0353, 0.0455, 0.0408, 0.0210, 0.0349, 0.0195, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0217, 0.0210, 0.0211, 0.0220, 0.0221, 0.0224, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:16:49,109 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-04-30 08:17:16,183 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:17:19,757 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 08:17:54,747 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6287, 2.7383, 2.6220, 4.1543, 3.2055, 3.9744, 1.5728, 3.0101], device='cuda:2'), covar=tensor([0.1319, 0.0695, 0.1062, 0.0126, 0.0246, 0.0321, 0.1540, 0.0732], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0168, 0.0187, 0.0172, 0.0203, 0.0213, 0.0190, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:17:56,547 INFO [train.py:904] (2/8) Epoch 16, batch 4550, loss[loss=0.2009, simple_loss=0.2888, pruned_loss=0.05648, over 16917.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2882, pruned_loss=0.05717, over 3229770.60 frames. ], batch size: 109, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:18:07,352 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156810.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:18:16,474 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3814, 3.3419, 2.5961, 2.0916, 2.2026, 2.0642, 3.5092, 3.0731], device='cuda:2'), covar=tensor([0.2853, 0.0687, 0.1748, 0.2500, 0.2494, 0.2139, 0.0504, 0.1110], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0264, 0.0295, 0.0295, 0.0291, 0.0240, 0.0282, 0.0320], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 08:18:19,417 INFO [optim.py:368] (2/8) 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 08:19:06,328 INFO [train.py:904] (2/8) Epoch 16, batch 4600, loss[loss=0.1873, simple_loss=0.2808, pruned_loss=0.04694, over 16514.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2888, pruned_loss=0.05715, over 3237699.08 frames. ], batch size: 75, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:19:36,318 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:19:55,731 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 4650, loss[loss=0.1943, simple_loss=0.2788, pruned_loss=0.05488, over 16694.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2872, pruned_loss=0.05679, over 3246707.70 frames. ], batch size: 57, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:20:37,729 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 08:20:39,077 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 08:20:45,037 INFO [optim.py:368] (2/8) 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 08:20:46,968 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:20:48,350 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7457, 3.7042, 4.1117, 1.9847, 4.3978, 4.3979, 3.0978, 3.1425], device='cuda:2'), covar=tensor([0.0786, 0.0236, 0.0211, 0.1204, 0.0056, 0.0100, 0.0450, 0.0469], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0106, 0.0093, 0.0139, 0.0074, 0.0119, 0.0126, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 08:21:12,277 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 08:21:25,715 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:25,805 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 4700, loss[loss=0.1843, simple_loss=0.2721, pruned_loss=0.04826, over 16552.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2844, pruned_loss=0.05599, over 3229576.64 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:36,260 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156994.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:22:47,496 INFO [train.py:904] (2/8) Epoch 16, batch 4750, loss[loss=0.2118, simple_loss=0.2887, pruned_loss=0.0675, over 11958.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2808, pruned_loss=0.05431, over 3203612.20 frames. ], batch size: 246, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:59,360 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157010.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:23:11,419 INFO [optim.py:368] (2/8) 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 08:23:59,602 INFO [train.py:904] (2/8) Epoch 16, batch 4800, loss[loss=0.1767, simple_loss=0.2758, pruned_loss=0.03883, over 16392.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.278, pruned_loss=0.05268, over 3202663.04 frames. ], batch size: 146, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:24:28,424 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157071.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:14,778 INFO [train.py:904] (2/8) Epoch 16, batch 4850, loss[loss=0.1973, simple_loss=0.2827, pruned_loss=0.05598, over 12361.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2788, pruned_loss=0.05231, over 3180034.19 frames. ], batch size: 247, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:25:20,198 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:39,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4094, 4.4531, 4.6438, 4.4289, 4.4756, 5.0103, 4.5375, 4.2654], device='cuda:2'), covar=tensor([0.1415, 0.1795, 0.1748, 0.1909, 0.2388, 0.0967, 0.1396, 0.2361], device='cuda:2'), in_proj_covar=tensor([0.0384, 0.0544, 0.0586, 0.0460, 0.0613, 0.0622, 0.0464, 0.0614], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 08:25:41,924 INFO [optim.py:368] (2/8) 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 08:26:16,190 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:26:30,906 INFO [train.py:904] (2/8) Epoch 16, batch 4900, loss[loss=0.1712, simple_loss=0.2599, pruned_loss=0.04128, over 16574.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2779, pruned_loss=0.05129, over 3160834.61 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:45,019 INFO [train.py:904] (2/8) Epoch 16, batch 4950, loss[loss=0.1932, simple_loss=0.2794, pruned_loss=0.05348, over 16569.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2771, pruned_loss=0.05045, over 3166119.34 frames. ], batch size: 62, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:46,693 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:27:59,785 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0192, 4.0907, 3.8980, 3.6395, 3.6046, 4.0059, 3.7071, 3.7550], device='cuda:2'), covar=tensor([0.0532, 0.0453, 0.0272, 0.0255, 0.0804, 0.0434, 0.0873, 0.0584], device='cuda:2'), in_proj_covar=tensor([0.0264, 0.0367, 0.0313, 0.0302, 0.0329, 0.0349, 0.0214, 0.0374], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:28:08,806 INFO [optim.py:368] (2/8) 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 08:28:41,075 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:57,707 INFO [train.py:904] (2/8) Epoch 16, batch 5000, loss[loss=0.2132, simple_loss=0.2992, pruned_loss=0.06365, over 16860.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2787, pruned_loss=0.05057, over 3170136.07 frames. ], batch size: 109, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:10,077 INFO [train.py:904] (2/8) Epoch 16, batch 5050, loss[loss=0.1725, simple_loss=0.2679, pruned_loss=0.03854, over 16581.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.279, pruned_loss=0.05056, over 3195322.89 frames. ], batch size: 62, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:33,953 INFO [optim.py:368] (2/8) 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 08:31:22,540 INFO [train.py:904] (2/8) Epoch 16, batch 5100, loss[loss=0.1567, simple_loss=0.2532, pruned_loss=0.03005, over 16899.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2779, pruned_loss=0.04998, over 3170483.00 frames. ], batch size: 96, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:31:29,234 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:31:40,583 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8587, 4.7973, 4.7736, 4.0278, 4.7945, 1.9342, 4.4964, 4.5897], device='cuda:2'), covar=tensor([0.0084, 0.0084, 0.0131, 0.0467, 0.0082, 0.2456, 0.0120, 0.0197], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0136, 0.0182, 0.0170, 0.0155, 0.0194, 0.0170, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:31:44,278 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:07,466 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:38,192 INFO [train.py:904] (2/8) Epoch 16, batch 5150, loss[loss=0.166, simple_loss=0.2713, pruned_loss=0.03033, over 16899.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2781, pruned_loss=0.049, over 3180543.95 frames. ], batch size: 96, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:32:42,195 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:32:43,932 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:02,222 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157417.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:04,660 INFO [optim.py:368] (2/8) 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 08:33:37,878 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8401, 4.0012, 2.9920, 2.4109, 2.7761, 2.5487, 4.1449, 3.5751], device='cuda:2'), covar=tensor([0.2309, 0.0554, 0.1655, 0.2238, 0.2280, 0.1663, 0.0421, 0.1040], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0261, 0.0292, 0.0293, 0.0286, 0.0236, 0.0279, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 08:33:40,802 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157443.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:42,433 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-30 08:33:53,087 INFO [train.py:904] (2/8) Epoch 16, batch 5200, loss[loss=0.2069, simple_loss=0.2873, pruned_loss=0.06323, over 16292.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2769, pruned_loss=0.04864, over 3186800.51 frames. ], batch size: 165, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:33:54,777 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:56,430 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 08:34:12,682 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:35:01,940 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:35:08,180 INFO [train.py:904] (2/8) Epoch 16, batch 5250, loss[loss=0.1783, simple_loss=0.273, pruned_loss=0.04179, over 16893.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2741, pruned_loss=0.04794, over 3192469.53 frames. ], batch size: 90, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:35:13,338 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157506.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:35:32,954 INFO [optim.py:368] (2/8) 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 08:36:05,774 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 5300, loss[loss=0.1483, simple_loss=0.2345, pruned_loss=0.03109, over 17003.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2712, pruned_loss=0.04679, over 3197657.37 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:36:45,039 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:36:46,068 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:36:46,436 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 08:36:49,796 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5898, 2.4060, 2.3596, 4.4644, 2.2995, 2.8171, 2.4807, 2.5992], device='cuda:2'), covar=tensor([0.1101, 0.3320, 0.2641, 0.0381, 0.3769, 0.2301, 0.3196, 0.2995], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0421, 0.0349, 0.0319, 0.0423, 0.0485, 0.0388, 0.0490], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:37:13,810 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3415, 3.5898, 3.7258, 1.9930, 3.1846, 2.4473, 3.7548, 3.7870], device='cuda:2'), covar=tensor([0.0227, 0.0677, 0.0545, 0.1916, 0.0790, 0.0885, 0.0530, 0.0850], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0148, 0.0139, 0.0125, 0.0139, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:37:17,187 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:35,557 INFO [train.py:904] (2/8) Epoch 16, batch 5350, loss[loss=0.1695, simple_loss=0.2562, pruned_loss=0.04143, over 17001.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2693, pruned_loss=0.04597, over 3218932.35 frames. ], batch size: 55, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:00,208 INFO [optim.py:368] (2/8) 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 08:38:13,578 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0349, 5.3508, 5.0707, 5.1054, 4.8365, 4.7887, 4.7499, 5.4482], device='cuda:2'), covar=tensor([0.1066, 0.0773, 0.0905, 0.0762, 0.0754, 0.0741, 0.1018, 0.0787], device='cuda:2'), in_proj_covar=tensor([0.0612, 0.0759, 0.0620, 0.0550, 0.0477, 0.0482, 0.0627, 0.0581], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:38:16,603 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 5400, loss[loss=0.1949, simple_loss=0.2847, pruned_loss=0.05253, over 15189.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2717, pruned_loss=0.04671, over 3207195.49 frames. ], batch size: 190, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:52,826 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8918, 1.9990, 2.2975, 3.2031, 2.0273, 2.1908, 2.1892, 2.1266], device='cuda:2'), covar=tensor([0.1229, 0.3308, 0.2324, 0.0636, 0.4048, 0.2534, 0.3270, 0.3244], device='cuda:2'), in_proj_covar=tensor([0.0380, 0.0421, 0.0350, 0.0320, 0.0424, 0.0486, 0.0389, 0.0491], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:39:09,627 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:04,949 INFO [train.py:904] (2/8) Epoch 16, batch 5450, loss[loss=0.1753, simple_loss=0.2618, pruned_loss=0.04435, over 16449.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2747, pruned_loss=0.04823, over 3208878.17 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:40:08,959 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0429, 2.0724, 2.2557, 3.5019, 2.0246, 2.3745, 2.2162, 2.2245], device='cuda:2'), covar=tensor([0.1210, 0.3304, 0.2514, 0.0537, 0.3997, 0.2404, 0.3273, 0.3120], device='cuda:2'), in_proj_covar=tensor([0.0380, 0.0421, 0.0350, 0.0320, 0.0423, 0.0485, 0.0388, 0.0490], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:40:19,873 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:23,494 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157714.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:29,841 INFO [optim.py:368] (2/8) 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 08:41:00,214 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:41:17,303 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-30 08:41:20,975 INFO [train.py:904] (2/8) Epoch 16, batch 5500, loss[loss=0.2518, simple_loss=0.3376, pruned_loss=0.083, over 16394.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2825, pruned_loss=0.05277, over 3191115.72 frames. ], batch size: 146, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:41:33,656 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:42:24,725 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5792, 4.8566, 4.6166, 4.6165, 4.4220, 4.3365, 4.3529, 4.9060], device='cuda:2'), covar=tensor([0.1135, 0.0825, 0.1012, 0.0825, 0.0757, 0.1273, 0.1063, 0.0895], device='cuda:2'), in_proj_covar=tensor([0.0607, 0.0754, 0.0616, 0.0546, 0.0473, 0.0480, 0.0622, 0.0575], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:42:30,909 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:42:35,997 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6106, 3.1081, 2.9871, 1.9108, 2.7257, 2.1119, 3.1146, 3.2923], device='cuda:2'), covar=tensor([0.0303, 0.0665, 0.0604, 0.1861, 0.0807, 0.0975, 0.0633, 0.0866], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0155, 0.0162, 0.0148, 0.0139, 0.0126, 0.0140, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:42:36,591 INFO [train.py:904] (2/8) Epoch 16, batch 5550, loss[loss=0.2534, simple_loss=0.3277, pruned_loss=0.08952, over 16216.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2893, pruned_loss=0.05775, over 3177054.49 frames. ], batch size: 165, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:04,447 INFO [optim.py:368] (2/8) 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 08:43:47,868 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:43:57,049 INFO [train.py:904] (2/8) Epoch 16, batch 5600, loss[loss=0.3286, simple_loss=0.3737, pruned_loss=0.1418, over 11092.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2948, pruned_loss=0.06275, over 3111576.18 frames. ], batch size: 247, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:59,500 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:44:07,220 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 08:44:14,252 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:44:55,123 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 08:45:21,486 INFO [train.py:904] (2/8) Epoch 16, batch 5650, loss[loss=0.2911, simple_loss=0.3598, pruned_loss=0.1112, over 11037.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2989, pruned_loss=0.06585, over 3107873.08 frames. ], batch size: 248, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:45:40,967 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:45:50,670 INFO [optim.py:368] (2/8) 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 08:45:53,641 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5130, 4.5658, 4.4145, 4.1710, 4.1273, 4.5218, 4.2481, 4.2247], device='cuda:2'), covar=tensor([0.0583, 0.0383, 0.0257, 0.0267, 0.0763, 0.0394, 0.0543, 0.0579], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0375, 0.0316, 0.0306, 0.0331, 0.0355, 0.0217, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:45:56,911 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:45:59,203 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0733, 2.1110, 2.3184, 3.6100, 2.0154, 2.4344, 2.2357, 2.2324], device='cuda:2'), covar=tensor([0.1269, 0.3202, 0.2387, 0.0536, 0.3920, 0.2323, 0.3013, 0.3152], device='cuda:2'), in_proj_covar=tensor([0.0377, 0.0419, 0.0348, 0.0318, 0.0422, 0.0483, 0.0386, 0.0488], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:46:39,301 INFO [train.py:904] (2/8) Epoch 16, batch 5700, loss[loss=0.2739, simple_loss=0.3284, pruned_loss=0.1097, over 11042.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3011, pruned_loss=0.06826, over 3090643.59 frames. ], batch size: 247, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:48:01,935 INFO [train.py:904] (2/8) Epoch 16, batch 5750, loss[loss=0.2186, simple_loss=0.3022, pruned_loss=0.0675, over 16559.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3039, pruned_loss=0.06959, over 3086394.01 frames. ], batch size: 62, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:48:04,666 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4632, 3.4214, 2.7045, 2.1632, 2.2758, 2.2723, 3.5485, 3.1709], device='cuda:2'), covar=tensor([0.2755, 0.0618, 0.1692, 0.2800, 0.2650, 0.1990, 0.0485, 0.1222], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0260, 0.0294, 0.0294, 0.0286, 0.0237, 0.0278, 0.0314], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 08:48:16,721 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:48:30,520 INFO [optim.py:368] (2/8) 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 08:49:00,721 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:00,777 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8106, 3.9165, 2.2031, 4.5662, 2.9143, 4.4310, 2.6009, 2.9884], device='cuda:2'), covar=tensor([0.0280, 0.0383, 0.1876, 0.0216, 0.0854, 0.0506, 0.1553, 0.0844], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0171, 0.0192, 0.0147, 0.0172, 0.0209, 0.0200, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:49:22,774 INFO [train.py:904] (2/8) Epoch 16, batch 5800, loss[loss=0.2178, simple_loss=0.2866, pruned_loss=0.0745, over 12109.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3034, pruned_loss=0.06869, over 3072872.02 frames. ], batch size: 247, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:49:35,950 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:36,064 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:50:16,598 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158086.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:50:41,300 INFO [train.py:904] (2/8) Epoch 16, batch 5850, loss[loss=0.2205, simple_loss=0.2908, pruned_loss=0.07509, over 12021.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3008, pruned_loss=0.06674, over 3080079.65 frames. ], batch size: 248, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:50:51,621 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:51:08,529 INFO [optim.py:368] (2/8) 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 08:51:12,988 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:52:03,755 INFO [train.py:904] (2/8) Epoch 16, batch 5900, loss[loss=0.2066, simple_loss=0.2996, pruned_loss=0.05685, over 16954.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2998, pruned_loss=0.06583, over 3097151.43 frames. ], batch size: 109, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:52:24,460 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:52:55,965 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:25,266 INFO [train.py:904] (2/8) Epoch 16, batch 5950, loss[loss=0.2195, simple_loss=0.3039, pruned_loss=0.06752, over 16775.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.3007, pruned_loss=0.06535, over 3084815.41 frames. ], batch size: 124, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:53:26,935 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-30 08:53:36,983 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:37,159 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6084, 3.5351, 4.0396, 1.9127, 4.2584, 4.2186, 3.1192, 3.1103], device='cuda:2'), covar=tensor([0.0727, 0.0235, 0.0146, 0.1158, 0.0048, 0.0111, 0.0390, 0.0409], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0105, 0.0091, 0.0138, 0.0073, 0.0117, 0.0124, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 08:53:38,263 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158210.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:53:55,266 INFO [optim.py:368] (2/8) 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 08:54:01,801 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:54:45,447 INFO [train.py:904] (2/8) Epoch 16, batch 6000, loss[loss=0.2083, simple_loss=0.2846, pruned_loss=0.06596, over 16449.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2994, pruned_loss=0.0643, over 3118879.33 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:54:45,447 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 08:54:56,473 INFO [train.py:938] (2/8) Epoch 16, validation: loss=0.1553, simple_loss=0.2682, pruned_loss=0.0212, over 944034.00 frames. 2023-04-30 08:54:56,474 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 08:55:27,050 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:56:13,221 INFO [train.py:904] (2/8) Epoch 16, batch 6050, loss[loss=0.2064, simple_loss=0.2982, pruned_loss=0.05733, over 16881.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2989, pruned_loss=0.06441, over 3104407.30 frames. ], batch size: 116, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:56:40,246 INFO [optim.py:368] (2/8) 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 08:56:57,640 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7768, 5.1049, 4.8152, 4.8666, 4.6247, 4.5785, 4.4799, 5.1801], device='cuda:2'), covar=tensor([0.1267, 0.0792, 0.1042, 0.0842, 0.0757, 0.0936, 0.1179, 0.0845], device='cuda:2'), in_proj_covar=tensor([0.0603, 0.0744, 0.0610, 0.0544, 0.0465, 0.0475, 0.0618, 0.0569], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 08:57:24,885 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-30 08:57:31,994 INFO [train.py:904] (2/8) Epoch 16, batch 6100, loss[loss=0.1822, simple_loss=0.2787, pruned_loss=0.04288, over 16929.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.298, pruned_loss=0.06345, over 3109327.22 frames. ], batch size: 96, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:58:24,036 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:58:37,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 08:58:45,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7824, 3.8149, 3.8829, 3.6814, 3.8531, 4.2414, 3.9041, 3.6520], device='cuda:2'), covar=tensor([0.2054, 0.1964, 0.2151, 0.2396, 0.2492, 0.1596, 0.1515, 0.2533], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0547, 0.0596, 0.0461, 0.0618, 0.0624, 0.0470, 0.0621], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 08:58:49,710 INFO [train.py:904] (2/8) Epoch 16, batch 6150, loss[loss=0.2071, simple_loss=0.2938, pruned_loss=0.06022, over 16722.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2961, pruned_loss=0.06259, over 3111532.89 frames. ], batch size: 124, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:59:11,307 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4612, 3.0103, 2.9461, 2.0316, 2.7007, 2.1173, 3.0593, 3.1276], device='cuda:2'), covar=tensor([0.0270, 0.0704, 0.0620, 0.1841, 0.0835, 0.0982, 0.0669, 0.0832], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0148, 0.0139, 0.0125, 0.0140, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 08:59:18,317 INFO [optim.py:368] (2/8) 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 08:59:57,063 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:00:07,999 INFO [train.py:904] (2/8) Epoch 16, batch 6200, loss[loss=0.2112, simple_loss=0.2846, pruned_loss=0.06891, over 11490.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2945, pruned_loss=0.06283, over 3103376.32 frames. ], batch size: 247, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:00:48,499 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158478.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:12,945 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 09:01:21,942 INFO [train.py:904] (2/8) Epoch 16, batch 6250, loss[loss=0.2026, simple_loss=0.2891, pruned_loss=0.05808, over 16892.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2936, pruned_loss=0.06219, over 3107649.68 frames. ], batch size: 109, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:01:34,519 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:44,104 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8559, 2.6874, 2.4691, 4.5214, 3.0442, 4.0506, 1.6576, 2.8523], device='cuda:2'), covar=tensor([0.1391, 0.0897, 0.1326, 0.0185, 0.0429, 0.0484, 0.1728, 0.0973], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0187, 0.0170, 0.0201, 0.0212, 0.0190, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 09:01:50,903 INFO [optim.py:368] (2/8) 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 09:02:20,198 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9953, 4.7303, 4.6945, 5.1369, 5.3500, 4.7493, 5.3172, 5.3408], device='cuda:2'), covar=tensor([0.1775, 0.1193, 0.2182, 0.0794, 0.0774, 0.1030, 0.0880, 0.0808], device='cuda:2'), in_proj_covar=tensor([0.0583, 0.0721, 0.0846, 0.0727, 0.0546, 0.0572, 0.0580, 0.0678], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:02:24,180 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7555, 4.7391, 4.6127, 4.3312, 4.2652, 4.7042, 4.5587, 4.3675], device='cuda:2'), covar=tensor([0.0691, 0.0682, 0.0298, 0.0305, 0.0945, 0.0579, 0.0418, 0.0759], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0370, 0.0312, 0.0300, 0.0326, 0.0350, 0.0215, 0.0371], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:02:38,738 INFO [train.py:904] (2/8) Epoch 16, batch 6300, loss[loss=0.2036, simple_loss=0.291, pruned_loss=0.05811, over 16806.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2937, pruned_loss=0.0619, over 3107052.88 frames. ], batch size: 116, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:02:45,560 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:03:28,161 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-04-30 09:03:55,407 INFO [train.py:904] (2/8) Epoch 16, batch 6350, loss[loss=0.2073, simple_loss=0.2936, pruned_loss=0.06051, over 16529.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2949, pruned_loss=0.06296, over 3110380.47 frames. ], batch size: 62, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:04:16,685 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6285, 2.2249, 1.8015, 1.9659, 2.6012, 2.2328, 2.5400, 2.7592], device='cuda:2'), covar=tensor([0.0168, 0.0362, 0.0529, 0.0472, 0.0223, 0.0365, 0.0218, 0.0226], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0217, 0.0211, 0.0210, 0.0218, 0.0219, 0.0221, 0.0213], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:04:24,045 INFO [optim.py:368] (2/8) 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 09:04:46,134 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7503, 3.6303, 3.8223, 3.5388, 3.7636, 4.1736, 3.8428, 3.5833], device='cuda:2'), covar=tensor([0.2015, 0.2596, 0.2605, 0.2951, 0.3123, 0.1852, 0.1625, 0.2839], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0550, 0.0601, 0.0464, 0.0622, 0.0630, 0.0474, 0.0624], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 09:05:11,913 INFO [train.py:904] (2/8) Epoch 16, batch 6400, loss[loss=0.1939, simple_loss=0.2737, pruned_loss=0.05709, over 17223.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2952, pruned_loss=0.06407, over 3110107.78 frames. ], batch size: 44, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:06:27,727 INFO [train.py:904] (2/8) Epoch 16, batch 6450, loss[loss=0.2289, simple_loss=0.2961, pruned_loss=0.08081, over 11696.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2948, pruned_loss=0.06236, over 3126078.67 frames. ], batch size: 247, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:06:42,469 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-04-30 09:06:56,388 INFO [optim.py:368] (2/8) 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 09:07:26,463 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:07:43,536 INFO [train.py:904] (2/8) Epoch 16, batch 6500, loss[loss=0.2038, simple_loss=0.2857, pruned_loss=0.06097, over 15377.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2928, pruned_loss=0.06204, over 3112039.86 frames. ], batch size: 190, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:08:21,067 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:22,793 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:31,543 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:52,225 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 09:09:01,975 INFO [train.py:904] (2/8) Epoch 16, batch 6550, loss[loss=0.213, simple_loss=0.31, pruned_loss=0.058, over 16720.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2954, pruned_loss=0.06246, over 3121203.84 frames. ], batch size: 134, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:09:19,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4313, 2.4574, 1.7458, 2.0463, 2.9160, 2.5034, 3.1388, 3.2080], device='cuda:2'), covar=tensor([0.0129, 0.0466, 0.0678, 0.0532, 0.0259, 0.0412, 0.0262, 0.0225], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0216, 0.0211, 0.0209, 0.0217, 0.0217, 0.0220, 0.0212], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:09:20,791 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7398, 4.7985, 4.6234, 4.2867, 4.2208, 4.7284, 4.5074, 4.4138], device='cuda:2'), covar=tensor([0.0763, 0.0791, 0.0282, 0.0332, 0.1059, 0.0494, 0.0517, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0371, 0.0312, 0.0300, 0.0327, 0.0349, 0.0215, 0.0372], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:09:33,175 INFO [optim.py:368] (2/8) 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 09:09:34,414 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6819, 4.3915, 4.3548, 3.0295, 3.8447, 4.3615, 3.9507, 2.4363], device='cuda:2'), covar=tensor([0.0460, 0.0038, 0.0043, 0.0321, 0.0081, 0.0096, 0.0073, 0.0383], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0074, 0.0076, 0.0129, 0.0088, 0.0098, 0.0087, 0.0122], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 09:09:38,604 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3337, 2.3380, 2.4026, 4.3787, 2.1755, 2.7603, 2.4028, 2.5467], device='cuda:2'), covar=tensor([0.1161, 0.3259, 0.2453, 0.0394, 0.3797, 0.2295, 0.3119, 0.2919], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0417, 0.0347, 0.0317, 0.0422, 0.0482, 0.0387, 0.0486], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:09:39,591 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:09:52,661 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:09:58,217 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:06,333 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:18,016 INFO [train.py:904] (2/8) Epoch 16, batch 6600, loss[loss=0.2138, simple_loss=0.2959, pruned_loss=0.06582, over 16907.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2972, pruned_loss=0.06319, over 3104643.24 frames. ], batch size: 109, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:10:35,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1825, 4.2619, 4.4317, 4.2768, 4.3075, 4.8123, 4.3621, 4.1388], device='cuda:2'), covar=tensor([0.1631, 0.1972, 0.2097, 0.1902, 0.2451, 0.1024, 0.1674, 0.2395], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0543, 0.0593, 0.0457, 0.0610, 0.0621, 0.0468, 0.0615], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 09:10:43,472 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0941, 1.8450, 2.5135, 2.9559, 2.7827, 3.1962, 1.7854, 3.2947], device='cuda:2'), covar=tensor([0.0165, 0.0472, 0.0276, 0.0226, 0.0240, 0.0172, 0.0558, 0.0129], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0185, 0.0171, 0.0174, 0.0185, 0.0142, 0.0185, 0.0134], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:10:50,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7641, 3.7530, 3.9196, 3.7090, 3.8396, 4.2440, 3.8967, 3.6719], device='cuda:2'), covar=tensor([0.2020, 0.2325, 0.2162, 0.2409, 0.2791, 0.1665, 0.1672, 0.2589], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0543, 0.0594, 0.0457, 0.0610, 0.0622, 0.0468, 0.0615], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 09:11:26,008 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:11:34,218 INFO [train.py:904] (2/8) Epoch 16, batch 6650, loss[loss=0.2105, simple_loss=0.2963, pruned_loss=0.06228, over 16809.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2982, pruned_loss=0.06447, over 3093273.82 frames. ], batch size: 116, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:11:38,829 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 09:12:04,649 INFO [optim.py:368] (2/8) 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 09:12:50,060 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 09:12:50,515 INFO [train.py:904] (2/8) Epoch 16, batch 6700, loss[loss=0.1843, simple_loss=0.274, pruned_loss=0.04726, over 16451.00 frames. ], tot_loss[loss=0.212, simple_loss=0.296, pruned_loss=0.06403, over 3101024.18 frames. ], batch size: 68, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:14:07,727 INFO [train.py:904] (2/8) Epoch 16, batch 6750, loss[loss=0.2028, simple_loss=0.2909, pruned_loss=0.05733, over 16412.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2947, pruned_loss=0.06416, over 3087943.85 frames. ], batch size: 146, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:14:37,800 INFO [optim.py:368] (2/8) 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 09:15:01,844 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0650, 2.4679, 2.5715, 1.9305, 2.6396, 2.7912, 2.4334, 2.3798], device='cuda:2'), covar=tensor([0.0694, 0.0219, 0.0215, 0.0887, 0.0093, 0.0249, 0.0425, 0.0411], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0105, 0.0091, 0.0137, 0.0073, 0.0117, 0.0124, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 09:15:05,252 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159040.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:15:23,303 INFO [train.py:904] (2/8) Epoch 16, batch 6800, loss[loss=0.1923, simple_loss=0.2867, pruned_loss=0.04889, over 16793.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2949, pruned_loss=0.06432, over 3081242.06 frames. ], batch size: 83, lr: 4.22e-03, grad_scale: 8.0 2023-04-30 09:15:50,692 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:15:58,887 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6456, 4.8422, 5.0312, 4.8341, 4.8919, 5.4031, 4.9027, 4.6886], device='cuda:2'), covar=tensor([0.1099, 0.1770, 0.2032, 0.1735, 0.2074, 0.0856, 0.1475, 0.2318], device='cuda:2'), in_proj_covar=tensor([0.0382, 0.0545, 0.0598, 0.0460, 0.0611, 0.0628, 0.0470, 0.0619], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 09:16:20,371 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:16:40,956 INFO [train.py:904] (2/8) Epoch 16, batch 6850, loss[loss=0.2487, simple_loss=0.3438, pruned_loss=0.07683, over 16301.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2963, pruned_loss=0.06426, over 3096356.89 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:17:12,962 INFO [optim.py:368] (2/8) 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 09:17:16,484 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 09:17:22,462 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:26,682 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:35,360 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:54,571 INFO [train.py:904] (2/8) Epoch 16, batch 6900, loss[loss=0.2254, simple_loss=0.3075, pruned_loss=0.07161, over 15342.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2983, pruned_loss=0.06367, over 3109855.75 frames. ], batch size: 190, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:18:07,183 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:18:56,498 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 09:19:13,870 INFO [train.py:904] (2/8) Epoch 16, batch 6950, loss[loss=0.2067, simple_loss=0.2941, pruned_loss=0.05966, over 16703.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.3002, pruned_loss=0.06538, over 3107574.67 frames. ], batch size: 83, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:19:42,140 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:19:48,719 INFO [optim.py:368] (2/8) 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 09:20:13,243 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-30 09:20:29,895 INFO [train.py:904] (2/8) Epoch 16, batch 7000, loss[loss=0.1932, simple_loss=0.2916, pruned_loss=0.04733, over 16381.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.3006, pruned_loss=0.06526, over 3090370.79 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:21:27,341 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:21:37,077 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 09:21:45,218 INFO [train.py:904] (2/8) Epoch 16, batch 7050, loss[loss=0.2273, simple_loss=0.309, pruned_loss=0.0728, over 15217.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.3009, pruned_loss=0.06447, over 3102107.85 frames. ], batch size: 191, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:22:18,893 INFO [optim.py:368] (2/8) 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 09:23:00,610 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:23:01,353 INFO [train.py:904] (2/8) Epoch 16, batch 7100, loss[loss=0.2336, simple_loss=0.2989, pruned_loss=0.08414, over 11559.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2997, pruned_loss=0.06468, over 3078256.08 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:23:54,856 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4356, 4.1741, 4.1366, 2.7915, 3.6728, 4.1602, 3.6928, 2.3849], device='cuda:2'), covar=tensor([0.0505, 0.0037, 0.0043, 0.0357, 0.0091, 0.0108, 0.0079, 0.0395], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0075, 0.0076, 0.0130, 0.0088, 0.0098, 0.0087, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 09:24:17,480 INFO [train.py:904] (2/8) Epoch 16, batch 7150, loss[loss=0.2356, simple_loss=0.2978, pruned_loss=0.08673, over 11245.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2984, pruned_loss=0.06484, over 3076863.31 frames. ], batch size: 246, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:24:49,380 INFO [optim.py:368] (2/8) 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 09:24:51,547 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:02,859 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:11,744 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:30,656 INFO [train.py:904] (2/8) Epoch 16, batch 7200, loss[loss=0.1939, simple_loss=0.2801, pruned_loss=0.05383, over 16795.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2957, pruned_loss=0.06273, over 3082597.22 frames. ], batch size: 39, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:26:15,844 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:16,039 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7946, 4.5902, 4.4760, 3.1490, 3.8869, 4.4717, 3.8930, 2.6196], device='cuda:2'), covar=tensor([0.0423, 0.0028, 0.0036, 0.0306, 0.0087, 0.0103, 0.0080, 0.0353], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0074, 0.0075, 0.0129, 0.0088, 0.0098, 0.0087, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 09:26:26,897 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:33,359 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:26:50,664 INFO [train.py:904] (2/8) Epoch 16, batch 7250, loss[loss=0.1906, simple_loss=0.2753, pruned_loss=0.05301, over 16450.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2932, pruned_loss=0.06136, over 3090462.74 frames. ], batch size: 68, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:27:11,556 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:27:14,190 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6191, 2.5219, 1.8660, 2.6645, 2.1637, 2.7424, 2.0883, 2.3540], device='cuda:2'), covar=tensor([0.0325, 0.0420, 0.1317, 0.0228, 0.0618, 0.0475, 0.1142, 0.0591], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0169, 0.0192, 0.0146, 0.0172, 0.0209, 0.0200, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 09:27:23,388 INFO [optim.py:368] (2/8) 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 09:27:47,293 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:28:06,272 INFO [train.py:904] (2/8) Epoch 16, batch 7300, loss[loss=0.2217, simple_loss=0.3004, pruned_loss=0.07149, over 16603.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2932, pruned_loss=0.06147, over 3101774.52 frames. ], batch size: 57, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:01,796 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 09:29:15,874 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7604, 1.7804, 1.5337, 1.5264, 1.8795, 1.5193, 1.6392, 1.8755], device='cuda:2'), covar=tensor([0.0133, 0.0265, 0.0363, 0.0293, 0.0190, 0.0255, 0.0162, 0.0167], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0216, 0.0210, 0.0210, 0.0216, 0.0217, 0.0219, 0.0211], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:29:22,389 INFO [train.py:904] (2/8) Epoch 16, batch 7350, loss[loss=0.2328, simple_loss=0.3299, pruned_loss=0.0679, over 15293.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2949, pruned_loss=0.06316, over 3075522.41 frames. ], batch size: 191, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:56,735 INFO [optim.py:368] (2/8) 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 09:30:07,730 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159646.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:39,219 INFO [train.py:904] (2/8) Epoch 16, batch 7400, loss[loss=0.2027, simple_loss=0.2804, pruned_loss=0.06251, over 16258.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2953, pruned_loss=0.06347, over 3075286.76 frames. ], batch size: 35, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:30:41,303 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-04-30 09:31:18,375 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5314, 4.5039, 4.3232, 3.5676, 4.3954, 1.5920, 4.1324, 4.0209], device='cuda:2'), covar=tensor([0.0100, 0.0078, 0.0191, 0.0401, 0.0101, 0.2918, 0.0138, 0.0255], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0134, 0.0181, 0.0166, 0.0153, 0.0192, 0.0167, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:31:41,344 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:31:57,478 INFO [train.py:904] (2/8) Epoch 16, batch 7450, loss[loss=0.2367, simple_loss=0.3073, pruned_loss=0.08304, over 11624.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2965, pruned_loss=0.06436, over 3061432.88 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:32:33,422 INFO [optim.py:368] (2/8) 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 09:32:35,921 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:32:49,022 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 09:33:02,577 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2766, 3.4222, 3.6142, 3.5798, 3.5815, 3.3931, 3.4387, 3.4852], device='cuda:2'), covar=tensor([0.0416, 0.0655, 0.0460, 0.0510, 0.0530, 0.0535, 0.0801, 0.0482], device='cuda:2'), in_proj_covar=tensor([0.0372, 0.0401, 0.0389, 0.0371, 0.0440, 0.0415, 0.0510, 0.0329], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 09:33:17,743 INFO [train.py:904] (2/8) Epoch 16, batch 7500, loss[loss=0.2048, simple_loss=0.2903, pruned_loss=0.05964, over 16444.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2971, pruned_loss=0.06386, over 3070516.34 frames. ], batch size: 146, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:33:25,919 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 09:33:50,299 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:34:37,012 INFO [train.py:904] (2/8) Epoch 16, batch 7550, loss[loss=0.2115, simple_loss=0.2882, pruned_loss=0.0674, over 16240.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2959, pruned_loss=0.06389, over 3064116.76 frames. ], batch size: 165, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:34:58,282 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:35:10,141 INFO [optim.py:368] (2/8) 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 09:35:21,260 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 09:35:53,836 INFO [train.py:904] (2/8) Epoch 16, batch 7600, loss[loss=0.2361, simple_loss=0.3007, pruned_loss=0.08574, over 11361.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2948, pruned_loss=0.06379, over 3070997.91 frames. ], batch size: 250, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:35:59,083 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:36:10,415 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:36:34,572 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0303, 2.0150, 2.2060, 3.6192, 1.9559, 2.3685, 2.1478, 2.1995], device='cuda:2'), covar=tensor([0.1306, 0.3817, 0.2698, 0.0557, 0.4310, 0.2542, 0.3521, 0.3207], device='cuda:2'), in_proj_covar=tensor([0.0378, 0.0420, 0.0347, 0.0317, 0.0425, 0.0483, 0.0388, 0.0488], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:37:09,859 INFO [train.py:904] (2/8) Epoch 16, batch 7650, loss[loss=0.2973, simple_loss=0.3442, pruned_loss=0.1252, over 11495.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2952, pruned_loss=0.0646, over 3052113.36 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:37:21,655 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:30,422 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:43,258 INFO [optim.py:368] (2/8) 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 09:37:57,624 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 09:38:13,602 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:38:22,085 INFO [train.py:904] (2/8) Epoch 16, batch 7700, loss[loss=0.253, simple_loss=0.3149, pruned_loss=0.09556, over 11313.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2948, pruned_loss=0.06441, over 3064868.61 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:38:50,529 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159971.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:15,434 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:26,412 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:41,049 INFO [train.py:904] (2/8) Epoch 16, batch 7750, loss[loss=0.2363, simple_loss=0.3158, pruned_loss=0.0784, over 16219.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2951, pruned_loss=0.06466, over 3061169.15 frames. ], batch size: 165, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:40:13,566 INFO [optim.py:368] (2/8) 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 09:40:53,466 INFO [train.py:904] (2/8) Epoch 16, batch 7800, loss[loss=0.2396, simple_loss=0.3137, pruned_loss=0.08271, over 15327.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2967, pruned_loss=0.06589, over 3056052.54 frames. ], batch size: 190, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:01,842 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 09:42:08,841 INFO [train.py:904] (2/8) Epoch 16, batch 7850, loss[loss=0.2204, simple_loss=0.3054, pruned_loss=0.06767, over 16299.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2976, pruned_loss=0.06555, over 3060981.23 frames. ], batch size: 165, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:43,406 INFO [optim.py:368] (2/8) 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 09:43:25,069 INFO [train.py:904] (2/8) Epoch 16, batch 7900, loss[loss=0.2005, simple_loss=0.2833, pruned_loss=0.0589, over 16485.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2961, pruned_loss=0.06435, over 3078516.54 frames. ], batch size: 35, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:43,670 INFO [train.py:904] (2/8) Epoch 16, batch 7950, loss[loss=0.1885, simple_loss=0.2751, pruned_loss=0.05099, over 16818.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.296, pruned_loss=0.0645, over 3080498.52 frames. ], batch size: 39, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:56,864 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:44:57,049 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7167, 3.7730, 2.3195, 4.3480, 2.8950, 4.3214, 2.4238, 3.0624], device='cuda:2'), covar=tensor([0.0242, 0.0333, 0.1560, 0.0210, 0.0818, 0.0495, 0.1502, 0.0692], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0169, 0.0192, 0.0147, 0.0172, 0.0210, 0.0201, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 09:45:16,536 INFO [optim.py:368] (2/8) 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 09:45:56,353 INFO [train.py:904] (2/8) Epoch 16, batch 8000, loss[loss=0.2223, simple_loss=0.3099, pruned_loss=0.06733, over 16713.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2968, pruned_loss=0.06519, over 3084830.26 frames. ], batch size: 124, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:46:17,933 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:46:23,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8202, 2.6914, 2.8139, 2.0874, 2.6773, 2.1381, 2.6388, 2.8955], device='cuda:2'), covar=tensor([0.0246, 0.0755, 0.0476, 0.1725, 0.0794, 0.0887, 0.0540, 0.0751], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0147, 0.0140, 0.0125, 0.0140, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 09:46:50,261 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:47:12,589 INFO [train.py:904] (2/8) Epoch 16, batch 8050, loss[loss=0.2259, simple_loss=0.305, pruned_loss=0.07338, over 16896.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2959, pruned_loss=0.06435, over 3098265.33 frames. ], batch size: 116, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:47:20,264 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 09:48:04,047 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:48:07,296 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8896, 3.9693, 4.3038, 4.2718, 4.2584, 3.9759, 4.0366, 3.9677], device='cuda:2'), covar=tensor([0.0369, 0.0551, 0.0378, 0.0431, 0.0499, 0.0427, 0.0893, 0.0567], device='cuda:2'), in_proj_covar=tensor([0.0376, 0.0406, 0.0396, 0.0376, 0.0449, 0.0420, 0.0513, 0.0334], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 09:48:29,600 INFO [train.py:904] (2/8) Epoch 16, batch 8100, loss[loss=0.2567, simple_loss=0.3187, pruned_loss=0.09731, over 11517.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2951, pruned_loss=0.06323, over 3104477.72 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:48:54,846 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:49:46,003 INFO [train.py:904] (2/8) Epoch 16, batch 8150, loss[loss=0.1893, simple_loss=0.2712, pruned_loss=0.05375, over 16665.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2929, pruned_loss=0.06257, over 3114037.45 frames. ], batch size: 76, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:50:21,736 INFO [optim.py:368] (2/8) 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 09:50:24,033 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 09:50:27,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7769, 3.1287, 2.5112, 5.0149, 3.8598, 4.2721, 1.6115, 3.0119], device='cuda:2'), covar=tensor([0.1467, 0.0755, 0.1421, 0.0162, 0.0435, 0.0462, 0.1796, 0.0919], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0167, 0.0190, 0.0172, 0.0205, 0.0213, 0.0194, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 09:51:05,065 INFO [train.py:904] (2/8) Epoch 16, batch 8200, loss[loss=0.21, simple_loss=0.2961, pruned_loss=0.06195, over 15321.00 frames. ], tot_loss[loss=0.207, simple_loss=0.29, pruned_loss=0.06198, over 3111702.00 frames. ], batch size: 190, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:27,314 INFO [train.py:904] (2/8) Epoch 16, batch 8250, loss[loss=0.1881, simple_loss=0.2695, pruned_loss=0.05338, over 11951.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2887, pruned_loss=0.05946, over 3078752.94 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:42,386 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:52:53,339 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9659, 2.8485, 2.8721, 2.2229, 2.6340, 2.1956, 2.5682, 2.9754], device='cuda:2'), covar=tensor([0.0378, 0.0717, 0.0471, 0.1606, 0.0832, 0.0901, 0.0737, 0.0704], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0155, 0.0162, 0.0148, 0.0140, 0.0126, 0.0141, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 09:53:04,606 INFO [optim.py:368] (2/8) 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 09:53:20,236 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 09:53:49,472 INFO [train.py:904] (2/8) Epoch 16, batch 8300, loss[loss=0.1769, simple_loss=0.2735, pruned_loss=0.04011, over 15408.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2862, pruned_loss=0.057, over 3036804.89 frames. ], batch size: 190, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:54:00,902 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:13,038 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:55:10,565 INFO [train.py:904] (2/8) Epoch 16, batch 8350, loss[loss=0.2199, simple_loss=0.2932, pruned_loss=0.07328, over 12364.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2856, pruned_loss=0.05508, over 3035725.14 frames. ], batch size: 246, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:55:30,914 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:55:43,556 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 09:55:48,787 INFO [optim.py:368] (2/8) 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 09:56:33,055 INFO [train.py:904] (2/8) Epoch 16, batch 8400, loss[loss=0.1817, simple_loss=0.2651, pruned_loss=0.04919, over 12535.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2829, pruned_loss=0.05272, over 3029504.08 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:56:34,007 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 09:56:50,686 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:57:54,560 INFO [train.py:904] (2/8) Epoch 16, batch 8450, loss[loss=0.1814, simple_loss=0.2634, pruned_loss=0.04974, over 12235.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2809, pruned_loss=0.05103, over 3025369.05 frames. ], batch size: 246, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:58:25,801 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8302, 4.7710, 4.6002, 3.7180, 4.6672, 1.6309, 4.3816, 4.3169], device='cuda:2'), covar=tensor([0.0090, 0.0103, 0.0197, 0.0452, 0.0118, 0.2960, 0.0166, 0.0264], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0132, 0.0178, 0.0162, 0.0150, 0.0189, 0.0164, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:58:31,816 INFO [optim.py:368] (2/8) 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 09:58:40,529 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.8676, 6.2123, 5.8746, 5.9605, 5.6080, 5.5570, 5.5628, 6.2576], device='cuda:2'), covar=tensor([0.1030, 0.0699, 0.0933, 0.0773, 0.0688, 0.0533, 0.1078, 0.0781], device='cuda:2'), in_proj_covar=tensor([0.0598, 0.0728, 0.0596, 0.0536, 0.0458, 0.0475, 0.0608, 0.0561], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:59:15,544 INFO [train.py:904] (2/8) Epoch 16, batch 8500, loss[loss=0.184, simple_loss=0.2763, pruned_loss=0.04583, over 16205.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2775, pruned_loss=0.04865, over 3030075.86 frames. ], batch size: 165, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:59:26,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7378, 2.6794, 2.5343, 3.8725, 2.5012, 3.9527, 1.4806, 2.9673], device='cuda:2'), covar=tensor([0.1316, 0.0667, 0.1023, 0.0160, 0.0132, 0.0345, 0.1602, 0.0669], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0163, 0.0186, 0.0167, 0.0199, 0.0208, 0.0190, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 09:59:41,847 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3790, 4.3802, 4.7557, 4.7165, 4.7022, 4.4399, 4.4348, 4.2941], device='cuda:2'), covar=tensor([0.0345, 0.0656, 0.0406, 0.0416, 0.0518, 0.0409, 0.0988, 0.0522], device='cuda:2'), in_proj_covar=tensor([0.0370, 0.0401, 0.0391, 0.0369, 0.0437, 0.0413, 0.0504, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 09:59:49,878 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8488, 2.2416, 1.7552, 1.8110, 2.5219, 2.2212, 2.6379, 2.7926], device='cuda:2'), covar=tensor([0.0186, 0.0427, 0.0607, 0.0526, 0.0270, 0.0389, 0.0197, 0.0241], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0213, 0.0207, 0.0207, 0.0212, 0.0213, 0.0215, 0.0208], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 09:59:50,052 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 10:00:26,478 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6121, 3.7182, 2.9340, 2.0908, 2.3049, 2.3674, 3.8937, 3.2513], device='cuda:2'), covar=tensor([0.2780, 0.0500, 0.1545, 0.3003, 0.2840, 0.2075, 0.0362, 0.1224], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0255, 0.0288, 0.0290, 0.0280, 0.0235, 0.0274, 0.0310], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:00:42,445 INFO [train.py:904] (2/8) Epoch 16, batch 8550, loss[loss=0.2035, simple_loss=0.3093, pruned_loss=0.04888, over 16860.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2753, pruned_loss=0.04767, over 3007578.88 frames. ], batch size: 102, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:01:13,741 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3234, 3.4045, 3.6650, 3.6292, 3.6339, 3.4751, 3.5001, 3.5325], device='cuda:2'), covar=tensor([0.0391, 0.0768, 0.0448, 0.0511, 0.0534, 0.0517, 0.0825, 0.0496], device='cuda:2'), in_proj_covar=tensor([0.0372, 0.0402, 0.0393, 0.0371, 0.0440, 0.0414, 0.0508, 0.0329], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:01:19,115 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 10:01:27,141 INFO [optim.py:368] (2/8) 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 10:02:17,150 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7125, 3.7648, 4.0990, 4.0555, 4.0826, 3.8462, 3.8624, 3.8809], device='cuda:2'), covar=tensor([0.0377, 0.0594, 0.0475, 0.0503, 0.0484, 0.0431, 0.0878, 0.0437], device='cuda:2'), in_proj_covar=tensor([0.0371, 0.0400, 0.0393, 0.0370, 0.0439, 0.0413, 0.0506, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:02:22,610 INFO [train.py:904] (2/8) Epoch 16, batch 8600, loss[loss=0.1864, simple_loss=0.2824, pruned_loss=0.04519, over 15415.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2759, pruned_loss=0.04672, over 3022959.84 frames. ], batch size: 191, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:04:02,573 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160901.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:04:03,213 INFO [train.py:904] (2/8) Epoch 16, batch 8650, loss[loss=0.1847, simple_loss=0.2702, pruned_loss=0.04964, over 12168.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2745, pruned_loss=0.04522, over 3028924.86 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:04:56,279 INFO [optim.py:368] (2/8) 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 10:05:17,622 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-04-30 10:05:52,474 INFO [train.py:904] (2/8) Epoch 16, batch 8700, loss[loss=0.1763, simple_loss=0.2723, pruned_loss=0.04014, over 15282.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2712, pruned_loss=0.04358, over 3029486.23 frames. ], batch size: 191, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:06:13,057 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160962.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:15,124 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 10:06:18,466 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:47,351 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0194, 3.3372, 3.3511, 2.1180, 3.0113, 3.3277, 3.2009, 1.8812], device='cuda:2'), covar=tensor([0.0541, 0.0044, 0.0048, 0.0426, 0.0097, 0.0078, 0.0075, 0.0504], device='cuda:2'), in_proj_covar=tensor([0.0131, 0.0073, 0.0075, 0.0130, 0.0088, 0.0097, 0.0086, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 10:07:29,336 INFO [train.py:904] (2/8) Epoch 16, batch 8750, loss[loss=0.1823, simple_loss=0.2806, pruned_loss=0.042, over 15462.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2699, pruned_loss=0.04261, over 3028765.29 frames. ], batch size: 192, lr: 4.20e-03, grad_scale: 4.0 2023-04-30 10:07:53,442 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161011.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 10:08:27,683 INFO [optim.py:368] (2/8) 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 10:08:28,250 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:09:20,779 INFO [train.py:904] (2/8) Epoch 16, batch 8800, loss[loss=0.1933, simple_loss=0.2855, pruned_loss=0.05052, over 16375.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.269, pruned_loss=0.04177, over 3043132.01 frames. ], batch size: 146, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:10:52,400 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 10:11:05,639 INFO [train.py:904] (2/8) Epoch 16, batch 8850, loss[loss=0.1861, simple_loss=0.2676, pruned_loss=0.05229, over 12415.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.272, pruned_loss=0.04144, over 3044164.79 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:55,943 INFO [optim.py:368] (2/8) 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 10:12:17,651 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-30 10:12:20,059 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 10:12:53,421 INFO [train.py:904] (2/8) Epoch 16, batch 8900, loss[loss=0.1881, simple_loss=0.2871, pruned_loss=0.04459, over 15368.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2725, pruned_loss=0.04092, over 3040396.46 frames. ], batch size: 191, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:12:55,311 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-30 10:14:59,678 INFO [train.py:904] (2/8) Epoch 16, batch 8950, loss[loss=0.1485, simple_loss=0.2442, pruned_loss=0.02636, over 16251.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2719, pruned_loss=0.04101, over 3071357.56 frames. ], batch size: 165, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:15:04,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5049, 3.5379, 2.1622, 3.9589, 2.6882, 3.8838, 2.2424, 2.7822], device='cuda:2'), covar=tensor([0.0288, 0.0385, 0.1684, 0.0139, 0.0865, 0.0493, 0.1667, 0.0769], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0163, 0.0186, 0.0140, 0.0167, 0.0200, 0.0195, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:15:12,012 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 10:15:29,263 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 10:15:50,299 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9401, 2.7869, 2.6686, 1.9975, 2.4945, 2.7325, 2.5799, 1.9112], device='cuda:2'), covar=tensor([0.0375, 0.0055, 0.0052, 0.0309, 0.0112, 0.0071, 0.0085, 0.0378], device='cuda:2'), in_proj_covar=tensor([0.0129, 0.0072, 0.0073, 0.0128, 0.0087, 0.0096, 0.0085, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 10:16:01,816 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:16:19,124 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 10:16:46,879 INFO [train.py:904] (2/8) Epoch 16, batch 9000, loss[loss=0.1627, simple_loss=0.2527, pruned_loss=0.03635, over 16833.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2682, pruned_loss=0.03962, over 3078506.00 frames. ], batch size: 124, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:16:46,880 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 10:16:56,905 INFO [train.py:938] (2/8) Epoch 16, validation: loss=0.1491, simple_loss=0.2531, pruned_loss=0.02259, over 944034.00 frames. 2023-04-30 10:16:56,906 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 10:17:08,076 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:17:48,844 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:20,897 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:39,996 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 10:18:40,780 INFO [train.py:904] (2/8) Epoch 16, batch 9050, loss[loss=0.1659, simple_loss=0.2533, pruned_loss=0.0392, over 12853.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2679, pruned_loss=0.03958, over 3060250.24 frames. ], batch size: 246, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:19:19,386 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161321.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:19:27,691 INFO [optim.py:368] (2/8) 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 10:20:22,817 INFO [train.py:904] (2/8) Epoch 16, batch 9100, loss[loss=0.1968, simple_loss=0.2883, pruned_loss=0.0527, over 16754.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2674, pruned_loss=0.0399, over 3074884.56 frames. ], batch size: 124, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:21:49,738 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6616, 2.9806, 2.7950, 5.0949, 3.7963, 4.5114, 1.5648, 3.4113], device='cuda:2'), covar=tensor([0.1439, 0.0754, 0.1179, 0.0169, 0.0259, 0.0350, 0.1694, 0.0648], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0162, 0.0185, 0.0165, 0.0194, 0.0206, 0.0190, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 10:22:19,350 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:22:19,975 INFO [train.py:904] (2/8) Epoch 16, batch 9150, loss[loss=0.1761, simple_loss=0.2711, pruned_loss=0.04054, over 16275.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2687, pruned_loss=0.04008, over 3070810.40 frames. ], batch size: 35, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:23:13,970 INFO [optim.py:368] (2/8) 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 10:23:17,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7334, 1.6759, 2.1985, 2.5845, 2.5614, 2.8762, 1.9186, 2.8678], device='cuda:2'), covar=tensor([0.0211, 0.0549, 0.0352, 0.0264, 0.0303, 0.0201, 0.0504, 0.0159], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0180, 0.0167, 0.0168, 0.0180, 0.0137, 0.0182, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:24:03,846 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6395, 2.1329, 1.8189, 1.9050, 2.4240, 2.0915, 2.1431, 2.5365], device='cuda:2'), covar=tensor([0.0138, 0.0359, 0.0446, 0.0415, 0.0233, 0.0339, 0.0163, 0.0232], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0214, 0.0208, 0.0208, 0.0213, 0.0214, 0.0214, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:24:04,520 INFO [train.py:904] (2/8) Epoch 16, batch 9200, loss[loss=0.1855, simple_loss=0.2894, pruned_loss=0.04081, over 16238.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2656, pruned_loss=0.03955, over 3081901.03 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:24:20,247 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:24:24,476 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:25:42,697 INFO [train.py:904] (2/8) Epoch 16, batch 9250, loss[loss=0.179, simple_loss=0.2702, pruned_loss=0.04393, over 16412.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2657, pruned_loss=0.03971, over 3077630.19 frames. ], batch size: 146, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:26:22,889 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:26:36,793 INFO [optim.py:368] (2/8) 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 10:27:34,801 INFO [train.py:904] (2/8) Epoch 16, batch 9300, loss[loss=0.1474, simple_loss=0.2365, pruned_loss=0.02914, over 16573.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2646, pruned_loss=0.03956, over 3074874.56 frames. ], batch size: 62, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:27:45,478 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:23,205 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:54,333 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:58,097 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5385, 4.5656, 4.9102, 4.8962, 4.8781, 4.6410, 4.5910, 4.5027], device='cuda:2'), covar=tensor([0.0319, 0.0470, 0.0379, 0.0337, 0.0455, 0.0319, 0.0787, 0.0397], device='cuda:2'), in_proj_covar=tensor([0.0355, 0.0382, 0.0378, 0.0354, 0.0423, 0.0397, 0.0484, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:29:21,684 INFO [train.py:904] (2/8) Epoch 16, batch 9350, loss[loss=0.1816, simple_loss=0.2718, pruned_loss=0.04571, over 16189.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2641, pruned_loss=0.0393, over 3069316.50 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:29:28,432 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:29:54,132 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1704, 1.8906, 2.0746, 3.8265, 1.8836, 2.1273, 2.0606, 1.9629], device='cuda:2'), covar=tensor([0.1328, 0.4478, 0.3163, 0.0569, 0.5600, 0.3300, 0.4037, 0.4523], device='cuda:2'), in_proj_covar=tensor([0.0367, 0.0408, 0.0342, 0.0308, 0.0415, 0.0468, 0.0377, 0.0474], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:30:02,157 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:12,558 INFO [optim.py:368] (2/8) 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 10:30:14,956 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4908, 5.7676, 5.5408, 5.5711, 5.2037, 5.1826, 5.2298, 5.8584], device='cuda:2'), covar=tensor([0.1003, 0.0828, 0.0862, 0.0687, 0.0779, 0.0635, 0.0978, 0.0763], device='cuda:2'), in_proj_covar=tensor([0.0583, 0.0715, 0.0581, 0.0526, 0.0452, 0.0466, 0.0594, 0.0548], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:31:03,992 INFO [train.py:904] (2/8) Epoch 16, batch 9400, loss[loss=0.1854, simple_loss=0.2871, pruned_loss=0.04187, over 16889.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2648, pruned_loss=0.03926, over 3080647.71 frames. ], batch size: 116, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:31:38,667 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:32:09,046 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9984, 2.0145, 2.3773, 2.9371, 2.7935, 3.3351, 2.0483, 3.2736], device='cuda:2'), covar=tensor([0.0180, 0.0453, 0.0329, 0.0263, 0.0264, 0.0130, 0.0477, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0178, 0.0165, 0.0167, 0.0179, 0.0135, 0.0179, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:32:38,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7786, 2.1235, 1.7580, 1.8468, 2.4263, 2.0971, 2.3348, 2.6179], device='cuda:2'), covar=tensor([0.0145, 0.0396, 0.0541, 0.0508, 0.0281, 0.0392, 0.0202, 0.0252], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0217, 0.0211, 0.0210, 0.0216, 0.0217, 0.0217, 0.0208], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:32:44,154 INFO [train.py:904] (2/8) Epoch 16, batch 9450, loss[loss=0.1689, simple_loss=0.2619, pruned_loss=0.03794, over 15507.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2658, pruned_loss=0.03929, over 3060401.18 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:32:56,261 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4351, 4.2044, 4.4830, 4.5983, 4.7801, 4.3217, 4.7660, 4.8072], device='cuda:2'), covar=tensor([0.1656, 0.1356, 0.1444, 0.0730, 0.0500, 0.0984, 0.0531, 0.0620], device='cuda:2'), in_proj_covar=tensor([0.0552, 0.0683, 0.0799, 0.0690, 0.0521, 0.0545, 0.0555, 0.0648], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:33:33,849 INFO [optim.py:368] (2/8) 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 10:33:50,920 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 10:34:23,902 INFO [train.py:904] (2/8) Epoch 16, batch 9500, loss[loss=0.1638, simple_loss=0.2595, pruned_loss=0.0341, over 16754.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2654, pruned_loss=0.03918, over 3057020.05 frames. ], batch size: 76, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:34:37,542 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161757.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:08,214 INFO [train.py:904] (2/8) Epoch 16, batch 9550, loss[loss=0.1799, simple_loss=0.2629, pruned_loss=0.04847, over 12370.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2647, pruned_loss=0.03906, over 3063996.88 frames. ], batch size: 247, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:36:38,131 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:40,706 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:37:00,497 INFO [optim.py:368] (2/8) 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 10:37:51,435 INFO [train.py:904] (2/8) Epoch 16, batch 9600, loss[loss=0.2019, simple_loss=0.2951, pruned_loss=0.05438, over 15323.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2662, pruned_loss=0.03986, over 3059782.38 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:38:29,859 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:38:38,439 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:38:38,915 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-30 10:39:02,358 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:37,497 INFO [train.py:904] (2/8) Epoch 16, batch 9650, loss[loss=0.1718, simple_loss=0.2681, pruned_loss=0.03773, over 16176.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2673, pruned_loss=0.0398, over 3054094.16 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:40:22,073 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:28,330 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 10:40:36,069 INFO [optim.py:368] (2/8) 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 10:40:49,471 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:55,598 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6111, 4.6221, 4.9951, 4.9685, 4.9677, 4.7405, 4.7039, 4.5358], device='cuda:2'), covar=tensor([0.0262, 0.0540, 0.0391, 0.0367, 0.0378, 0.0334, 0.0693, 0.0340], device='cuda:2'), in_proj_covar=tensor([0.0357, 0.0383, 0.0378, 0.0353, 0.0421, 0.0396, 0.0485, 0.0314], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:41:23,666 INFO [train.py:904] (2/8) Epoch 16, batch 9700, loss[loss=0.1796, simple_loss=0.2729, pruned_loss=0.04312, over 15343.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2668, pruned_loss=0.03999, over 3059132.41 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:41:33,352 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8957, 2.0463, 2.2854, 3.2087, 2.0822, 2.2282, 2.2108, 2.1812], device='cuda:2'), covar=tensor([0.1099, 0.3770, 0.2553, 0.0574, 0.4466, 0.2635, 0.3486, 0.3481], device='cuda:2'), in_proj_covar=tensor([0.0369, 0.0408, 0.0343, 0.0309, 0.0417, 0.0467, 0.0378, 0.0473], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:41:42,154 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9216, 2.0776, 2.3331, 3.2307, 2.1090, 2.2420, 2.2625, 2.1923], device='cuda:2'), covar=tensor([0.1046, 0.3642, 0.2543, 0.0570, 0.4342, 0.2596, 0.3128, 0.3515], device='cuda:2'), in_proj_covar=tensor([0.0368, 0.0408, 0.0342, 0.0309, 0.0416, 0.0466, 0.0378, 0.0473], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:42:06,725 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1401, 4.1627, 4.5070, 4.5047, 4.4993, 4.2683, 4.2634, 4.1984], device='cuda:2'), covar=tensor([0.0334, 0.0669, 0.0510, 0.0467, 0.0450, 0.0425, 0.0801, 0.0408], device='cuda:2'), in_proj_covar=tensor([0.0359, 0.0384, 0.0380, 0.0355, 0.0423, 0.0398, 0.0486, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:42:12,387 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4054, 3.4729, 2.0226, 3.8045, 2.5496, 3.7847, 2.2693, 2.8571], device='cuda:2'), covar=tensor([0.0258, 0.0313, 0.1585, 0.0177, 0.0843, 0.0512, 0.1399, 0.0687], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0162, 0.0184, 0.0139, 0.0166, 0.0198, 0.0194, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:43:08,495 INFO [train.py:904] (2/8) Epoch 16, batch 9750, loss[loss=0.17, simple_loss=0.2649, pruned_loss=0.03752, over 16824.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2652, pruned_loss=0.0399, over 3056144.38 frames. ], batch size: 124, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:43:58,494 INFO [optim.py:368] (2/8) 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 10:44:40,091 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8311, 1.3422, 1.7562, 1.7227, 1.8527, 1.9464, 1.6042, 1.8946], device='cuda:2'), covar=tensor([0.0268, 0.0375, 0.0200, 0.0280, 0.0273, 0.0213, 0.0385, 0.0139], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0176, 0.0164, 0.0166, 0.0178, 0.0134, 0.0178, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:44:46,280 INFO [train.py:904] (2/8) Epoch 16, batch 9800, loss[loss=0.1608, simple_loss=0.2627, pruned_loss=0.02941, over 16419.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2652, pruned_loss=0.03889, over 3057114.98 frames. ], batch size: 68, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:44:56,613 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:45:34,283 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:31,256 INFO [train.py:904] (2/8) Epoch 16, batch 9850, loss[loss=0.1919, simple_loss=0.2836, pruned_loss=0.05013, over 16838.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2664, pruned_loss=0.03863, over 3056951.60 frames. ], batch size: 124, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:46:38,614 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:59,768 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:47:23,357 INFO [optim.py:368] (2/8) 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 10:47:51,505 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:48:22,256 INFO [train.py:904] (2/8) Epoch 16, batch 9900, loss[loss=0.1565, simple_loss=0.2446, pruned_loss=0.03422, over 12688.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2665, pruned_loss=0.0385, over 3056595.87 frames. ], batch size: 248, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:48:35,340 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3643, 3.0313, 2.6582, 2.2114, 2.1780, 2.2784, 2.9766, 2.8604], device='cuda:2'), covar=tensor([0.2646, 0.0796, 0.1533, 0.2576, 0.2399, 0.1972, 0.0478, 0.1415], device='cuda:2'), in_proj_covar=tensor([0.0306, 0.0251, 0.0284, 0.0286, 0.0270, 0.0232, 0.0270, 0.0304], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:48:52,698 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:49:08,582 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:50:19,043 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9152, 4.0336, 2.3741, 4.5783, 3.0897, 4.5022, 2.4302, 3.2941], device='cuda:2'), covar=tensor([0.0210, 0.0253, 0.1399, 0.0150, 0.0663, 0.0336, 0.1461, 0.0555], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0162, 0.0183, 0.0138, 0.0165, 0.0197, 0.0193, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 10:50:22,159 INFO [train.py:904] (2/8) Epoch 16, batch 9950, loss[loss=0.17, simple_loss=0.2649, pruned_loss=0.0375, over 16778.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2688, pruned_loss=0.03894, over 3071138.32 frames. ], batch size: 76, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:51:11,939 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7497, 4.6769, 4.5938, 4.1477, 4.2298, 4.6238, 4.4563, 4.3273], device='cuda:2'), covar=tensor([0.0589, 0.0733, 0.0385, 0.0369, 0.1054, 0.0584, 0.0426, 0.0762], device='cuda:2'), in_proj_covar=tensor([0.0254, 0.0351, 0.0296, 0.0284, 0.0303, 0.0330, 0.0202, 0.0351], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-04-30 10:51:26,523 INFO [optim.py:368] (2/8) 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 10:51:36,396 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:52:23,901 INFO [train.py:904] (2/8) Epoch 16, batch 10000, loss[loss=0.178, simple_loss=0.2623, pruned_loss=0.04689, over 12788.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2674, pruned_loss=0.03859, over 3085863.41 frames. ], batch size: 247, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:53:46,616 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 16, batch 10050, loss[loss=0.186, simple_loss=0.2819, pruned_loss=0.04507, over 16233.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2674, pruned_loss=0.03826, over 3080540.10 frames. ], batch size: 166, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:54:54,659 INFO [optim.py:368] (2/8) 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 10:55:24,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6378, 4.8370, 5.0059, 4.8677, 4.9324, 5.3639, 4.8065, 4.5341], device='cuda:2'), covar=tensor([0.1041, 0.1700, 0.1639, 0.1690, 0.2037, 0.0902, 0.1542, 0.2498], device='cuda:2'), in_proj_covar=tensor([0.0357, 0.0515, 0.0566, 0.0434, 0.0576, 0.0599, 0.0449, 0.0581], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 10:55:38,457 INFO [train.py:904] (2/8) Epoch 16, batch 10100, loss[loss=0.1529, simple_loss=0.2513, pruned_loss=0.02728, over 16706.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2673, pruned_loss=0.03828, over 3090083.05 frames. ], batch size: 83, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:57:23,030 INFO [train.py:904] (2/8) Epoch 17, batch 0, loss[loss=0.2215, simple_loss=0.3133, pruned_loss=0.0648, over 16774.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3133, pruned_loss=0.0648, over 16774.00 frames. ], batch size: 57, lr: 4.05e-03, grad_scale: 8.0 2023-04-30 10:57:23,030 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 10:57:30,741 INFO [train.py:938] (2/8) Epoch 17, validation: loss=0.1481, simple_loss=0.2518, pruned_loss=0.02217, over 944034.00 frames. 2023-04-30 10:57:30,742 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 10:58:09,502 INFO [optim.py:368] (2/8) 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 10:58:14,142 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:58:39,852 INFO [train.py:904] (2/8) Epoch 17, batch 50, loss[loss=0.214, simple_loss=0.2892, pruned_loss=0.06943, over 16753.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2762, pruned_loss=0.0532, over 752830.82 frames. ], batch size: 83, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 10:59:06,416 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:59:13,258 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4677, 2.1780, 2.2807, 4.3071, 2.2630, 2.6300, 2.3047, 2.4317], device='cuda:2'), covar=tensor([0.1095, 0.3747, 0.2734, 0.0438, 0.3895, 0.2477, 0.3404, 0.3327], device='cuda:2'), in_proj_covar=tensor([0.0372, 0.0410, 0.0345, 0.0312, 0.0419, 0.0470, 0.0382, 0.0476], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:59:35,891 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7554, 2.3571, 1.9101, 2.1695, 2.8190, 2.5391, 2.8372, 2.9291], device='cuda:2'), covar=tensor([0.0219, 0.0382, 0.0516, 0.0404, 0.0205, 0.0304, 0.0210, 0.0233], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0219, 0.0213, 0.0212, 0.0217, 0.0218, 0.0219, 0.0210], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 10:59:47,951 INFO [train.py:904] (2/8) Epoch 17, batch 100, loss[loss=0.1779, simple_loss=0.2637, pruned_loss=0.04606, over 16834.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2698, pruned_loss=0.05048, over 1325482.77 frames. ], batch size: 102, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:00:00,696 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:00:12,326 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:00:26,384 INFO [optim.py:368] (2/8) 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 11:00:36,830 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4267, 3.4526, 3.4630, 2.8545, 3.2451, 2.0247, 3.1805, 2.7506], device='cuda:2'), covar=tensor([0.0133, 0.0104, 0.0158, 0.0185, 0.0086, 0.2219, 0.0125, 0.0212], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0131, 0.0176, 0.0159, 0.0151, 0.0192, 0.0164, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:00:56,544 INFO [train.py:904] (2/8) Epoch 17, batch 150, loss[loss=0.1933, simple_loss=0.2616, pruned_loss=0.06247, over 16861.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2683, pruned_loss=0.04952, over 1767567.73 frames. ], batch size: 96, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:01:23,562 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162572.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:01:44,466 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:02:05,939 INFO [train.py:904] (2/8) Epoch 17, batch 200, loss[loss=0.1637, simple_loss=0.2568, pruned_loss=0.03533, over 17198.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2679, pruned_loss=0.04891, over 2118918.21 frames. ], batch size: 44, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:02:43,590 INFO [optim.py:368] (2/8) 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 11:03:05,789 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0363, 5.5311, 5.6675, 5.4465, 5.5168, 6.0851, 5.6170, 5.3403], device='cuda:2'), covar=tensor([0.0970, 0.1907, 0.2159, 0.2037, 0.2810, 0.1036, 0.1346, 0.2148], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0542, 0.0596, 0.0455, 0.0605, 0.0627, 0.0469, 0.0605], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:03:12,317 INFO [train.py:904] (2/8) Epoch 17, batch 250, loss[loss=0.1654, simple_loss=0.2445, pruned_loss=0.04314, over 16810.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2656, pruned_loss=0.04804, over 2396565.48 frames. ], batch size: 102, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:03:22,247 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1881, 5.5808, 5.3432, 5.3774, 5.0888, 4.9492, 5.0344, 5.6990], device='cuda:2'), covar=tensor([0.1343, 0.1019, 0.1135, 0.0859, 0.0921, 0.0847, 0.1175, 0.0939], device='cuda:2'), in_proj_covar=tensor([0.0608, 0.0750, 0.0609, 0.0551, 0.0473, 0.0482, 0.0626, 0.0577], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:04:20,330 INFO [train.py:904] (2/8) Epoch 17, batch 300, loss[loss=0.1845, simple_loss=0.2608, pruned_loss=0.05412, over 16775.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2624, pruned_loss=0.0467, over 2605871.13 frames. ], batch size: 89, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:04:55,534 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8882, 2.7626, 2.6519, 4.3184, 3.5911, 4.2223, 1.5109, 3.0138], device='cuda:2'), covar=tensor([0.1322, 0.0633, 0.1024, 0.0156, 0.0127, 0.0323, 0.1520, 0.0745], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0164, 0.0187, 0.0168, 0.0194, 0.0209, 0.0193, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:04:59,732 INFO [optim.py:368] (2/8) 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 11:05:03,720 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:05:29,583 INFO [train.py:904] (2/8) Epoch 17, batch 350, loss[loss=0.1613, simple_loss=0.2557, pruned_loss=0.03348, over 17193.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2613, pruned_loss=0.04641, over 2765607.96 frames. ], batch size: 46, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:05:45,417 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2474, 2.1440, 2.3335, 3.8942, 2.1916, 2.5295, 2.2575, 2.3448], device='cuda:2'), covar=tensor([0.1259, 0.3594, 0.2779, 0.0606, 0.3726, 0.2495, 0.3571, 0.3002], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0419, 0.0353, 0.0319, 0.0426, 0.0480, 0.0390, 0.0488], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:06:07,974 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:06:27,112 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9535, 5.5842, 5.7392, 5.4879, 5.5910, 6.1456, 5.5622, 5.3149], device='cuda:2'), covar=tensor([0.1015, 0.1961, 0.2253, 0.2153, 0.2704, 0.0944, 0.1518, 0.2333], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0550, 0.0605, 0.0463, 0.0617, 0.0638, 0.0477, 0.0617], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:06:36,784 INFO [train.py:904] (2/8) Epoch 17, batch 400, loss[loss=0.1771, simple_loss=0.2602, pruned_loss=0.047, over 16532.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2604, pruned_loss=0.04557, over 2898638.94 frames. ], batch size: 68, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:07:10,942 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:07:15,898 INFO [optim.py:368] (2/8) 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 11:07:42,816 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4765, 3.6224, 3.8555, 1.9788, 3.1998, 2.4571, 3.8662, 3.7776], device='cuda:2'), covar=tensor([0.0244, 0.0842, 0.0499, 0.2014, 0.0727, 0.0949, 0.0559, 0.1048], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0149, 0.0139, 0.0126, 0.0139, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:07:46,582 INFO [train.py:904] (2/8) Epoch 17, batch 450, loss[loss=0.1738, simple_loss=0.2672, pruned_loss=0.04023, over 16713.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2576, pruned_loss=0.0446, over 2997869.61 frames. ], batch size: 57, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:08:06,775 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 11:08:34,344 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:35,574 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:55,331 INFO [train.py:904] (2/8) Epoch 17, batch 500, loss[loss=0.167, simple_loss=0.2449, pruned_loss=0.0445, over 16821.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2563, pruned_loss=0.04389, over 3071703.50 frames. ], batch size: 102, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:09:32,594 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:10:01,802 INFO [train.py:904] (2/8) Epoch 17, batch 550, loss[loss=0.2085, simple_loss=0.2875, pruned_loss=0.06477, over 16699.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2556, pruned_loss=0.0438, over 3118855.79 frames. ], batch size: 124, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:10:16,471 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1606, 2.1317, 2.2970, 3.8060, 2.1623, 2.5439, 2.2260, 2.3022], device='cuda:2'), covar=tensor([0.1304, 0.3499, 0.2850, 0.0629, 0.3759, 0.2351, 0.3605, 0.2972], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0420, 0.0354, 0.0321, 0.0427, 0.0483, 0.0391, 0.0492], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:10:29,222 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-04-30 11:11:10,218 INFO [train.py:904] (2/8) Epoch 17, batch 600, loss[loss=0.1657, simple_loss=0.2578, pruned_loss=0.03682, over 17254.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2552, pruned_loss=0.04469, over 3162228.78 frames. ], batch size: 52, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:42,199 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4329, 5.3878, 5.2908, 4.7506, 4.8655, 5.3380, 5.3337, 4.9435], device='cuda:2'), covar=tensor([0.0539, 0.0539, 0.0310, 0.0349, 0.1156, 0.0418, 0.0244, 0.0775], device='cuda:2'), in_proj_covar=tensor([0.0274, 0.0383, 0.0321, 0.0309, 0.0331, 0.0360, 0.0220, 0.0385], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:11:47,294 INFO [optim.py:368] (2/8) 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 11:12:16,979 INFO [train.py:904] (2/8) Epoch 17, batch 650, loss[loss=0.1605, simple_loss=0.2411, pruned_loss=0.03989, over 16677.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2539, pruned_loss=0.04396, over 3198029.12 frames. ], batch size: 134, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:13:09,328 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:13:25,534 INFO [train.py:904] (2/8) Epoch 17, batch 700, loss[loss=0.1683, simple_loss=0.2613, pruned_loss=0.03759, over 17121.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.254, pruned_loss=0.04369, over 3234098.14 frames. ], batch size: 49, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:14:04,486 INFO [optim.py:368] (2/8) 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 11:14:34,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4013, 3.3750, 3.6498, 1.7532, 3.7522, 3.7610, 2.9951, 2.8176], device='cuda:2'), covar=tensor([0.0759, 0.0221, 0.0197, 0.1208, 0.0099, 0.0185, 0.0409, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0106, 0.0093, 0.0139, 0.0074, 0.0118, 0.0126, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:14:34,382 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=163151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:14:35,142 INFO [train.py:904] (2/8) Epoch 17, batch 750, loss[loss=0.1734, simple_loss=0.2512, pruned_loss=0.04777, over 16766.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2539, pruned_loss=0.04405, over 3249618.10 frames. ], batch size: 124, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:14:57,131 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:15:16,212 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-30 11:15:17,984 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:15:44,387 INFO [train.py:904] (2/8) Epoch 17, batch 800, loss[loss=0.1719, simple_loss=0.2668, pruned_loss=0.03849, over 17288.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2537, pruned_loss=0.04369, over 3271540.90 frames. ], batch size: 52, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:16:03,280 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163215.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:23,806 INFO [optim.py:368] (2/8) 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 11:16:36,598 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:40,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3740, 2.2499, 2.3452, 4.2464, 2.2094, 2.6548, 2.3727, 2.4043], device='cuda:2'), covar=tensor([0.1202, 0.3649, 0.2787, 0.0487, 0.3911, 0.2530, 0.3473, 0.3528], device='cuda:2'), in_proj_covar=tensor([0.0383, 0.0422, 0.0354, 0.0323, 0.0428, 0.0485, 0.0393, 0.0493], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:16:53,792 INFO [train.py:904] (2/8) Epoch 17, batch 850, loss[loss=0.1757, simple_loss=0.2536, pruned_loss=0.04893, over 16991.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2544, pruned_loss=0.04345, over 3283423.80 frames. ], batch size: 41, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:17:01,557 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8713, 5.2406, 5.2995, 5.1601, 5.1771, 5.7609, 5.2126, 4.9834], device='cuda:2'), covar=tensor([0.1357, 0.2280, 0.2577, 0.2228, 0.2653, 0.1227, 0.1714, 0.2631], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0562, 0.0621, 0.0473, 0.0631, 0.0655, 0.0490, 0.0634], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:18:01,130 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 17, batch 900, loss[loss=0.1449, simple_loss=0.2321, pruned_loss=0.0289, over 16988.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2541, pruned_loss=0.04287, over 3294237.72 frames. ], batch size: 41, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:18:26,582 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1269, 4.1962, 4.4899, 4.4741, 4.5189, 4.2045, 4.2619, 4.1237], device='cuda:2'), covar=tensor([0.0383, 0.0620, 0.0407, 0.0407, 0.0472, 0.0446, 0.0821, 0.0596], device='cuda:2'), in_proj_covar=tensor([0.0387, 0.0418, 0.0408, 0.0383, 0.0456, 0.0431, 0.0525, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:18:40,396 INFO [optim.py:368] (2/8) 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 11:18:42,637 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6857, 2.5657, 2.1953, 2.5811, 2.9159, 2.7723, 3.3331, 3.2086], device='cuda:2'), covar=tensor([0.0127, 0.0418, 0.0494, 0.0413, 0.0262, 0.0379, 0.0243, 0.0226], device='cuda:2'), in_proj_covar=tensor([0.0190, 0.0230, 0.0221, 0.0220, 0.0228, 0.0230, 0.0234, 0.0222], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:18:52,635 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 11:19:09,573 INFO [train.py:904] (2/8) Epoch 17, batch 950, loss[loss=0.1574, simple_loss=0.2474, pruned_loss=0.0337, over 17117.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2548, pruned_loss=0.04301, over 3312441.47 frames. ], batch size: 47, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:19:33,338 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3357, 4.3504, 4.7227, 4.7066, 4.7188, 4.4271, 4.4471, 4.2711], device='cuda:2'), covar=tensor([0.0426, 0.0779, 0.0476, 0.0575, 0.0587, 0.0567, 0.1031, 0.0643], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0419, 0.0411, 0.0385, 0.0459, 0.0432, 0.0528, 0.0344], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:20:00,164 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8466, 4.0439, 2.5915, 4.6292, 3.2398, 4.6514, 2.7981, 3.3384], device='cuda:2'), covar=tensor([0.0272, 0.0341, 0.1423, 0.0235, 0.0703, 0.0465, 0.1266, 0.0630], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0171, 0.0192, 0.0152, 0.0173, 0.0212, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:20:13,082 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 11:20:17,961 INFO [train.py:904] (2/8) Epoch 17, batch 1000, loss[loss=0.18, simple_loss=0.2477, pruned_loss=0.05618, over 12188.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2535, pruned_loss=0.04286, over 3312848.10 frames. ], batch size: 247, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:20:54,826 INFO [optim.py:368] (2/8) 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 11:20:59,956 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0655, 4.5524, 3.2992, 2.3979, 2.8352, 2.6472, 4.8501, 3.7592], device='cuda:2'), covar=tensor([0.2398, 0.0471, 0.1485, 0.2626, 0.2784, 0.1849, 0.0287, 0.1270], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0262, 0.0295, 0.0298, 0.0286, 0.0241, 0.0281, 0.0320], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:21:18,941 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:21:26,450 INFO [train.py:904] (2/8) Epoch 17, batch 1050, loss[loss=0.1594, simple_loss=0.2577, pruned_loss=0.03049, over 17040.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2528, pruned_loss=0.04256, over 3315973.82 frames. ], batch size: 50, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:22:10,635 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163483.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:22:36,965 INFO [train.py:904] (2/8) Epoch 17, batch 1100, loss[loss=0.1879, simple_loss=0.2632, pruned_loss=0.05632, over 11865.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2523, pruned_loss=0.04232, over 3317109.20 frames. ], batch size: 246, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:22:40,130 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-30 11:23:08,644 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2764, 4.0847, 4.3291, 4.4907, 4.6050, 4.1248, 4.3974, 4.5735], device='cuda:2'), covar=tensor([0.1634, 0.1152, 0.1393, 0.0680, 0.0594, 0.1344, 0.2064, 0.0660], device='cuda:2'), in_proj_covar=tensor([0.0617, 0.0765, 0.0903, 0.0769, 0.0579, 0.0610, 0.0622, 0.0721], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:23:14,822 INFO [optim.py:368] (2/8) 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 11:23:16,187 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:23:43,927 INFO [train.py:904] (2/8) Epoch 17, batch 1150, loss[loss=0.1512, simple_loss=0.2433, pruned_loss=0.02954, over 17212.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2525, pruned_loss=0.04229, over 3325548.67 frames. ], batch size: 44, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:23:46,860 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7497, 3.8847, 2.3228, 4.4109, 2.9788, 4.4539, 2.6503, 3.2314], device='cuda:2'), covar=tensor([0.0314, 0.0361, 0.1607, 0.0282, 0.0790, 0.0439, 0.1335, 0.0637], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0154, 0.0174, 0.0214, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:24:20,078 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 11:24:43,076 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:24:52,261 INFO [train.py:904] (2/8) Epoch 17, batch 1200, loss[loss=0.16, simple_loss=0.2448, pruned_loss=0.0376, over 15538.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2514, pruned_loss=0.04146, over 3334692.42 frames. ], batch size: 190, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:24:58,689 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2507, 4.3857, 4.7446, 4.7243, 4.7532, 4.3927, 4.4551, 4.2780], device='cuda:2'), covar=tensor([0.0464, 0.0744, 0.0468, 0.0482, 0.0424, 0.0483, 0.0821, 0.0676], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0422, 0.0413, 0.0388, 0.0460, 0.0436, 0.0532, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:25:29,955 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 17, batch 1250, loss[loss=0.1804, simple_loss=0.2526, pruned_loss=0.05416, over 16903.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2519, pruned_loss=0.04209, over 3321858.22 frames. ], batch size: 96, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:06,150 INFO [train.py:904] (2/8) Epoch 17, batch 1300, loss[loss=0.1785, simple_loss=0.2692, pruned_loss=0.04393, over 17138.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2519, pruned_loss=0.04191, over 3327016.29 frames. ], batch size: 49, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:44,998 INFO [optim.py:368] (2/8) 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 11:28:08,473 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 17, batch 1350, loss[loss=0.1845, simple_loss=0.2574, pruned_loss=0.05586, over 16913.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2527, pruned_loss=0.04208, over 3328308.74 frames. ], batch size: 116, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:28:48,609 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-30 11:29:15,148 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:29:25,093 INFO [train.py:904] (2/8) Epoch 17, batch 1400, loss[loss=0.1564, simple_loss=0.2381, pruned_loss=0.03735, over 16893.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2525, pruned_loss=0.04174, over 3331019.48 frames. ], batch size: 96, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:29:52,570 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8702, 2.4873, 2.6187, 4.6761, 2.5293, 2.8055, 2.6541, 2.6995], device='cuda:2'), covar=tensor([0.1029, 0.3409, 0.2630, 0.0446, 0.3807, 0.2646, 0.3217, 0.3576], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0424, 0.0357, 0.0325, 0.0429, 0.0490, 0.0395, 0.0497], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:30:05,129 INFO [optim.py:368] (2/8) 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 11:30:36,590 INFO [train.py:904] (2/8) Epoch 17, batch 1450, loss[loss=0.167, simple_loss=0.2457, pruned_loss=0.04413, over 16672.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2518, pruned_loss=0.04168, over 3334907.05 frames. ], batch size: 134, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:31:26,373 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7293, 4.6382, 4.6578, 4.3177, 4.2897, 4.6973, 4.5471, 4.4028], device='cuda:2'), covar=tensor([0.0632, 0.0768, 0.0322, 0.0313, 0.1032, 0.0514, 0.0467, 0.0729], device='cuda:2'), in_proj_covar=tensor([0.0284, 0.0395, 0.0331, 0.0322, 0.0344, 0.0373, 0.0228, 0.0398], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:31:38,105 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:31:45,963 INFO [train.py:904] (2/8) Epoch 17, batch 1500, loss[loss=0.1602, simple_loss=0.2592, pruned_loss=0.03065, over 17262.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2519, pruned_loss=0.04174, over 3343586.69 frames. ], batch size: 45, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:32:24,619 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:32:56,210 INFO [train.py:904] (2/8) Epoch 17, batch 1550, loss[loss=0.1781, simple_loss=0.2732, pruned_loss=0.04153, over 17143.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2531, pruned_loss=0.04267, over 3336620.65 frames. ], batch size: 47, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:33:00,285 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4556, 2.2471, 2.1326, 4.3753, 2.2775, 2.6404, 2.2930, 2.3738], device='cuda:2'), covar=tensor([0.1151, 0.3555, 0.2990, 0.0457, 0.3981, 0.2611, 0.3463, 0.3574], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0426, 0.0360, 0.0328, 0.0432, 0.0493, 0.0397, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:33:14,964 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 11:34:07,076 INFO [train.py:904] (2/8) Epoch 17, batch 1600, loss[loss=0.1702, simple_loss=0.2442, pruned_loss=0.04808, over 16722.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.254, pruned_loss=0.0429, over 3341996.46 frames. ], batch size: 124, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:34:45,104 INFO [optim.py:368] (2/8) 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 11:35:10,863 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8692, 2.3590, 1.6975, 1.9598, 2.7681, 2.5267, 2.9416, 2.8498], device='cuda:2'), covar=tensor([0.0231, 0.0475, 0.0704, 0.0586, 0.0304, 0.0407, 0.0271, 0.0316], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0232, 0.0222, 0.0222, 0.0231, 0.0232, 0.0237, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:35:15,603 INFO [train.py:904] (2/8) Epoch 17, batch 1650, loss[loss=0.1963, simple_loss=0.2861, pruned_loss=0.05322, over 16674.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2552, pruned_loss=0.04341, over 3346490.25 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:35:39,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1978, 3.3215, 3.5939, 2.4607, 3.2255, 3.6186, 3.3662, 2.0804], device='cuda:2'), covar=tensor([0.0471, 0.0139, 0.0050, 0.0344, 0.0105, 0.0092, 0.0085, 0.0417], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0079, 0.0078, 0.0133, 0.0091, 0.0102, 0.0090, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:36:08,266 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:36:24,913 INFO [train.py:904] (2/8) Epoch 17, batch 1700, loss[loss=0.1911, simple_loss=0.2862, pruned_loss=0.04797, over 16788.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2573, pruned_loss=0.04463, over 3346749.79 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:36:40,680 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2577, 4.2785, 4.6980, 4.6300, 4.6639, 4.3464, 4.3870, 4.2256], device='cuda:2'), covar=tensor([0.0391, 0.0639, 0.0350, 0.0442, 0.0524, 0.0436, 0.0836, 0.0685], device='cuda:2'), in_proj_covar=tensor([0.0393, 0.0425, 0.0415, 0.0390, 0.0462, 0.0438, 0.0535, 0.0346], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:36:50,108 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7040, 4.5469, 4.6049, 4.2544, 4.2571, 4.6486, 4.4700, 4.3912], device='cuda:2'), covar=tensor([0.0713, 0.0731, 0.0392, 0.0381, 0.1078, 0.0525, 0.0516, 0.0724], device='cuda:2'), in_proj_covar=tensor([0.0287, 0.0400, 0.0335, 0.0328, 0.0350, 0.0378, 0.0230, 0.0404], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:37:01,951 INFO [optim.py:368] (2/8) 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 11:37:02,620 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3336, 2.1419, 2.3640, 4.0338, 2.1888, 2.5199, 2.2633, 2.3564], device='cuda:2'), covar=tensor([0.1326, 0.3958, 0.2584, 0.0543, 0.3948, 0.2574, 0.3841, 0.3180], device='cuda:2'), in_proj_covar=tensor([0.0387, 0.0427, 0.0360, 0.0327, 0.0433, 0.0493, 0.0397, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:37:06,474 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3907, 3.2907, 3.4827, 1.8854, 3.5840, 3.5833, 2.9510, 2.7753], device='cuda:2'), covar=tensor([0.0716, 0.0209, 0.0169, 0.1105, 0.0096, 0.0207, 0.0388, 0.0399], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0107, 0.0094, 0.0139, 0.0075, 0.0121, 0.0126, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:37:13,408 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0639, 4.1436, 4.4125, 2.0782, 4.5202, 4.6308, 3.3157, 3.5648], device='cuda:2'), covar=tensor([0.0682, 0.0185, 0.0181, 0.1226, 0.0082, 0.0168, 0.0405, 0.0363], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0107, 0.0094, 0.0139, 0.0075, 0.0121, 0.0126, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:37:32,281 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:37:32,928 INFO [train.py:904] (2/8) Epoch 17, batch 1750, loss[loss=0.1953, simple_loss=0.2716, pruned_loss=0.05952, over 16434.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2581, pruned_loss=0.0449, over 3335581.04 frames. ], batch size: 146, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:37:36,762 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:38:06,502 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-30 11:38:41,937 INFO [train.py:904] (2/8) Epoch 17, batch 1800, loss[loss=0.2478, simple_loss=0.3166, pruned_loss=0.08949, over 12148.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2588, pruned_loss=0.04461, over 3332393.67 frames. ], batch size: 246, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:38:58,339 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3965, 2.2194, 2.2650, 4.2050, 2.2025, 2.6242, 2.2948, 2.3987], device='cuda:2'), covar=tensor([0.1215, 0.3727, 0.2935, 0.0512, 0.4222, 0.2552, 0.3577, 0.3338], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0425, 0.0358, 0.0325, 0.0429, 0.0491, 0.0395, 0.0497], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:39:01,672 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:39:07,996 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 11:39:19,765 INFO [optim.py:368] (2/8) 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 11:39:49,954 INFO [train.py:904] (2/8) Epoch 17, batch 1850, loss[loss=0.1829, simple_loss=0.2785, pruned_loss=0.04365, over 17086.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2597, pruned_loss=0.04449, over 3325155.51 frames. ], batch size: 53, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:40:03,078 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 11:40:10,961 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7289, 3.1433, 2.9987, 5.1019, 4.2413, 4.5643, 1.6506, 3.1782], device='cuda:2'), covar=tensor([0.1362, 0.0667, 0.1029, 0.0139, 0.0215, 0.0361, 0.1582, 0.0794], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0172, 0.0198, 0.0211, 0.0192, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:41:01,367 INFO [train.py:904] (2/8) Epoch 17, batch 1900, loss[loss=0.1716, simple_loss=0.2695, pruned_loss=0.03688, over 17010.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2591, pruned_loss=0.04383, over 3322959.93 frames. ], batch size: 50, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:26,274 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8573, 1.9162, 2.3839, 2.7390, 2.7262, 2.7436, 1.9357, 3.0426], device='cuda:2'), covar=tensor([0.0166, 0.0427, 0.0302, 0.0229, 0.0257, 0.0220, 0.0445, 0.0115], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0186, 0.0173, 0.0177, 0.0187, 0.0144, 0.0186, 0.0137], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:41:41,203 INFO [optim.py:368] (2/8) 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 11:41:43,316 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9416, 4.7200, 4.9278, 5.1329, 5.3470, 4.6328, 5.2940, 5.3331], device='cuda:2'), covar=tensor([0.1648, 0.1139, 0.1677, 0.0696, 0.0509, 0.0973, 0.0502, 0.0553], device='cuda:2'), in_proj_covar=tensor([0.0621, 0.0771, 0.0918, 0.0776, 0.0582, 0.0621, 0.0626, 0.0728], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:41:52,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1138, 3.3639, 3.6026, 2.0997, 3.0528, 2.5270, 3.6072, 3.6506], device='cuda:2'), covar=tensor([0.0282, 0.0964, 0.0565, 0.1968, 0.0850, 0.0910, 0.0658, 0.1048], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0159, 0.0164, 0.0151, 0.0142, 0.0128, 0.0142, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:42:12,314 INFO [train.py:904] (2/8) Epoch 17, batch 1950, loss[loss=0.1944, simple_loss=0.2716, pruned_loss=0.05861, over 16700.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2592, pruned_loss=0.04378, over 3317794.40 frames. ], batch size: 134, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:42:24,657 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:42:25,725 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8612, 5.2469, 4.9605, 4.9789, 4.7261, 4.6810, 4.6387, 5.3071], device='cuda:2'), covar=tensor([0.1268, 0.0865, 0.1076, 0.0816, 0.0875, 0.1089, 0.1254, 0.0968], device='cuda:2'), in_proj_covar=tensor([0.0647, 0.0791, 0.0648, 0.0589, 0.0503, 0.0507, 0.0657, 0.0614], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:42:59,745 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4432, 2.2173, 2.2795, 4.2170, 2.1877, 2.6429, 2.2886, 2.4068], device='cuda:2'), covar=tensor([0.1153, 0.3652, 0.2937, 0.0472, 0.4089, 0.2459, 0.3396, 0.3660], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0427, 0.0359, 0.0327, 0.0432, 0.0494, 0.0397, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:43:23,961 INFO [train.py:904] (2/8) Epoch 17, batch 2000, loss[loss=0.1612, simple_loss=0.2381, pruned_loss=0.0421, over 16507.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2587, pruned_loss=0.04391, over 3321226.56 frames. ], batch size: 75, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:43:26,102 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 11:43:51,055 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 11:44:25,526 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:44:32,478 INFO [train.py:904] (2/8) Epoch 17, batch 2050, loss[loss=0.1955, simple_loss=0.2636, pruned_loss=0.06373, over 16774.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2584, pruned_loss=0.0441, over 3322990.58 frames. ], batch size: 124, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:45:41,559 INFO [train.py:904] (2/8) Epoch 17, batch 2100, loss[loss=0.1709, simple_loss=0.2591, pruned_loss=0.0414, over 11597.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2596, pruned_loss=0.04396, over 3322413.46 frames. ], batch size: 247, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:45:54,964 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:46:13,724 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7435, 4.8560, 5.0127, 4.8485, 4.8221, 5.4768, 5.0077, 4.7136], device='cuda:2'), covar=tensor([0.1426, 0.1982, 0.2552, 0.2004, 0.2960, 0.1114, 0.1588, 0.2603], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0570, 0.0630, 0.0483, 0.0651, 0.0665, 0.0497, 0.0646], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:46:20,650 INFO [optim.py:368] (2/8) 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 11:46:50,964 INFO [train.py:904] (2/8) Epoch 17, batch 2150, loss[loss=0.1818, simple_loss=0.255, pruned_loss=0.05433, over 16820.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2608, pruned_loss=0.04522, over 3323886.03 frames. ], batch size: 83, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:47:13,589 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:27,015 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:58,403 INFO [train.py:904] (2/8) Epoch 17, batch 2200, loss[loss=0.1899, simple_loss=0.275, pruned_loss=0.05236, over 16210.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2618, pruned_loss=0.04612, over 3315435.58 frames. ], batch size: 165, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:48:36,637 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:48:37,253 INFO [optim.py:368] (2/8) 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 11:48:44,215 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9071, 2.6217, 2.6446, 2.0394, 2.5535, 2.7005, 2.5581, 1.8537], device='cuda:2'), covar=tensor([0.0409, 0.0091, 0.0078, 0.0343, 0.0117, 0.0121, 0.0116, 0.0385], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0077, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 11:48:44,457 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-30 11:48:50,049 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:48:51,335 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-30 11:48:58,206 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2277, 4.1430, 4.5382, 2.3036, 4.7413, 4.7516, 3.3725, 3.7272], device='cuda:2'), covar=tensor([0.0622, 0.0198, 0.0188, 0.1088, 0.0063, 0.0181, 0.0384, 0.0348], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0095, 0.0139, 0.0075, 0.0122, 0.0127, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 11:49:02,752 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 11:49:06,804 INFO [train.py:904] (2/8) Epoch 17, batch 2250, loss[loss=0.1797, simple_loss=0.2589, pruned_loss=0.05025, over 16464.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2624, pruned_loss=0.04664, over 3302741.66 frames. ], batch size: 75, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:49:15,027 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164658.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:15,835 INFO [train.py:904] (2/8) Epoch 17, batch 2300, loss[loss=0.1551, simple_loss=0.2451, pruned_loss=0.03254, over 17001.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2626, pruned_loss=0.04638, over 3310134.29 frames. ], batch size: 41, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:50:34,545 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:38,843 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 11:51:16,709 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:16,832 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:23,950 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1045, 2.0827, 2.2017, 3.7428, 2.0991, 2.3951, 2.1880, 2.2246], device='cuda:2'), covar=tensor([0.1373, 0.3684, 0.2840, 0.0636, 0.3847, 0.2542, 0.3593, 0.3234], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0429, 0.0359, 0.0327, 0.0431, 0.0496, 0.0398, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:51:24,576 INFO [train.py:904] (2/8) Epoch 17, batch 2350, loss[loss=0.1826, simple_loss=0.258, pruned_loss=0.05356, over 16443.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2633, pruned_loss=0.04743, over 3305980.33 frames. ], batch size: 146, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:23,520 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:24,328 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 11:52:34,376 INFO [train.py:904] (2/8) Epoch 17, batch 2400, loss[loss=0.1894, simple_loss=0.2688, pruned_loss=0.05502, over 16658.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2632, pruned_loss=0.0473, over 3317624.12 frames. ], batch size: 134, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:41,608 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:46,693 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:53:12,670 INFO [optim.py:368] (2/8) 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 11:53:15,526 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6839, 2.5589, 2.4249, 3.9620, 3.2099, 3.9733, 1.4955, 2.8036], device='cuda:2'), covar=tensor([0.1358, 0.0689, 0.1078, 0.0151, 0.0165, 0.0352, 0.1523, 0.0823], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0164, 0.0185, 0.0172, 0.0198, 0.0210, 0.0190, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:53:19,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6923, 3.8422, 2.3977, 4.4232, 3.0761, 4.4183, 2.5255, 3.1708], device='cuda:2'), covar=tensor([0.0292, 0.0345, 0.1420, 0.0278, 0.0710, 0.0457, 0.1315, 0.0713], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0157, 0.0173, 0.0217, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:53:41,603 INFO [train.py:904] (2/8) Epoch 17, batch 2450, loss[loss=0.1546, simple_loss=0.2531, pruned_loss=0.02808, over 17136.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2637, pruned_loss=0.04655, over 3317232.19 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:53:47,271 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6453, 4.9935, 4.7676, 4.7512, 4.5296, 4.4766, 4.4578, 5.0599], device='cuda:2'), covar=tensor([0.1102, 0.0763, 0.0890, 0.0784, 0.0786, 0.1185, 0.1059, 0.0856], device='cuda:2'), in_proj_covar=tensor([0.0647, 0.0797, 0.0649, 0.0593, 0.0506, 0.0509, 0.0662, 0.0618], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:53:51,211 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164859.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:54:46,718 INFO [train.py:904] (2/8) Epoch 17, batch 2500, loss[loss=0.2034, simple_loss=0.2865, pruned_loss=0.06014, over 16253.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2631, pruned_loss=0.04618, over 3322727.63 frames. ], batch size: 165, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:55:12,208 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1849, 2.1831, 2.2897, 3.9640, 2.1641, 2.5360, 2.2296, 2.3289], device='cuda:2'), covar=tensor([0.1349, 0.3589, 0.2687, 0.0567, 0.3777, 0.2465, 0.3645, 0.3067], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0427, 0.0357, 0.0326, 0.0429, 0.0494, 0.0396, 0.0500], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:55:17,576 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:55:26,643 INFO [optim.py:368] (2/8) 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 11:55:29,668 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 11:55:30,296 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:55:55,232 INFO [train.py:904] (2/8) Epoch 17, batch 2550, loss[loss=0.1886, simple_loss=0.2794, pruned_loss=0.04891, over 16507.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2635, pruned_loss=0.04642, over 3319720.80 frames. ], batch size: 68, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:02,086 INFO [train.py:904] (2/8) Epoch 17, batch 2600, loss[loss=0.1687, simple_loss=0.2575, pruned_loss=0.03997, over 16478.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2633, pruned_loss=0.04593, over 3322128.70 frames. ], batch size: 146, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:17,259 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:20,593 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165016.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:31,746 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6263, 2.8645, 2.6351, 4.8557, 3.8633, 4.3507, 1.6449, 3.1043], device='cuda:2'), covar=tensor([0.1385, 0.0751, 0.1224, 0.0176, 0.0246, 0.0382, 0.1525, 0.0787], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0174, 0.0199, 0.0212, 0.0192, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 11:57:41,412 INFO [optim.py:368] (2/8) 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 11:57:43,681 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6514, 4.9944, 4.7194, 4.8004, 4.5180, 4.4763, 4.4887, 5.0513], device='cuda:2'), covar=tensor([0.1252, 0.0922, 0.1121, 0.0803, 0.0929, 0.1263, 0.1200, 0.0999], device='cuda:2'), in_proj_covar=tensor([0.0647, 0.0800, 0.0651, 0.0594, 0.0508, 0.0509, 0.0663, 0.0618], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 11:58:08,461 INFO [train.py:904] (2/8) Epoch 17, batch 2650, loss[loss=0.1848, simple_loss=0.2855, pruned_loss=0.04199, over 16435.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2642, pruned_loss=0.0456, over 3329916.27 frames. ], batch size: 68, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:58:26,258 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:58:29,368 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 11:59:18,038 INFO [train.py:904] (2/8) Epoch 17, batch 2700, loss[loss=0.1659, simple_loss=0.2575, pruned_loss=0.03709, over 16779.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2642, pruned_loss=0.04513, over 3332989.65 frames. ], batch size: 102, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:59:18,391 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:47,017 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 11:59:47,972 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:59,034 INFO [optim.py:368] (2/8) 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 12:00:28,482 INFO [train.py:904] (2/8) Epoch 17, batch 2750, loss[loss=0.1743, simple_loss=0.2651, pruned_loss=0.04169, over 17137.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2641, pruned_loss=0.04441, over 3334979.91 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:00:59,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8537, 2.8983, 2.4377, 4.3474, 3.3755, 4.1778, 1.7045, 2.8921], device='cuda:2'), covar=tensor([0.1311, 0.0706, 0.1264, 0.0181, 0.0245, 0.0418, 0.1568, 0.0841], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0174, 0.0199, 0.0212, 0.0192, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:01:13,240 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:01:38,385 INFO [train.py:904] (2/8) Epoch 17, batch 2800, loss[loss=0.1706, simple_loss=0.2714, pruned_loss=0.03488, over 17226.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2637, pruned_loss=0.04419, over 3331293.86 frames. ], batch size: 52, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:02:10,299 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:02:19,320 INFO [optim.py:368] (2/8) 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 12:02:24,071 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165234.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:02:48,648 INFO [train.py:904] (2/8) Epoch 17, batch 2850, loss[loss=0.1639, simple_loss=0.2651, pruned_loss=0.03136, over 17058.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2636, pruned_loss=0.0442, over 3318169.91 frames. ], batch size: 50, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:03:18,270 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:18,406 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9273, 4.6780, 4.9884, 5.1603, 5.3527, 4.6834, 5.3213, 5.3427], device='cuda:2'), covar=tensor([0.1781, 0.1330, 0.1631, 0.0739, 0.0558, 0.0942, 0.0571, 0.0626], device='cuda:2'), in_proj_covar=tensor([0.0633, 0.0779, 0.0930, 0.0789, 0.0593, 0.0629, 0.0633, 0.0738], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:03:31,430 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:33,884 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2579, 5.8341, 5.9665, 5.7099, 5.7709, 6.3216, 5.8432, 5.5856], device='cuda:2'), covar=tensor([0.0875, 0.1893, 0.2026, 0.1899, 0.2517, 0.0881, 0.1357, 0.2316], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0572, 0.0631, 0.0485, 0.0650, 0.0666, 0.0500, 0.0650], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:03:47,546 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6279, 4.6191, 4.7729, 4.6132, 4.6683, 5.2286, 4.7197, 4.4779], device='cuda:2'), covar=tensor([0.1495, 0.2222, 0.2415, 0.2176, 0.2830, 0.1079, 0.1636, 0.2676], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0571, 0.0631, 0.0484, 0.0649, 0.0665, 0.0499, 0.0649], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:03:58,711 INFO [train.py:904] (2/8) Epoch 17, batch 2900, loss[loss=0.1855, simple_loss=0.2783, pruned_loss=0.0463, over 16697.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2624, pruned_loss=0.04466, over 3319891.84 frames. ], batch size: 62, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:04:17,010 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:04:39,535 INFO [optim.py:368] (2/8) 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 12:05:09,242 INFO [train.py:904] (2/8) Epoch 17, batch 2950, loss[loss=0.1906, simple_loss=0.2797, pruned_loss=0.05074, over 17064.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.262, pruned_loss=0.04493, over 3323937.67 frames. ], batch size: 55, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:05:19,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1420, 5.5975, 5.7910, 5.4794, 5.6311, 6.1565, 5.6515, 5.3924], device='cuda:2'), covar=tensor([0.0941, 0.1945, 0.2094, 0.2326, 0.2608, 0.0977, 0.1420, 0.2417], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0575, 0.0636, 0.0489, 0.0654, 0.0671, 0.0502, 0.0654], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:05:23,902 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:20,615 INFO [train.py:904] (2/8) Epoch 17, batch 3000, loss[loss=0.1645, simple_loss=0.2481, pruned_loss=0.04045, over 15891.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2618, pruned_loss=0.0454, over 3319623.82 frames. ], batch size: 35, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:06:20,616 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 12:06:29,121 INFO [train.py:938] (2/8) Epoch 17, validation: loss=0.1364, simple_loss=0.2422, pruned_loss=0.0153, over 944034.00 frames. 2023-04-30 12:06:29,122 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 12:06:29,378 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:32,068 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5818, 1.6953, 2.2098, 2.4394, 2.5623, 2.5023, 1.8316, 2.7507], device='cuda:2'), covar=tensor([0.0188, 0.0439, 0.0286, 0.0276, 0.0246, 0.0255, 0.0439, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0188, 0.0175, 0.0180, 0.0189, 0.0146, 0.0188, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:07:09,773 INFO [optim.py:368] (2/8) 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 12:07:36,650 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165450.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:07:38,650 INFO [train.py:904] (2/8) Epoch 17, batch 3050, loss[loss=0.1834, simple_loss=0.2789, pruned_loss=0.04399, over 17060.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2615, pruned_loss=0.04506, over 3323078.22 frames. ], batch size: 50, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:08:15,873 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:08:45,828 INFO [train.py:904] (2/8) Epoch 17, batch 3100, loss[loss=0.1825, simple_loss=0.2736, pruned_loss=0.04572, over 17078.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2609, pruned_loss=0.04525, over 3322386.84 frames. ], batch size: 53, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:09:27,660 INFO [optim.py:368] (2/8) 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 12:09:54,979 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:09:55,708 INFO [train.py:904] (2/8) Epoch 17, batch 3150, loss[loss=0.1397, simple_loss=0.2315, pruned_loss=0.02394, over 17159.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.259, pruned_loss=0.04485, over 3327558.39 frames. ], batch size: 46, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:11:06,274 INFO [train.py:904] (2/8) Epoch 17, batch 3200, loss[loss=0.164, simple_loss=0.2532, pruned_loss=0.0374, over 15849.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2583, pruned_loss=0.04457, over 3323210.30 frames. ], batch size: 35, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:11:20,410 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:11:31,173 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8143, 3.7004, 4.0362, 2.0949, 4.3523, 4.3584, 3.1107, 3.3049], device='cuda:2'), covar=tensor([0.0779, 0.0273, 0.0255, 0.1168, 0.0069, 0.0221, 0.0456, 0.0392], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0096, 0.0139, 0.0076, 0.0123, 0.0128, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:11:49,641 INFO [optim.py:368] (2/8) 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 12:12:15,438 INFO [train.py:904] (2/8) Epoch 17, batch 3250, loss[loss=0.1993, simple_loss=0.2746, pruned_loss=0.06195, over 16747.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2586, pruned_loss=0.04499, over 3327250.25 frames. ], batch size: 124, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:13:23,328 INFO [train.py:904] (2/8) Epoch 17, batch 3300, loss[loss=0.2199, simple_loss=0.3006, pruned_loss=0.06966, over 12197.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2595, pruned_loss=0.04525, over 3330978.50 frames. ], batch size: 246, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:13:55,180 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-04-30 12:14:06,777 INFO [optim.py:368] (2/8) 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 12:14:30,233 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165750.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:14:32,749 INFO [train.py:904] (2/8) Epoch 17, batch 3350, loss[loss=0.1472, simple_loss=0.239, pruned_loss=0.02764, over 17188.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2599, pruned_loss=0.04519, over 3319912.87 frames. ], batch size: 46, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:11,077 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165779.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:12,382 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9436, 4.1299, 2.7661, 4.7884, 3.2223, 4.7054, 2.8836, 3.4090], device='cuda:2'), covar=tensor([0.0284, 0.0340, 0.1268, 0.0222, 0.0685, 0.0456, 0.1220, 0.0653], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0159, 0.0174, 0.0219, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:15:27,934 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3593, 2.2553, 2.3505, 4.1758, 2.1440, 2.6277, 2.2891, 2.4482], device='cuda:2'), covar=tensor([0.1210, 0.3332, 0.2643, 0.0549, 0.3854, 0.2406, 0.3561, 0.2978], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0427, 0.0356, 0.0326, 0.0430, 0.0495, 0.0398, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:15:42,358 INFO [train.py:904] (2/8) Epoch 17, batch 3400, loss[loss=0.1835, simple_loss=0.2668, pruned_loss=0.05007, over 16482.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2604, pruned_loss=0.04546, over 3311791.20 frames. ], batch size: 146, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:44,680 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:56,251 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:18,379 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:22,579 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-30 12:16:27,736 INFO [optim.py:368] (2/8) 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 12:16:54,370 INFO [train.py:904] (2/8) Epoch 17, batch 3450, loss[loss=0.1689, simple_loss=0.2491, pruned_loss=0.04433, over 15407.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2589, pruned_loss=0.04423, over 3321889.82 frames. ], batch size: 191, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:17:03,116 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0273, 3.1533, 3.2179, 2.0631, 2.8895, 2.3009, 3.5234, 3.5373], device='cuda:2'), covar=tensor([0.0244, 0.0891, 0.0605, 0.1896, 0.0832, 0.1000, 0.0539, 0.0830], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0160, 0.0164, 0.0150, 0.0141, 0.0128, 0.0141, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:17:11,563 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165864.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:17:29,612 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5708, 3.6541, 2.3003, 3.9280, 2.8931, 3.8537, 2.4063, 2.8727], device='cuda:2'), covar=tensor([0.0259, 0.0379, 0.1368, 0.0283, 0.0678, 0.0693, 0.1265, 0.0673], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0176, 0.0194, 0.0159, 0.0174, 0.0218, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:17:33,042 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4511, 2.1307, 2.2361, 4.3210, 2.0716, 2.4589, 2.2682, 2.2591], device='cuda:2'), covar=tensor([0.1191, 0.4055, 0.2994, 0.0495, 0.4614, 0.2835, 0.3608, 0.4041], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0427, 0.0356, 0.0326, 0.0429, 0.0495, 0.0397, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:18:05,607 INFO [train.py:904] (2/8) Epoch 17, batch 3500, loss[loss=0.2004, simple_loss=0.2718, pruned_loss=0.06456, over 11524.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2575, pruned_loss=0.04339, over 3324461.15 frames. ], batch size: 246, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:18:13,132 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:18:33,823 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-30 12:18:49,909 INFO [optim.py:368] (2/8) 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 12:19:16,033 INFO [train.py:904] (2/8) Epoch 17, batch 3550, loss[loss=0.1453, simple_loss=0.2297, pruned_loss=0.03046, over 16798.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2576, pruned_loss=0.04348, over 3318863.47 frames. ], batch size: 42, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:20:27,926 INFO [train.py:904] (2/8) Epoch 17, batch 3600, loss[loss=0.181, simple_loss=0.2534, pruned_loss=0.05432, over 16898.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2561, pruned_loss=0.04333, over 3324952.52 frames. ], batch size: 109, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:20:32,751 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8418, 4.9875, 5.1450, 4.9618, 4.9800, 5.6295, 5.1410, 4.8414], device='cuda:2'), covar=tensor([0.1175, 0.2056, 0.2147, 0.2077, 0.2789, 0.0998, 0.1578, 0.2513], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0579, 0.0636, 0.0489, 0.0656, 0.0669, 0.0500, 0.0655], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:21:08,665 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 12:21:12,107 INFO [optim.py:368] (2/8) 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 12:21:23,636 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 12:21:40,966 INFO [train.py:904] (2/8) Epoch 17, batch 3650, loss[loss=0.1551, simple_loss=0.227, pruned_loss=0.04154, over 16387.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2551, pruned_loss=0.04374, over 3321613.96 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:21:44,517 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:22:48,276 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 12:22:55,839 INFO [train.py:904] (2/8) Epoch 17, batch 3700, loss[loss=0.1668, simple_loss=0.2414, pruned_loss=0.04616, over 16775.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2537, pruned_loss=0.04505, over 3302891.49 frames. ], batch size: 124, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:23:02,184 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:23:15,464 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 12:24:03,646 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166147.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:10,329 INFO [train.py:904] (2/8) Epoch 17, batch 3750, loss[loss=0.1743, simple_loss=0.2457, pruned_loss=0.05141, over 16572.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2542, pruned_loss=0.04664, over 3287248.79 frames. ], batch size: 75, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:24:17,262 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-30 12:24:21,421 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:24,422 INFO [train.py:904] (2/8) Epoch 17, batch 3800, loss[loss=0.1892, simple_loss=0.265, pruned_loss=0.05671, over 16408.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2554, pruned_loss=0.04754, over 3281285.79 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:25:32,324 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:33,815 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:26:10,618 INFO [optim.py:368] (2/8) 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 12:26:33,889 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 12:26:38,287 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-30 12:26:38,503 INFO [train.py:904] (2/8) Epoch 17, batch 3850, loss[loss=0.1827, simple_loss=0.2641, pruned_loss=0.05061, over 16689.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2562, pruned_loss=0.04842, over 3282170.86 frames. ], batch size: 134, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:26:44,003 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166255.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:27:02,370 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7102, 3.8642, 2.4480, 4.1862, 2.9913, 4.2756, 2.4051, 2.8702], device='cuda:2'), covar=tensor([0.0237, 0.0310, 0.1325, 0.0216, 0.0622, 0.0393, 0.1317, 0.0717], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0175, 0.0193, 0.0159, 0.0173, 0.0218, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:27:15,362 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 12:27:27,311 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1257, 2.0735, 2.2515, 3.8058, 2.1149, 2.3643, 2.1897, 2.2337], device='cuda:2'), covar=tensor([0.1260, 0.3665, 0.2655, 0.0573, 0.3727, 0.2600, 0.3640, 0.3064], device='cuda:2'), in_proj_covar=tensor([0.0392, 0.0432, 0.0358, 0.0329, 0.0431, 0.0500, 0.0401, 0.0507], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:27:52,934 INFO [train.py:904] (2/8) Epoch 17, batch 3900, loss[loss=0.1741, simple_loss=0.2596, pruned_loss=0.04432, over 17207.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2563, pruned_loss=0.04938, over 3287094.48 frames. ], batch size: 45, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:28:01,072 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:28:11,230 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4043, 3.7395, 3.9463, 2.7379, 3.5822, 4.0555, 3.7557, 2.3947], device='cuda:2'), covar=tensor([0.0476, 0.0143, 0.0043, 0.0341, 0.0085, 0.0079, 0.0070, 0.0402], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0076, 0.0076, 0.0129, 0.0089, 0.0099, 0.0087, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:28:37,861 INFO [optim.py:368] (2/8) 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 12:29:07,076 INFO [train.py:904] (2/8) Epoch 17, batch 3950, loss[loss=0.1749, simple_loss=0.2555, pruned_loss=0.0472, over 16571.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2564, pruned_loss=0.05014, over 3290722.82 frames. ], batch size: 62, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:29:12,648 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2195, 2.1208, 2.7212, 3.1296, 3.0555, 3.5963, 2.2871, 3.4488], device='cuda:2'), covar=tensor([0.0173, 0.0402, 0.0266, 0.0224, 0.0234, 0.0096, 0.0433, 0.0088], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0189, 0.0174, 0.0179, 0.0189, 0.0146, 0.0189, 0.0139], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:29:25,602 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7473, 4.6814, 4.6752, 4.3757, 4.3586, 4.6861, 4.5045, 4.4756], device='cuda:2'), covar=tensor([0.0605, 0.0614, 0.0302, 0.0266, 0.0870, 0.0454, 0.0477, 0.0631], device='cuda:2'), in_proj_covar=tensor([0.0290, 0.0403, 0.0342, 0.0329, 0.0352, 0.0383, 0.0232, 0.0410], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:29:30,611 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:18,541 INFO [train.py:904] (2/8) Epoch 17, batch 4000, loss[loss=0.1693, simple_loss=0.2521, pruned_loss=0.04325, over 16668.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2561, pruned_loss=0.05035, over 3299021.40 frames. ], batch size: 76, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:30:25,180 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:31,136 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 12:30:31,295 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6596, 2.6753, 1.8035, 2.6506, 2.0622, 2.8700, 2.0463, 2.2847], device='cuda:2'), covar=tensor([0.0281, 0.0316, 0.1406, 0.0194, 0.0682, 0.0314, 0.1215, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0175, 0.0194, 0.0158, 0.0174, 0.0218, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:31:03,720 INFO [optim.py:368] (2/8) 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 12:31:31,404 INFO [train.py:904] (2/8) Epoch 17, batch 4050, loss[loss=0.1774, simple_loss=0.2622, pruned_loss=0.04627, over 16667.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2563, pruned_loss=0.0492, over 3298039.97 frames. ], batch size: 76, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:31:34,056 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:31:41,949 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:09,539 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5539, 5.8804, 5.5869, 5.6753, 5.3046, 5.2177, 5.2914, 5.9847], device='cuda:2'), covar=tensor([0.1149, 0.0776, 0.1047, 0.0764, 0.0818, 0.0607, 0.1036, 0.0775], device='cuda:2'), in_proj_covar=tensor([0.0649, 0.0808, 0.0653, 0.0598, 0.0511, 0.0512, 0.0669, 0.0620], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:32:14,611 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:43,835 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 12:32:44,112 INFO [train.py:904] (2/8) Epoch 17, batch 4100, loss[loss=0.2273, simple_loss=0.3188, pruned_loss=0.06794, over 16246.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2589, pruned_loss=0.04892, over 3288113.74 frames. ], batch size: 165, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:32:46,281 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166503.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:51,044 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:54,560 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:56,717 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6451, 3.8120, 2.3326, 4.3529, 2.9934, 4.3205, 2.6087, 3.0418], device='cuda:2'), covar=tensor([0.0274, 0.0307, 0.1531, 0.0147, 0.0689, 0.0410, 0.1212, 0.0695], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0174, 0.0193, 0.0156, 0.0173, 0.0217, 0.0201, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:33:26,217 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5202, 5.5698, 5.4229, 5.0074, 5.0570, 5.4650, 5.2988, 5.1712], device='cuda:2'), covar=tensor([0.0507, 0.0300, 0.0209, 0.0226, 0.0794, 0.0336, 0.0253, 0.0523], device='cuda:2'), in_proj_covar=tensor([0.0288, 0.0398, 0.0339, 0.0325, 0.0348, 0.0379, 0.0230, 0.0405], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:33:29,263 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:31,669 INFO [optim.py:368] (2/8) 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 12:33:45,785 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:59,691 INFO [train.py:904] (2/8) Epoch 17, batch 4150, loss[loss=0.1776, simple_loss=0.2691, pruned_loss=0.04308, over 16843.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2655, pruned_loss=0.05127, over 3251257.66 frames. ], batch size: 102, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:34:28,430 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7428, 1.7860, 1.6145, 1.4413, 1.8755, 1.5555, 1.6835, 1.9316], device='cuda:2'), covar=tensor([0.0150, 0.0270, 0.0341, 0.0309, 0.0161, 0.0220, 0.0149, 0.0181], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0227, 0.0217, 0.0219, 0.0229, 0.0227, 0.0232, 0.0224], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:34:28,433 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:00,453 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:14,002 INFO [train.py:904] (2/8) Epoch 17, batch 4200, loss[loss=0.2027, simple_loss=0.2925, pruned_loss=0.05644, over 16892.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2723, pruned_loss=0.05294, over 3208466.85 frames. ], batch size: 109, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:00,020 INFO [optim.py:368] (2/8) 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 12:36:17,451 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-30 12:36:27,721 INFO [train.py:904] (2/8) Epoch 17, batch 4250, loss[loss=0.1754, simple_loss=0.2732, pruned_loss=0.03879, over 16678.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2763, pruned_loss=0.05306, over 3194337.28 frames. ], batch size: 76, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:42,893 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:37:39,012 INFO [train.py:904] (2/8) Epoch 17, batch 4300, loss[loss=0.2169, simple_loss=0.3054, pruned_loss=0.06416, over 16297.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2774, pruned_loss=0.05218, over 3195624.39 frames. ], batch size: 165, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:37:48,959 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 12:37:51,446 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:38:24,694 INFO [optim.py:368] (2/8) 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 12:38:28,902 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:38:52,931 INFO [train.py:904] (2/8) Epoch 17, batch 4350, loss[loss=0.2131, simple_loss=0.3, pruned_loss=0.06306, over 16596.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2808, pruned_loss=0.05312, over 3207991.89 frames. ], batch size: 62, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:39:01,857 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:39:10,308 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:39:57,543 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:05,351 INFO [train.py:904] (2/8) Epoch 17, batch 4400, loss[loss=0.2061, simple_loss=0.2914, pruned_loss=0.06037, over 16470.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2825, pruned_loss=0.05391, over 3216443.93 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:40:06,831 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:37,992 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 12:40:55,203 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166851.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:41:16,587 INFO [train.py:904] (2/8) Epoch 17, batch 4450, loss[loss=0.1926, simple_loss=0.2818, pruned_loss=0.05167, over 16607.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2854, pruned_loss=0.0552, over 3215999.65 frames. ], batch size: 62, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:41:36,423 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:07,444 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:26,941 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7737, 3.9385, 2.4440, 4.7331, 3.1083, 4.5436, 2.6218, 3.1341], device='cuda:2'), covar=tensor([0.0253, 0.0291, 0.1501, 0.0093, 0.0711, 0.0383, 0.1259, 0.0640], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0171, 0.0191, 0.0151, 0.0170, 0.0212, 0.0198, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:42:28,858 INFO [train.py:904] (2/8) Epoch 17, batch 4500, loss[loss=0.1911, simple_loss=0.2847, pruned_loss=0.04875, over 16243.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2861, pruned_loss=0.05586, over 3214442.07 frames. ], batch size: 165, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:07,210 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 12:43:40,951 INFO [train.py:904] (2/8) Epoch 17, batch 4550, loss[loss=0.227, simple_loss=0.3172, pruned_loss=0.06837, over 16693.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2871, pruned_loss=0.0569, over 3198944.76 frames. ], batch size: 134, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:57,163 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:35,595 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:53,170 INFO [train.py:904] (2/8) Epoch 17, batch 4600, loss[loss=0.1944, simple_loss=0.2859, pruned_loss=0.05143, over 16793.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.288, pruned_loss=0.0569, over 3198954.59 frames. ], batch size: 96, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:45:07,098 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:45:38,257 INFO [optim.py:368] (2/8) 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 12:45:47,228 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 12:46:05,392 INFO [train.py:904] (2/8) Epoch 17, batch 4650, loss[loss=0.2035, simple_loss=0.2856, pruned_loss=0.06066, over 16764.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2868, pruned_loss=0.05683, over 3205559.83 frames. ], batch size: 124, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:46:32,344 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4088, 5.3724, 5.2705, 4.9222, 4.9606, 5.3308, 5.2499, 5.0506], device='cuda:2'), covar=tensor([0.0486, 0.0307, 0.0214, 0.0210, 0.0830, 0.0260, 0.0198, 0.0500], device='cuda:2'), in_proj_covar=tensor([0.0273, 0.0381, 0.0325, 0.0312, 0.0334, 0.0361, 0.0222, 0.0387], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:47:00,062 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:13,908 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167100.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:16,412 INFO [train.py:904] (2/8) Epoch 17, batch 4700, loss[loss=0.1826, simple_loss=0.2732, pruned_loss=0.04595, over 16542.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2845, pruned_loss=0.05591, over 3212079.41 frames. ], batch size: 68, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:47:18,952 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167104.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:40,921 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5133, 3.4026, 2.8190, 2.2312, 2.4131, 2.3103, 3.7483, 3.2167], device='cuda:2'), covar=tensor([0.2925, 0.0949, 0.1715, 0.2570, 0.2526, 0.1938, 0.0505, 0.1324], device='cuda:2'), in_proj_covar=tensor([0.0318, 0.0261, 0.0296, 0.0300, 0.0291, 0.0242, 0.0283, 0.0321], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:47:42,085 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167120.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:54,232 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 12:48:00,930 INFO [optim.py:368] (2/8) 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 12:48:06,766 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:28,392 INFO [train.py:904] (2/8) Epoch 17, batch 4750, loss[loss=0.205, simple_loss=0.2863, pruned_loss=0.06184, over 12013.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2802, pruned_loss=0.05365, over 3209576.72 frames. ], batch size: 247, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:48:42,057 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:48,041 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:48,188 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:51,254 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2325, 2.1601, 2.2104, 3.9504, 2.2049, 2.5874, 2.2922, 2.3596], device='cuda:2'), covar=tensor([0.1256, 0.3506, 0.2823, 0.0509, 0.3899, 0.2436, 0.3452, 0.3225], device='cuda:2'), in_proj_covar=tensor([0.0389, 0.0429, 0.0354, 0.0324, 0.0430, 0.0496, 0.0397, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:49:16,407 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:19,052 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167187.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:40,730 INFO [train.py:904] (2/8) Epoch 17, batch 4800, loss[loss=0.1837, simple_loss=0.276, pruned_loss=0.04566, over 16451.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2762, pruned_loss=0.05124, over 3218736.16 frames. ], batch size: 75, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:49:58,978 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:27,971 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167235.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:34,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3199, 4.2654, 4.2250, 3.4314, 4.2266, 1.7203, 3.9969, 3.7845], device='cuda:2'), covar=tensor([0.0096, 0.0107, 0.0144, 0.0416, 0.0097, 0.2696, 0.0151, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0140, 0.0187, 0.0171, 0.0161, 0.0197, 0.0176, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:50:58,072 INFO [train.py:904] (2/8) Epoch 17, batch 4850, loss[loss=0.1755, simple_loss=0.2702, pruned_loss=0.0404, over 16573.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2771, pruned_loss=0.05054, over 3207717.71 frames. ], batch size: 68, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:51:22,030 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5350, 2.2141, 1.8205, 2.0523, 2.5712, 2.2691, 2.3142, 2.7549], device='cuda:2'), covar=tensor([0.0151, 0.0419, 0.0524, 0.0442, 0.0234, 0.0380, 0.0187, 0.0239], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0223, 0.0217, 0.0217, 0.0226, 0.0225, 0.0228, 0.0222], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:51:45,652 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7199, 3.9488, 3.1001, 2.2826, 2.6377, 2.5648, 4.1450, 3.5123], device='cuda:2'), covar=tensor([0.2665, 0.0562, 0.1563, 0.2630, 0.2434, 0.1705, 0.0403, 0.1068], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0262, 0.0296, 0.0301, 0.0291, 0.0241, 0.0282, 0.0321], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:51:46,659 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:52:03,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4464, 3.4894, 1.9600, 3.9064, 2.5245, 3.8477, 2.0325, 2.7256], device='cuda:2'), covar=tensor([0.0243, 0.0309, 0.1717, 0.0118, 0.0925, 0.0492, 0.1751, 0.0793], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0169, 0.0187, 0.0148, 0.0169, 0.0208, 0.0195, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:52:12,188 INFO [train.py:904] (2/8) Epoch 17, batch 4900, loss[loss=0.1817, simple_loss=0.2674, pruned_loss=0.04801, over 16715.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2766, pruned_loss=0.04959, over 3200396.60 frames. ], batch size: 76, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:52:55,966 INFO [optim.py:368] (2/8) 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 12:53:24,977 INFO [train.py:904] (2/8) Epoch 17, batch 4950, loss[loss=0.1975, simple_loss=0.2902, pruned_loss=0.05239, over 16425.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2762, pruned_loss=0.04931, over 3204639.99 frames. ], batch size: 68, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:54:22,663 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:54:37,753 INFO [train.py:904] (2/8) Epoch 17, batch 5000, loss[loss=0.1875, simple_loss=0.2723, pruned_loss=0.05132, over 17022.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.278, pruned_loss=0.04946, over 3203138.28 frames. ], batch size: 53, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:02,872 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:19,949 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:46,703 INFO [train.py:904] (2/8) Epoch 17, batch 5050, loss[loss=0.216, simple_loss=0.2988, pruned_loss=0.06657, over 12196.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2782, pruned_loss=0.04918, over 3199418.97 frames. ], batch size: 246, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:51,606 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:57,164 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:58,651 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-30 12:56:08,481 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:43,535 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:56,776 INFO [train.py:904] (2/8) Epoch 17, batch 5100, loss[loss=0.1794, simple_loss=0.2662, pruned_loss=0.0463, over 16557.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2766, pruned_loss=0.04822, over 3206809.89 frames. ], batch size: 62, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:57:04,231 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0297, 2.7356, 2.8258, 2.0763, 2.6387, 2.1051, 2.7203, 2.8886], device='cuda:2'), covar=tensor([0.0284, 0.0753, 0.0533, 0.1683, 0.0768, 0.0885, 0.0598, 0.0767], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0159, 0.0165, 0.0150, 0.0142, 0.0127, 0.0141, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 12:57:38,076 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8526, 2.3059, 1.8713, 2.0770, 2.7012, 2.4271, 2.5935, 2.8463], device='cuda:2'), covar=tensor([0.0128, 0.0410, 0.0491, 0.0450, 0.0234, 0.0338, 0.0203, 0.0249], device='cuda:2'), in_proj_covar=tensor([0.0187, 0.0223, 0.0216, 0.0217, 0.0225, 0.0225, 0.0226, 0.0221], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 12:57:39,789 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 17, batch 5150, loss[loss=0.1789, simple_loss=0.2778, pruned_loss=0.04002, over 16841.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2762, pruned_loss=0.04744, over 3209414.29 frames. ], batch size: 102, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:58:09,498 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6981, 3.9792, 3.1014, 2.2516, 2.6721, 2.5277, 4.2141, 3.5463], device='cuda:2'), covar=tensor([0.2554, 0.0535, 0.1503, 0.2554, 0.2361, 0.1738, 0.0375, 0.1060], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0262, 0.0294, 0.0298, 0.0289, 0.0240, 0.0282, 0.0319], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 12:58:56,841 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:59:14,586 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3048, 4.3401, 4.6503, 4.5828, 4.6036, 4.3294, 4.3201, 4.2154], device='cuda:2'), covar=tensor([0.0308, 0.0554, 0.0345, 0.0447, 0.0462, 0.0377, 0.0884, 0.0480], device='cuda:2'), in_proj_covar=tensor([0.0370, 0.0403, 0.0395, 0.0371, 0.0441, 0.0413, 0.0511, 0.0331], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 12:59:21,203 INFO [train.py:904] (2/8) Epoch 17, batch 5200, loss[loss=0.1468, simple_loss=0.2363, pruned_loss=0.02861, over 16898.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2741, pruned_loss=0.04628, over 3208823.72 frames. ], batch size: 96, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:00:07,261 INFO [optim.py:368] (2/8) 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 13:00:07,563 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167633.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:00:36,320 INFO [train.py:904] (2/8) Epoch 17, batch 5250, loss[loss=0.2052, simple_loss=0.296, pruned_loss=0.05717, over 12377.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2728, pruned_loss=0.04655, over 3189629.90 frames. ], batch size: 246, lr: 3.99e-03, grad_scale: 16.0 2023-04-30 13:01:38,584 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4646, 4.4141, 4.3215, 3.6546, 4.3723, 1.7243, 4.1348, 4.0439], device='cuda:2'), covar=tensor([0.0087, 0.0092, 0.0165, 0.0345, 0.0092, 0.2691, 0.0141, 0.0216], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0140, 0.0185, 0.0170, 0.0160, 0.0196, 0.0174, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:01:48,353 INFO [train.py:904] (2/8) Epoch 17, batch 5300, loss[loss=0.1514, simple_loss=0.2361, pruned_loss=0.03331, over 17126.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2696, pruned_loss=0.04577, over 3184017.39 frames. ], batch size: 49, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:02:12,761 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:02:29,049 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 13:02:35,320 INFO [optim.py:368] (2/8) 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 13:03:01,793 INFO [train.py:904] (2/8) Epoch 17, batch 5350, loss[loss=0.1782, simple_loss=0.2646, pruned_loss=0.04586, over 17174.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2679, pruned_loss=0.04496, over 3202197.27 frames. ], batch size: 46, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:03:08,220 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:13,807 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:23,270 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:41,683 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:53,567 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:14,426 INFO [train.py:904] (2/8) Epoch 17, batch 5400, loss[loss=0.2157, simple_loss=0.2988, pruned_loss=0.0663, over 12301.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2708, pruned_loss=0.04575, over 3202796.91 frames. ], batch size: 246, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:04:17,676 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:24,144 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:37,027 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-30 13:04:43,572 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4039, 3.6607, 3.6110, 2.0497, 3.0445, 2.5875, 3.8454, 3.9445], device='cuda:2'), covar=tensor([0.0250, 0.0727, 0.0661, 0.1993, 0.0866, 0.0896, 0.0604, 0.0918], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0141, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 13:04:44,946 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 13:04:54,049 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 13:05:31,682 INFO [train.py:904] (2/8) Epoch 17, batch 5450, loss[loss=0.2288, simple_loss=0.312, pruned_loss=0.07276, over 16393.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2736, pruned_loss=0.04734, over 3180430.05 frames. ], batch size: 146, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:05:37,003 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3322, 5.6378, 5.3341, 5.4215, 5.1058, 5.0157, 5.0349, 5.7271], device='cuda:2'), covar=tensor([0.1006, 0.0690, 0.0843, 0.0720, 0.0748, 0.0696, 0.1083, 0.0744], device='cuda:2'), in_proj_covar=tensor([0.0617, 0.0760, 0.0619, 0.0560, 0.0479, 0.0486, 0.0630, 0.0589], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:05:54,030 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1747, 4.1805, 4.0803, 3.3946, 4.1210, 1.7296, 3.8998, 3.7010], device='cuda:2'), covar=tensor([0.0130, 0.0114, 0.0174, 0.0366, 0.0103, 0.2751, 0.0153, 0.0219], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0139, 0.0185, 0.0171, 0.0160, 0.0196, 0.0174, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:06:48,592 INFO [train.py:904] (2/8) Epoch 17, batch 5500, loss[loss=0.2354, simple_loss=0.3225, pruned_loss=0.07412, over 16446.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2809, pruned_loss=0.05165, over 3167698.46 frames. ], batch size: 146, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:07:39,670 INFO [optim.py:368] (2/8) 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 13:08:05,268 INFO [train.py:904] (2/8) Epoch 17, batch 5550, loss[loss=0.2102, simple_loss=0.2985, pruned_loss=0.061, over 17238.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.288, pruned_loss=0.05679, over 3153827.60 frames. ], batch size: 52, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:08:40,766 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 13:08:53,049 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7952, 1.3910, 1.6982, 1.6893, 1.8590, 1.8981, 1.5413, 1.7379], device='cuda:2'), covar=tensor([0.0191, 0.0310, 0.0164, 0.0238, 0.0206, 0.0138, 0.0321, 0.0108], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0186, 0.0172, 0.0177, 0.0186, 0.0142, 0.0188, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:09:31,455 INFO [train.py:904] (2/8) Epoch 17, batch 5600, loss[loss=0.2455, simple_loss=0.3251, pruned_loss=0.08297, over 15379.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2934, pruned_loss=0.06163, over 3110912.89 frames. ], batch size: 191, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:10:27,346 INFO [optim.py:368] (2/8) 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 13:10:27,902 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:10:54,415 INFO [train.py:904] (2/8) Epoch 17, batch 5650, loss[loss=0.217, simple_loss=0.3011, pruned_loss=0.0665, over 16871.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2985, pruned_loss=0.06632, over 3076656.66 frames. ], batch size: 109, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:11:30,929 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:11:46,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9477, 2.7221, 2.6288, 2.0575, 2.5801, 2.6736, 2.5460, 1.8805], device='cuda:2'), covar=tensor([0.0380, 0.0063, 0.0075, 0.0310, 0.0109, 0.0116, 0.0106, 0.0384], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0077, 0.0078, 0.0132, 0.0091, 0.0102, 0.0090, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 13:11:50,850 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:04,683 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168096.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:12:12,320 INFO [train.py:904] (2/8) Epoch 17, batch 5700, loss[loss=0.2287, simple_loss=0.316, pruned_loss=0.07066, over 16706.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3014, pruned_loss=0.06866, over 3070004.93 frames. ], batch size: 134, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:12:45,719 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:13:04,294 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:13:31,682 INFO [train.py:904] (2/8) Epoch 17, batch 5750, loss[loss=0.216, simple_loss=0.3025, pruned_loss=0.06471, over 16696.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3039, pruned_loss=0.07008, over 3057537.80 frames. ], batch size: 134, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:14:08,155 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 13:14:50,309 INFO [train.py:904] (2/8) Epoch 17, batch 5800, loss[loss=0.2031, simple_loss=0.2921, pruned_loss=0.05705, over 16157.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3031, pruned_loss=0.06834, over 3051432.74 frames. ], batch size: 165, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:15:41,189 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168233.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:15:43,217 INFO [optim.py:368] (2/8) 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 13:16:05,933 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0196, 2.0804, 2.2606, 3.4902, 2.0587, 2.3778, 2.1890, 2.2385], device='cuda:2'), covar=tensor([0.1231, 0.3382, 0.2628, 0.0554, 0.4054, 0.2377, 0.3403, 0.3141], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0421, 0.0351, 0.0317, 0.0423, 0.0487, 0.0391, 0.0493], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:16:09,621 INFO [train.py:904] (2/8) Epoch 17, batch 5850, loss[loss=0.2198, simple_loss=0.2852, pruned_loss=0.07716, over 11653.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.3003, pruned_loss=0.06601, over 3072710.85 frames. ], batch size: 246, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:16:45,325 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 13:17:19,302 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:17:32,943 INFO [train.py:904] (2/8) Epoch 17, batch 5900, loss[loss=0.2282, simple_loss=0.3111, pruned_loss=0.07268, over 16585.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2997, pruned_loss=0.06527, over 3098854.61 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:18:03,973 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9241, 2.1091, 2.4455, 3.1800, 2.1832, 2.3761, 2.2902, 2.2348], device='cuda:2'), covar=tensor([0.1111, 0.3062, 0.2039, 0.0582, 0.3862, 0.2099, 0.2861, 0.2977], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0420, 0.0349, 0.0316, 0.0422, 0.0485, 0.0390, 0.0491], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:18:28,196 INFO [optim.py:368] (2/8) 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 13:18:52,631 INFO [train.py:904] (2/8) Epoch 17, batch 5950, loss[loss=0.2078, simple_loss=0.3015, pruned_loss=0.05703, over 16415.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.3003, pruned_loss=0.06453, over 3081991.29 frames. ], batch size: 146, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:19:29,693 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168375.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:19:52,882 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168391.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 13:20:12,025 INFO [train.py:904] (2/8) Epoch 17, batch 6000, loss[loss=0.2163, simple_loss=0.2991, pruned_loss=0.06669, over 15377.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2992, pruned_loss=0.06431, over 3073398.45 frames. ], batch size: 190, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:20:12,026 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 13:20:21,973 INFO [train.py:938] (2/8) Epoch 17, validation: loss=0.1535, simple_loss=0.2667, pruned_loss=0.0202, over 944034.00 frames. 2023-04-30 13:20:21,974 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 13:20:53,933 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:20:53,997 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:21:15,436 INFO [optim.py:368] (2/8) 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 13:21:25,834 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6349, 2.6248, 1.8773, 2.6999, 2.1777, 2.7922, 2.0991, 2.3436], device='cuda:2'), covar=tensor([0.0265, 0.0407, 0.1346, 0.0213, 0.0683, 0.0486, 0.1129, 0.0572], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0192, 0.0151, 0.0172, 0.0211, 0.0200, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 13:21:39,921 INFO [train.py:904] (2/8) Epoch 17, batch 6050, loss[loss=0.2204, simple_loss=0.3112, pruned_loss=0.06477, over 16474.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2978, pruned_loss=0.06393, over 3069453.21 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:22:09,421 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:22:26,862 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 13:22:32,242 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5517, 4.5146, 4.3944, 3.6295, 4.4399, 1.6304, 4.1712, 4.1560], device='cuda:2'), covar=tensor([0.0105, 0.0095, 0.0183, 0.0375, 0.0108, 0.2750, 0.0158, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0137, 0.0184, 0.0168, 0.0158, 0.0194, 0.0172, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:22:59,845 INFO [train.py:904] (2/8) Epoch 17, batch 6100, loss[loss=0.2681, simple_loss=0.33, pruned_loss=0.1031, over 11424.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2974, pruned_loss=0.06252, over 3095708.54 frames. ], batch size: 247, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:23:55,308 INFO [optim.py:368] (2/8) 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 13:24:18,357 INFO [train.py:904] (2/8) Epoch 17, batch 6150, loss[loss=0.2354, simple_loss=0.2978, pruned_loss=0.08652, over 11295.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2953, pruned_loss=0.06195, over 3096000.57 frames. ], batch size: 248, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:24:22,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4365, 4.8415, 5.2113, 5.0837, 5.0881, 4.7157, 4.4030, 4.4911], device='cuda:2'), covar=tensor([0.0677, 0.0735, 0.0431, 0.0628, 0.0677, 0.0606, 0.1558, 0.0619], device='cuda:2'), in_proj_covar=tensor([0.0384, 0.0418, 0.0407, 0.0386, 0.0456, 0.0430, 0.0528, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 13:25:17,267 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:25:36,300 INFO [train.py:904] (2/8) Epoch 17, batch 6200, loss[loss=0.189, simple_loss=0.2756, pruned_loss=0.05122, over 16932.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2938, pruned_loss=0.06172, over 3081414.17 frames. ], batch size: 116, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:25:38,566 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7442, 5.0407, 5.2392, 5.0599, 5.1110, 5.6471, 5.1202, 4.9538], device='cuda:2'), covar=tensor([0.1127, 0.1811, 0.2274, 0.1842, 0.2235, 0.0887, 0.1416, 0.2175], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0563, 0.0618, 0.0474, 0.0636, 0.0650, 0.0487, 0.0636], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 13:26:31,025 INFO [optim.py:368] (2/8) 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 13:26:53,705 INFO [train.py:904] (2/8) Epoch 17, batch 6250, loss[loss=0.215, simple_loss=0.292, pruned_loss=0.06901, over 17029.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2935, pruned_loss=0.06139, over 3084558.50 frames. ], batch size: 53, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:27:53,800 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168691.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:28:10,636 INFO [train.py:904] (2/8) Epoch 17, batch 6300, loss[loss=0.2092, simple_loss=0.2999, pruned_loss=0.05923, over 16669.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2933, pruned_loss=0.0609, over 3089329.77 frames. ], batch size: 89, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:29:06,429 INFO [optim.py:368] (2/8) 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 13:29:09,334 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168739.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:29:29,582 INFO [train.py:904] (2/8) Epoch 17, batch 6350, loss[loss=0.2265, simple_loss=0.2999, pruned_loss=0.07653, over 15421.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2954, pruned_loss=0.06331, over 3048022.19 frames. ], batch size: 190, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:29:39,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3031, 3.3495, 3.6320, 1.8993, 3.7424, 3.8098, 2.9185, 2.7891], device='cuda:2'), covar=tensor([0.0819, 0.0216, 0.0172, 0.1124, 0.0074, 0.0148, 0.0365, 0.0451], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0104, 0.0092, 0.0136, 0.0074, 0.0117, 0.0124, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 13:30:46,693 INFO [train.py:904] (2/8) Epoch 17, batch 6400, loss[loss=0.2097, simple_loss=0.2947, pruned_loss=0.06237, over 16891.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2956, pruned_loss=0.06421, over 3050661.30 frames. ], batch size: 116, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:31:42,104 INFO [optim.py:368] (2/8) 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 13:32:03,554 INFO [train.py:904] (2/8) Epoch 17, batch 6450, loss[loss=0.2122, simple_loss=0.2818, pruned_loss=0.07127, over 11387.00 frames. ], tot_loss[loss=0.211, simple_loss=0.295, pruned_loss=0.06349, over 3047508.55 frames. ], batch size: 248, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:32:55,132 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5627, 2.5537, 1.8490, 2.6533, 2.1563, 2.7380, 2.1078, 2.3359], device='cuda:2'), covar=tensor([0.0319, 0.0478, 0.1238, 0.0317, 0.0632, 0.0627, 0.1126, 0.0583], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0170, 0.0191, 0.0150, 0.0171, 0.0210, 0.0198, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 13:33:02,317 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:33:07,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3537, 4.2260, 4.4403, 4.5746, 4.7158, 4.3077, 4.6576, 4.7166], device='cuda:2'), covar=tensor([0.1670, 0.1199, 0.1387, 0.0638, 0.0557, 0.1067, 0.0708, 0.0644], device='cuda:2'), in_proj_covar=tensor([0.0588, 0.0724, 0.0861, 0.0740, 0.0554, 0.0589, 0.0593, 0.0690], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:33:20,845 INFO [train.py:904] (2/8) Epoch 17, batch 6500, loss[loss=0.1835, simple_loss=0.2673, pruned_loss=0.04983, over 17029.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2929, pruned_loss=0.06277, over 3066927.28 frames. ], batch size: 50, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:33:25,236 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2388, 3.0690, 3.3069, 1.6989, 3.4403, 3.5289, 2.8433, 2.6666], device='cuda:2'), covar=tensor([0.0833, 0.0255, 0.0186, 0.1309, 0.0081, 0.0187, 0.0403, 0.0493], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0104, 0.0092, 0.0136, 0.0074, 0.0117, 0.0124, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 13:33:41,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7874, 3.8240, 4.2394, 2.0701, 4.3997, 4.4278, 3.1965, 3.3133], device='cuda:2'), covar=tensor([0.0762, 0.0186, 0.0148, 0.1119, 0.0056, 0.0126, 0.0368, 0.0402], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0104, 0.0092, 0.0136, 0.0074, 0.0118, 0.0124, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 13:33:46,177 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7961, 1.7915, 1.6212, 1.5552, 1.9269, 1.6152, 1.6785, 1.9459], device='cuda:2'), covar=tensor([0.0161, 0.0255, 0.0337, 0.0311, 0.0196, 0.0245, 0.0170, 0.0200], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0222, 0.0214, 0.0216, 0.0224, 0.0222, 0.0223, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:34:15,420 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:34:16,254 INFO [optim.py:368] (2/8) 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 13:34:40,482 INFO [train.py:904] (2/8) Epoch 17, batch 6550, loss[loss=0.2102, simple_loss=0.3062, pruned_loss=0.05714, over 16621.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2963, pruned_loss=0.06389, over 3077630.02 frames. ], batch size: 62, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:35:03,789 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:35:06,317 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:35:17,855 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1764, 2.1637, 2.1758, 3.8808, 2.0680, 2.5512, 2.2533, 2.3581], device='cuda:2'), covar=tensor([0.1241, 0.3420, 0.2662, 0.0485, 0.3912, 0.2290, 0.3311, 0.3113], device='cuda:2'), in_proj_covar=tensor([0.0382, 0.0421, 0.0350, 0.0317, 0.0425, 0.0487, 0.0392, 0.0493], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:35:58,162 INFO [train.py:904] (2/8) Epoch 17, batch 6600, loss[loss=0.2, simple_loss=0.2912, pruned_loss=0.0544, over 16333.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2976, pruned_loss=0.06375, over 3089778.53 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:36:37,602 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:36:40,042 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 13:36:42,775 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 13:36:50,873 INFO [optim.py:368] (2/8) 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 13:37:04,155 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8428, 4.8355, 4.6485, 3.9968, 4.7407, 1.7419, 4.4911, 4.4508], device='cuda:2'), covar=tensor([0.0072, 0.0067, 0.0175, 0.0347, 0.0082, 0.2579, 0.0130, 0.0205], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0138, 0.0185, 0.0169, 0.0158, 0.0196, 0.0172, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:37:13,177 INFO [train.py:904] (2/8) Epoch 17, batch 6650, loss[loss=0.1844, simple_loss=0.2781, pruned_loss=0.04533, over 16900.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2974, pruned_loss=0.06398, over 3098877.79 frames. ], batch size: 96, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:37:51,879 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 13:37:59,817 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:38:30,530 INFO [train.py:904] (2/8) Epoch 17, batch 6700, loss[loss=0.1896, simple_loss=0.2719, pruned_loss=0.05366, over 16624.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2956, pruned_loss=0.06371, over 3094763.93 frames. ], batch size: 57, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:38:53,685 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:08,812 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 13:39:23,140 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1915, 2.9799, 3.3132, 1.6042, 3.4346, 3.5068, 2.6837, 2.5960], device='cuda:2'), covar=tensor([0.0854, 0.0253, 0.0162, 0.1300, 0.0078, 0.0171, 0.0449, 0.0477], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0104, 0.0093, 0.0136, 0.0074, 0.0118, 0.0124, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 13:39:26,532 INFO [optim.py:368] (2/8) 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 13:39:34,897 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:47,624 INFO [train.py:904] (2/8) Epoch 17, batch 6750, loss[loss=0.2188, simple_loss=0.2931, pruned_loss=0.07231, over 11905.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2945, pruned_loss=0.06371, over 3091785.33 frames. ], batch size: 248, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:39:58,623 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:40:26,200 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:41:04,381 INFO [train.py:904] (2/8) Epoch 17, batch 6800, loss[loss=0.2084, simple_loss=0.3093, pruned_loss=0.05373, over 16920.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2944, pruned_loss=0.06306, over 3104278.12 frames. ], batch size: 96, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:41:33,126 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:42:02,686 INFO [optim.py:368] (2/8) 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 13:42:09,742 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1147, 1.8660, 2.5216, 3.0741, 2.8846, 3.4041, 2.0179, 3.4492], device='cuda:2'), covar=tensor([0.0160, 0.0461, 0.0297, 0.0216, 0.0235, 0.0151, 0.0504, 0.0125], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0184, 0.0170, 0.0173, 0.0184, 0.0141, 0.0186, 0.0136], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:42:23,107 INFO [train.py:904] (2/8) Epoch 17, batch 6850, loss[loss=0.1765, simple_loss=0.2745, pruned_loss=0.03919, over 16793.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2952, pruned_loss=0.06342, over 3114603.01 frames. ], batch size: 39, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:43:37,900 INFO [train.py:904] (2/8) Epoch 17, batch 6900, loss[loss=0.2005, simple_loss=0.2937, pruned_loss=0.05363, over 16706.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2974, pruned_loss=0.06298, over 3113945.01 frames. ], batch size: 134, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:43:45,420 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2362, 2.0180, 1.6506, 1.7563, 2.2632, 1.9669, 2.0395, 2.4023], device='cuda:2'), covar=tensor([0.0171, 0.0371, 0.0481, 0.0425, 0.0231, 0.0350, 0.0184, 0.0228], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0224, 0.0216, 0.0218, 0.0224, 0.0223, 0.0224, 0.0218], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:44:10,362 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:44:14,314 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 13:44:32,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4146, 2.9389, 2.7097, 2.2525, 2.3050, 2.2629, 2.9144, 2.9478], device='cuda:2'), covar=tensor([0.2163, 0.0740, 0.1401, 0.2149, 0.2023, 0.1865, 0.0442, 0.1014], device='cuda:2'), in_proj_covar=tensor([0.0318, 0.0260, 0.0296, 0.0299, 0.0289, 0.0241, 0.0283, 0.0320], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 13:44:33,263 INFO [optim.py:368] (2/8) 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 13:44:55,623 INFO [train.py:904] (2/8) Epoch 17, batch 6950, loss[loss=0.1946, simple_loss=0.2901, pruned_loss=0.04959, over 16772.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2995, pruned_loss=0.06474, over 3105189.20 frames. ], batch size: 89, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:45:23,791 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 13:46:11,848 INFO [train.py:904] (2/8) Epoch 17, batch 7000, loss[loss=0.1943, simple_loss=0.2941, pruned_loss=0.04724, over 16268.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2993, pruned_loss=0.06363, over 3101547.10 frames. ], batch size: 165, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:47:07,101 INFO [optim.py:368] (2/8) 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 13:47:07,471 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:47:28,661 INFO [train.py:904] (2/8) Epoch 17, batch 7050, loss[loss=0.2085, simple_loss=0.2989, pruned_loss=0.05905, over 16437.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.3005, pruned_loss=0.06431, over 3077718.70 frames. ], batch size: 146, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:48:00,527 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:48:43,058 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7423, 5.0084, 4.7711, 4.7698, 4.5106, 4.5625, 4.4695, 5.1128], device='cuda:2'), covar=tensor([0.1130, 0.0891, 0.1021, 0.0859, 0.0847, 0.0980, 0.1108, 0.0874], device='cuda:2'), in_proj_covar=tensor([0.0626, 0.0765, 0.0623, 0.0568, 0.0480, 0.0490, 0.0633, 0.0591], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:48:47,181 INFO [train.py:904] (2/8) Epoch 17, batch 7100, loss[loss=0.2129, simple_loss=0.2887, pruned_loss=0.06852, over 16700.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2981, pruned_loss=0.0634, over 3088181.02 frames. ], batch size: 62, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:49:08,053 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:49:42,142 INFO [optim.py:368] (2/8) 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 13:50:02,978 INFO [train.py:904] (2/8) Epoch 17, batch 7150, loss[loss=0.2097, simple_loss=0.2987, pruned_loss=0.06031, over 16444.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2959, pruned_loss=0.06251, over 3106021.18 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:50:14,841 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2844, 4.3932, 4.7098, 4.6552, 4.6800, 4.3246, 4.3623, 4.2230], device='cuda:2'), covar=tensor([0.0352, 0.0525, 0.0354, 0.0427, 0.0479, 0.0434, 0.0990, 0.0553], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0413, 0.0404, 0.0382, 0.0454, 0.0428, 0.0523, 0.0338], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 13:50:49,512 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 13:51:19,558 INFO [train.py:904] (2/8) Epoch 17, batch 7200, loss[loss=0.2003, simple_loss=0.2864, pruned_loss=0.05709, over 15413.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2939, pruned_loss=0.06125, over 3094049.26 frames. ], batch size: 191, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:51:48,668 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:51,358 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:55,193 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:51:55,655 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-30 13:52:16,449 INFO [optim.py:368] (2/8) 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 13:52:38,958 INFO [train.py:904] (2/8) Epoch 17, batch 7250, loss[loss=0.2125, simple_loss=0.2892, pruned_loss=0.06793, over 11541.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.291, pruned_loss=0.05971, over 3096519.16 frames. ], batch size: 247, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:53:08,591 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:53:25,511 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:57,113 INFO [train.py:904] (2/8) Epoch 17, batch 7300, loss[loss=0.1969, simple_loss=0.2895, pruned_loss=0.05215, over 16373.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2908, pruned_loss=0.05985, over 3089439.69 frames. ], batch size: 146, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:54:41,385 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:54:55,061 INFO [optim.py:368] (2/8) 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 13:54:55,481 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:10,696 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:17,627 INFO [train.py:904] (2/8) Epoch 17, batch 7350, loss[loss=0.2899, simple_loss=0.347, pruned_loss=0.1164, over 11011.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2916, pruned_loss=0.06087, over 3069979.79 frames. ], batch size: 250, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:55:49,789 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:04,019 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 13:56:12,135 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,315 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:19,805 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:31,750 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8964, 4.1307, 3.9301, 4.0206, 3.6633, 3.7729, 3.8268, 4.1119], device='cuda:2'), covar=tensor([0.1000, 0.0871, 0.1004, 0.0747, 0.0754, 0.1543, 0.0896, 0.0965], device='cuda:2'), in_proj_covar=tensor([0.0620, 0.0757, 0.0618, 0.0562, 0.0477, 0.0488, 0.0625, 0.0586], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:56:37,950 INFO [train.py:904] (2/8) Epoch 17, batch 7400, loss[loss=0.1925, simple_loss=0.285, pruned_loss=0.05002, over 16740.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2929, pruned_loss=0.06149, over 3051005.09 frames. ], batch size: 83, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:56:48,634 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:59,912 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:06,531 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 13:57:07,983 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:36,684 INFO [optim.py:368] (2/8) 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 13:57:45,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9929, 5.2514, 4.9618, 5.0436, 4.7764, 4.7485, 4.6986, 5.2977], device='cuda:2'), covar=tensor([0.1029, 0.0839, 0.1031, 0.0842, 0.0817, 0.0960, 0.1100, 0.0995], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0758, 0.0620, 0.0562, 0.0478, 0.0488, 0.0626, 0.0587], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:57:52,287 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:54,586 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 13:57:55,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9046, 4.1630, 3.9544, 4.0544, 3.7137, 3.7946, 3.8764, 4.1283], device='cuda:2'), covar=tensor([0.1057, 0.0969, 0.1054, 0.0833, 0.0815, 0.1632, 0.0932, 0.1065], device='cuda:2'), in_proj_covar=tensor([0.0623, 0.0758, 0.0621, 0.0563, 0.0478, 0.0489, 0.0627, 0.0588], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 13:57:59,464 INFO [train.py:904] (2/8) Epoch 17, batch 7450, loss[loss=0.2044, simple_loss=0.2833, pruned_loss=0.06272, over 16234.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2932, pruned_loss=0.06198, over 3047553.97 frames. ], batch size: 35, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 13:58:04,768 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3840, 3.1343, 2.5784, 2.0456, 2.1828, 2.1729, 3.3286, 3.0315], device='cuda:2'), covar=tensor([0.3082, 0.0934, 0.1867, 0.2697, 0.2436, 0.2155, 0.0588, 0.1286], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0263, 0.0301, 0.0303, 0.0292, 0.0245, 0.0287, 0.0323], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 13:58:18,999 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:58:20,668 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 13:58:56,813 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9706, 3.2576, 3.0632, 5.2122, 4.1608, 4.5514, 1.8820, 3.3991], device='cuda:2'), covar=tensor([0.1312, 0.0684, 0.1124, 0.0154, 0.0387, 0.0388, 0.1620, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0174, 0.0202, 0.0211, 0.0194, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 13:59:20,320 INFO [train.py:904] (2/8) Epoch 17, batch 7500, loss[loss=0.2016, simple_loss=0.2874, pruned_loss=0.05791, over 16918.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2935, pruned_loss=0.06098, over 3059383.83 frames. ], batch size: 109, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 13:59:34,678 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0077, 3.4924, 3.5050, 2.2023, 3.2062, 3.4767, 3.3238, 1.8841], device='cuda:2'), covar=tensor([0.0532, 0.0047, 0.0055, 0.0411, 0.0101, 0.0114, 0.0088, 0.0471], device='cuda:2'), in_proj_covar=tensor([0.0132, 0.0075, 0.0077, 0.0130, 0.0089, 0.0101, 0.0088, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 13:59:50,449 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 14:00:05,224 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7978, 3.7177, 3.8586, 3.9783, 4.0487, 3.7099, 4.0088, 4.0878], device='cuda:2'), covar=tensor([0.1577, 0.1121, 0.1315, 0.0688, 0.0613, 0.1809, 0.0884, 0.0727], device='cuda:2'), in_proj_covar=tensor([0.0583, 0.0722, 0.0858, 0.0737, 0.0558, 0.0587, 0.0597, 0.0689], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:00:17,626 INFO [optim.py:368] (2/8) 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 14:00:37,735 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 14:00:39,248 INFO [train.py:904] (2/8) Epoch 17, batch 7550, loss[loss=0.2119, simple_loss=0.2932, pruned_loss=0.06528, over 15259.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2927, pruned_loss=0.061, over 3069671.14 frames. ], batch size: 191, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:00:50,181 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 14:00:51,348 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 14:01:19,598 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:01:27,690 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 14:02:00,832 INFO [train.py:904] (2/8) Epoch 17, batch 7600, loss[loss=0.2008, simple_loss=0.2849, pruned_loss=0.05829, over 16267.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2922, pruned_loss=0.06167, over 3072147.14 frames. ], batch size: 35, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:02:28,045 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0264, 3.0714, 1.8242, 3.2191, 2.3278, 3.3023, 2.0059, 2.4587], device='cuda:2'), covar=tensor([0.0278, 0.0404, 0.1665, 0.0201, 0.0816, 0.0615, 0.1588, 0.0818], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0170, 0.0193, 0.0150, 0.0172, 0.0211, 0.0200, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 14:02:45,362 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7345, 4.0522, 3.1668, 2.3771, 2.8009, 2.6341, 4.3331, 3.6130], device='cuda:2'), covar=tensor([0.2799, 0.0617, 0.1586, 0.2480, 0.2469, 0.1822, 0.0448, 0.1162], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0263, 0.0300, 0.0303, 0.0292, 0.0245, 0.0288, 0.0323], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:02:58,188 INFO [optim.py:368] (2/8) 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 14:03:05,019 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1467, 5.1917, 4.9445, 4.3010, 5.0726, 1.8026, 4.7956, 4.7546], device='cuda:2'), covar=tensor([0.0077, 0.0063, 0.0170, 0.0343, 0.0078, 0.2616, 0.0097, 0.0176], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0136, 0.0183, 0.0168, 0.0156, 0.0194, 0.0170, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:03:08,020 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170045.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:03:17,364 INFO [train.py:904] (2/8) Epoch 17, batch 7650, loss[loss=0.2372, simple_loss=0.3134, pruned_loss=0.08049, over 15346.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2924, pruned_loss=0.06206, over 3078498.40 frames. ], batch size: 190, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:07,613 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:12,435 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4827, 3.5760, 2.0971, 4.0247, 2.7132, 3.9954, 2.1624, 2.7534], device='cuda:2'), covar=tensor([0.0275, 0.0356, 0.1625, 0.0172, 0.0774, 0.0487, 0.1585, 0.0760], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0170, 0.0192, 0.0149, 0.0172, 0.0210, 0.0199, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 14:04:33,137 INFO [train.py:904] (2/8) Epoch 17, batch 7700, loss[loss=0.2591, simple_loss=0.3205, pruned_loss=0.09885, over 11563.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2926, pruned_loss=0.06247, over 3078344.83 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:34,716 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:39,597 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:20,822 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9335, 5.3079, 5.5072, 5.3376, 5.3243, 5.9153, 5.4025, 5.2388], device='cuda:2'), covar=tensor([0.0987, 0.1691, 0.1893, 0.1683, 0.2340, 0.0870, 0.1480, 0.2132], device='cuda:2'), in_proj_covar=tensor([0.0387, 0.0559, 0.0615, 0.0468, 0.0632, 0.0646, 0.0487, 0.0629], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:05:29,557 INFO [optim.py:368] (2/8) 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 14:05:34,958 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:50,026 INFO [train.py:904] (2/8) Epoch 17, batch 7750, loss[loss=0.2455, simple_loss=0.3086, pruned_loss=0.09124, over 11591.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.293, pruned_loss=0.06285, over 3052316.67 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:05:52,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5983, 2.9692, 3.1346, 1.9741, 2.7091, 2.0560, 3.1885, 3.1913], device='cuda:2'), covar=tensor([0.0304, 0.0847, 0.0599, 0.2015, 0.0924, 0.1062, 0.0713, 0.1025], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0158, 0.0163, 0.0150, 0.0142, 0.0126, 0.0141, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 14:06:28,170 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 14:07:07,224 INFO [train.py:904] (2/8) Epoch 17, batch 7800, loss[loss=0.1859, simple_loss=0.2849, pruned_loss=0.04344, over 16408.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2948, pruned_loss=0.06416, over 3043807.81 frames. ], batch size: 68, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:07:41,338 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8820, 2.6747, 2.6198, 1.9205, 2.5430, 2.6460, 2.5758, 1.9440], device='cuda:2'), covar=tensor([0.0409, 0.0082, 0.0075, 0.0342, 0.0119, 0.0132, 0.0122, 0.0371], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0076, 0.0078, 0.0132, 0.0091, 0.0102, 0.0089, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:08:03,352 INFO [optim.py:368] (2/8) 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 14:08:23,347 INFO [train.py:904] (2/8) Epoch 17, batch 7850, loss[loss=0.1967, simple_loss=0.2872, pruned_loss=0.05311, over 16830.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2957, pruned_loss=0.06378, over 3048265.81 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:08:34,035 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1703, 4.0815, 4.2916, 4.4179, 4.5197, 4.1261, 4.4230, 4.5307], device='cuda:2'), covar=tensor([0.1824, 0.1205, 0.1396, 0.0679, 0.0579, 0.1306, 0.0855, 0.0673], device='cuda:2'), in_proj_covar=tensor([0.0584, 0.0720, 0.0855, 0.0732, 0.0556, 0.0583, 0.0594, 0.0689], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:08:55,543 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 14:09:01,029 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:09:31,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3260, 4.3873, 4.2038, 3.9117, 3.8905, 4.2829, 3.9921, 4.0166], device='cuda:2'), covar=tensor([0.0601, 0.0558, 0.0295, 0.0291, 0.0801, 0.0490, 0.0712, 0.0635], device='cuda:2'), in_proj_covar=tensor([0.0266, 0.0377, 0.0316, 0.0302, 0.0324, 0.0352, 0.0217, 0.0376], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:09:38,277 INFO [train.py:904] (2/8) Epoch 17, batch 7900, loss[loss=0.2035, simple_loss=0.2918, pruned_loss=0.05756, over 16485.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2949, pruned_loss=0.06319, over 3058202.16 frames. ], batch size: 68, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:10:12,457 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:10:36,114 INFO [optim.py:368] (2/8) 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 14:10:55,565 INFO [train.py:904] (2/8) Epoch 17, batch 7950, loss[loss=0.2019, simple_loss=0.2787, pruned_loss=0.06256, over 16537.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2948, pruned_loss=0.06351, over 3070792.49 frames. ], batch size: 62, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:11:49,332 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:05,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4246, 3.4754, 2.6953, 2.0283, 2.2598, 2.1693, 3.6417, 3.2178], device='cuda:2'), covar=tensor([0.3134, 0.0789, 0.1894, 0.3047, 0.2915, 0.2204, 0.0555, 0.1356], device='cuda:2'), in_proj_covar=tensor([0.0320, 0.0260, 0.0297, 0.0299, 0.0288, 0.0242, 0.0285, 0.0320], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:12:13,319 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:14,139 INFO [train.py:904] (2/8) Epoch 17, batch 8000, loss[loss=0.2202, simple_loss=0.3032, pruned_loss=0.06858, over 15270.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2949, pruned_loss=0.06348, over 3081027.38 frames. ], batch size: 190, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:12:15,831 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170403.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:00,979 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 14:13:15,263 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:28,431 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170451.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:29,849 INFO [train.py:904] (2/8) Epoch 17, batch 8050, loss[loss=0.1828, simple_loss=0.2807, pruned_loss=0.04242, over 16798.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2947, pruned_loss=0.06321, over 3086799.60 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:13:44,342 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170461.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:28,129 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:34,635 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4581, 3.5393, 2.1594, 3.8500, 2.6683, 3.9299, 2.1832, 2.8156], device='cuda:2'), covar=tensor([0.0249, 0.0333, 0.1567, 0.0210, 0.0743, 0.0497, 0.1531, 0.0692], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0169, 0.0191, 0.0149, 0.0171, 0.0209, 0.0198, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 14:14:46,137 INFO [train.py:904] (2/8) Epoch 17, batch 8100, loss[loss=0.1908, simple_loss=0.2857, pruned_loss=0.04792, over 16682.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2939, pruned_loss=0.06244, over 3094481.82 frames. ], batch size: 83, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:15:16,586 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170522.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 14:15:42,385 INFO [optim.py:368] (2/8) 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 14:16:00,983 INFO [train.py:904] (2/8) Epoch 17, batch 8150, loss[loss=0.177, simple_loss=0.266, pruned_loss=0.04397, over 17258.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.291, pruned_loss=0.06108, over 3109705.13 frames. ], batch size: 52, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:16:23,472 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 14:16:27,579 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170569.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:17:16,333 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6293, 4.7071, 5.0805, 5.0380, 5.0012, 4.6986, 4.6998, 4.5007], device='cuda:2'), covar=tensor([0.0299, 0.0542, 0.0381, 0.0381, 0.0509, 0.0376, 0.0946, 0.0498], device='cuda:2'), in_proj_covar=tensor([0.0377, 0.0411, 0.0405, 0.0381, 0.0454, 0.0424, 0.0522, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 14:17:18,489 INFO [train.py:904] (2/8) Epoch 17, batch 8200, loss[loss=0.1723, simple_loss=0.2685, pruned_loss=0.03805, over 16722.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2887, pruned_loss=0.06017, over 3124567.39 frames. ], batch size: 83, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:18:05,960 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:18:21,306 INFO [optim.py:368] (2/8) 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 14:18:41,069 INFO [train.py:904] (2/8) Epoch 17, batch 8250, loss[loss=0.1871, simple_loss=0.2711, pruned_loss=0.05156, over 12240.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2874, pruned_loss=0.05792, over 3093610.44 frames. ], batch size: 247, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:19:05,648 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7515, 1.8836, 2.2934, 2.7737, 2.6505, 3.1005, 2.1733, 3.0716], device='cuda:2'), covar=tensor([0.0200, 0.0470, 0.0323, 0.0268, 0.0265, 0.0194, 0.0398, 0.0142], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0184, 0.0171, 0.0173, 0.0183, 0.0142, 0.0186, 0.0135], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:20:04,049 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:20:05,298 INFO [train.py:904] (2/8) Epoch 17, batch 8300, loss[loss=0.1762, simple_loss=0.2755, pruned_loss=0.03844, over 16739.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2854, pruned_loss=0.05535, over 3079098.44 frames. ], batch size: 124, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:21:09,571 INFO [optim.py:368] (2/8) 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 14:21:16,625 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8313, 3.8162, 3.9555, 3.7300, 3.8790, 4.2919, 3.9282, 3.6212], device='cuda:2'), covar=tensor([0.1920, 0.2239, 0.2287, 0.2518, 0.2872, 0.1602, 0.1555, 0.2672], device='cuda:2'), in_proj_covar=tensor([0.0378, 0.0546, 0.0600, 0.0459, 0.0615, 0.0634, 0.0474, 0.0614], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:21:25,415 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:21:29,852 INFO [train.py:904] (2/8) Epoch 17, batch 8350, loss[loss=0.196, simple_loss=0.2951, pruned_loss=0.04841, over 16636.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2854, pruned_loss=0.05408, over 3063951.14 frames. ], batch size: 134, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:21:35,139 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3674, 3.4488, 3.6816, 3.6603, 3.6632, 3.4769, 3.5164, 3.5687], device='cuda:2'), covar=tensor([0.0384, 0.0837, 0.0527, 0.0484, 0.0487, 0.0574, 0.0893, 0.0555], device='cuda:2'), in_proj_covar=tensor([0.0375, 0.0409, 0.0401, 0.0379, 0.0451, 0.0421, 0.0519, 0.0333], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 14:22:51,857 INFO [train.py:904] (2/8) Epoch 17, batch 8400, loss[loss=0.1734, simple_loss=0.2597, pruned_loss=0.04356, over 12206.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2823, pruned_loss=0.05171, over 3066765.02 frames. ], batch size: 246, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:23:16,980 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170817.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 14:23:25,224 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0556, 3.7901, 4.2045, 2.1006, 4.4379, 4.4666, 3.2208, 3.3881], device='cuda:2'), covar=tensor([0.0602, 0.0203, 0.0239, 0.1172, 0.0050, 0.0116, 0.0366, 0.0376], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0102, 0.0090, 0.0134, 0.0072, 0.0115, 0.0121, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 14:23:52,184 INFO [optim.py:368] (2/8) 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 14:24:10,747 INFO [train.py:904] (2/8) Epoch 17, batch 8450, loss[loss=0.1648, simple_loss=0.2594, pruned_loss=0.03507, over 17035.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2807, pruned_loss=0.04996, over 3076162.75 frames. ], batch size: 55, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:25:32,473 INFO [train.py:904] (2/8) Epoch 17, batch 8500, loss[loss=0.1561, simple_loss=0.2416, pruned_loss=0.03535, over 11930.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2773, pruned_loss=0.04779, over 3077852.42 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:25:52,622 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 14:26:10,265 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:14,649 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:34,258 INFO [optim.py:368] (2/8) 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 14:26:56,006 INFO [train.py:904] (2/8) Epoch 17, batch 8550, loss[loss=0.1931, simple_loss=0.2865, pruned_loss=0.04987, over 16213.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2746, pruned_loss=0.04677, over 3044774.19 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:27:13,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7111, 2.8096, 2.5850, 4.3122, 2.8967, 4.1008, 1.5382, 3.1056], device='cuda:2'), covar=tensor([0.1423, 0.0761, 0.1175, 0.0190, 0.0171, 0.0404, 0.1697, 0.0684], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0165, 0.0186, 0.0172, 0.0199, 0.0209, 0.0192, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 14:28:08,373 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:28:33,714 INFO [train.py:904] (2/8) Epoch 17, batch 8600, loss[loss=0.178, simple_loss=0.2678, pruned_loss=0.04409, over 17062.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2749, pruned_loss=0.04576, over 3043566.77 frames. ], batch size: 53, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:29:52,153 INFO [optim.py:368] (2/8) 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 14:30:15,670 INFO [train.py:904] (2/8) Epoch 17, batch 8650, loss[loss=0.1703, simple_loss=0.2611, pruned_loss=0.03978, over 12221.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2731, pruned_loss=0.04412, over 3049675.81 frames. ], batch size: 250, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:30:51,143 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8525, 4.2456, 4.2965, 3.2957, 3.7949, 4.3108, 3.9518, 2.7073], device='cuda:2'), covar=tensor([0.0393, 0.0033, 0.0027, 0.0252, 0.0077, 0.0062, 0.0049, 0.0355], device='cuda:2'), in_proj_covar=tensor([0.0130, 0.0074, 0.0074, 0.0128, 0.0088, 0.0099, 0.0086, 0.0121], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:31:43,993 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171092.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:32:02,416 INFO [train.py:904] (2/8) Epoch 17, batch 8700, loss[loss=0.1727, simple_loss=0.2622, pruned_loss=0.0416, over 16938.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2703, pruned_loss=0.04279, over 3067008.21 frames. ], batch size: 116, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:32:05,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7975, 5.0461, 5.2504, 4.9587, 5.0973, 5.6394, 5.0932, 4.7766], device='cuda:2'), covar=tensor([0.0922, 0.1931, 0.1962, 0.2007, 0.2442, 0.0841, 0.1532, 0.2139], device='cuda:2'), in_proj_covar=tensor([0.0371, 0.0539, 0.0591, 0.0451, 0.0606, 0.0629, 0.0471, 0.0606], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:32:23,274 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7782, 5.0946, 4.8539, 4.8386, 4.5995, 4.5468, 4.5320, 5.1485], device='cuda:2'), covar=tensor([0.1097, 0.0816, 0.0957, 0.0756, 0.0749, 0.0974, 0.1158, 0.0830], device='cuda:2'), in_proj_covar=tensor([0.0607, 0.0742, 0.0604, 0.0549, 0.0467, 0.0479, 0.0615, 0.0577], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:32:31,753 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171117.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 14:33:13,274 INFO [optim.py:368] (2/8) 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 14:33:38,783 INFO [train.py:904] (2/8) Epoch 17, batch 8750, loss[loss=0.2036, simple_loss=0.2983, pruned_loss=0.05443, over 15245.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.27, pruned_loss=0.04218, over 3066157.30 frames. ], batch size: 191, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:33:43,330 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:34:13,578 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:35:21,422 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171196.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:35:33,195 INFO [train.py:904] (2/8) Epoch 17, batch 8800, loss[loss=0.164, simple_loss=0.2639, pruned_loss=0.03205, over 16689.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.269, pruned_loss=0.04138, over 3060737.27 frames. ], batch size: 76, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:36:22,009 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:36:54,975 INFO [optim.py:368] (2/8) 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 14:37:18,432 INFO [train.py:904] (2/8) Epoch 17, batch 8850, loss[loss=0.161, simple_loss=0.2523, pruned_loss=0.03487, over 12592.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2713, pruned_loss=0.04099, over 3048115.77 frames. ], batch size: 246, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:37:28,345 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:03,154 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:27,219 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:39:01,947 INFO [train.py:904] (2/8) Epoch 17, batch 8900, loss[loss=0.1838, simple_loss=0.286, pruned_loss=0.04076, over 16196.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.272, pruned_loss=0.04047, over 3071321.46 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:40:38,364 INFO [optim.py:368] (2/8) 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 14:40:50,855 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:41:04,757 INFO [train.py:904] (2/8) Epoch 17, batch 8950, loss[loss=0.1817, simple_loss=0.2654, pruned_loss=0.04894, over 12976.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2719, pruned_loss=0.04094, over 3075288.38 frames. ], batch size: 250, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,003 INFO [train.py:904] (2/8) Epoch 17, batch 9000, loss[loss=0.1543, simple_loss=0.2469, pruned_loss=0.0308, over 16647.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2684, pruned_loss=0.03925, over 3104869.85 frames. ], batch size: 134, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,003 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 14:43:02,943 INFO [train.py:938] (2/8) Epoch 17, validation: loss=0.1478, simple_loss=0.2519, pruned_loss=0.02187, over 944034.00 frames. 2023-04-30 14:43:02,944 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 14:43:11,002 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:43:17,924 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 14:43:40,116 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 14:43:54,015 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 14:44:23,317 INFO [optim.py:368] (2/8) 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 14:44:40,962 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:44:47,395 INFO [train.py:904] (2/8) Epoch 17, batch 9050, loss[loss=0.1786, simple_loss=0.2689, pruned_loss=0.04417, over 12649.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2681, pruned_loss=0.03942, over 3072158.59 frames. ], batch size: 247, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:46:21,434 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 14:46:24,656 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:46:30,493 INFO [train.py:904] (2/8) Epoch 17, batch 9100, loss[loss=0.1666, simple_loss=0.2694, pruned_loss=0.03188, over 16773.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2683, pruned_loss=0.04003, over 3085938.72 frames. ], batch size: 76, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:48:02,850 INFO [optim.py:368] (2/8) 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 14:48:28,379 INFO [train.py:904] (2/8) Epoch 17, batch 9150, loss[loss=0.1544, simple_loss=0.2522, pruned_loss=0.02833, over 16723.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2684, pruned_loss=0.03955, over 3082608.42 frames. ], batch size: 89, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:48:28,877 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:48:44,372 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:49:09,144 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 14:49:36,077 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:50:13,333 INFO [train.py:904] (2/8) Epoch 17, batch 9200, loss[loss=0.1754, simple_loss=0.2693, pruned_loss=0.0407, over 15361.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.264, pruned_loss=0.03865, over 3079088.74 frames. ], batch size: 192, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:51:10,342 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:51:16,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7029, 2.1432, 1.8633, 1.9960, 2.4741, 2.1686, 2.2699, 2.5746], device='cuda:2'), covar=tensor([0.0142, 0.0427, 0.0495, 0.0437, 0.0291, 0.0369, 0.0221, 0.0261], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0220, 0.0213, 0.0213, 0.0219, 0.0218, 0.0216, 0.0209], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:51:27,838 INFO [optim.py:368] (2/8) 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 14:51:50,813 INFO [train.py:904] (2/8) Epoch 17, batch 9250, loss[loss=0.1634, simple_loss=0.2478, pruned_loss=0.03953, over 12515.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2631, pruned_loss=0.03816, over 3068874.06 frames. ], batch size: 247, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:53:01,677 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0779, 3.3781, 3.8100, 2.0350, 3.2609, 2.3985, 3.5993, 3.4661], device='cuda:2'), covar=tensor([0.0269, 0.0874, 0.0415, 0.1957, 0.0674, 0.0921, 0.0646, 0.1029], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0149, 0.0158, 0.0145, 0.0137, 0.0123, 0.0137, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 14:53:04,768 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5075, 4.6424, 4.7583, 4.5706, 4.6610, 5.1724, 4.7126, 4.4186], device='cuda:2'), covar=tensor([0.1136, 0.1861, 0.2111, 0.1907, 0.2366, 0.0911, 0.1370, 0.2090], device='cuda:2'), in_proj_covar=tensor([0.0360, 0.0526, 0.0577, 0.0441, 0.0590, 0.0613, 0.0460, 0.0590], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 14:53:42,073 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:53:42,881 INFO [train.py:904] (2/8) Epoch 17, batch 9300, loss[loss=0.1753, simple_loss=0.2549, pruned_loss=0.0479, over 12499.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.262, pruned_loss=0.03782, over 3075532.74 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:53:56,797 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 14:55:09,819 INFO [optim.py:368] (2/8) 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 14:55:20,709 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171748.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:55:27,770 INFO [train.py:904] (2/8) Epoch 17, batch 9350, loss[loss=0.1659, simple_loss=0.2514, pruned_loss=0.04025, over 12307.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2623, pruned_loss=0.03809, over 3070398.71 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:55:45,852 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5014, 3.5297, 2.0784, 3.9610, 2.6218, 3.8546, 2.1377, 2.8273], device='cuda:2'), covar=tensor([0.0274, 0.0355, 0.1664, 0.0190, 0.0860, 0.0543, 0.1638, 0.0713], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0165, 0.0187, 0.0145, 0.0167, 0.0202, 0.0195, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 14:56:04,938 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9752, 4.2548, 4.0607, 4.0976, 3.7716, 3.8745, 3.8804, 4.2220], device='cuda:2'), covar=tensor([0.1079, 0.1050, 0.0954, 0.0829, 0.0839, 0.1482, 0.0941, 0.1113], device='cuda:2'), in_proj_covar=tensor([0.0603, 0.0739, 0.0595, 0.0545, 0.0463, 0.0473, 0.0611, 0.0572], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:56:57,423 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:57:07,775 INFO [train.py:904] (2/8) Epoch 17, batch 9400, loss[loss=0.1696, simple_loss=0.2761, pruned_loss=0.03161, over 16633.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2624, pruned_loss=0.03794, over 3067551.54 frames. ], batch size: 57, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:29,671 INFO [optim.py:368] (2/8) 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 14:58:48,377 INFO [train.py:904] (2/8) Epoch 17, batch 9450, loss[loss=0.2012, simple_loss=0.2786, pruned_loss=0.06191, over 12107.00 frames. ], tot_loss[loss=0.17, simple_loss=0.264, pruned_loss=0.03795, over 3066824.23 frames. ], batch size: 247, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:48,912 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171852.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:58:52,137 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:59:37,636 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 14:59:48,665 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3550, 3.2421, 3.3209, 3.5256, 3.5371, 3.2964, 3.5285, 3.5827], device='cuda:2'), covar=tensor([0.1548, 0.1175, 0.1451, 0.0780, 0.0821, 0.2668, 0.1061, 0.0980], device='cuda:2'), in_proj_covar=tensor([0.0559, 0.0687, 0.0811, 0.0703, 0.0531, 0.0554, 0.0568, 0.0663], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:59:51,427 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4416, 3.1396, 2.6492, 2.1542, 2.2046, 2.1967, 3.1193, 2.8552], device='cuda:2'), covar=tensor([0.2814, 0.0727, 0.1625, 0.2756, 0.3036, 0.2167, 0.0530, 0.1504], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0252, 0.0289, 0.0291, 0.0274, 0.0236, 0.0274, 0.0309], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 14:59:58,892 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 15:00:26,651 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:00:31,027 INFO [train.py:904] (2/8) Epoch 17, batch 9500, loss[loss=0.1508, simple_loss=0.2356, pruned_loss=0.03302, over 12757.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2633, pruned_loss=0.0379, over 3054207.32 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:00:48,715 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6593, 4.9536, 4.7341, 4.7886, 4.4892, 4.4743, 4.3718, 5.0074], device='cuda:2'), covar=tensor([0.1071, 0.0940, 0.0921, 0.0823, 0.0783, 0.1045, 0.1108, 0.0839], device='cuda:2'), in_proj_covar=tensor([0.0599, 0.0732, 0.0592, 0.0540, 0.0460, 0.0469, 0.0605, 0.0567], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:01:41,237 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:01:51,140 INFO [optim.py:368] (2/8) 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 15:02:14,948 INFO [train.py:904] (2/8) Epoch 17, batch 9550, loss[loss=0.1826, simple_loss=0.2789, pruned_loss=0.04317, over 16816.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2631, pruned_loss=0.03809, over 3060124.48 frames. ], batch size: 116, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:03:46,434 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6223, 3.7137, 2.2724, 4.2648, 2.6941, 4.1515, 2.4019, 3.0186], device='cuda:2'), covar=tensor([0.0269, 0.0353, 0.1588, 0.0212, 0.0905, 0.0417, 0.1496, 0.0696], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0165, 0.0188, 0.0145, 0.0168, 0.0203, 0.0196, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 15:03:49,317 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171998.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:03:57,376 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172001.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:03:58,206 INFO [train.py:904] (2/8) Epoch 17, batch 9600, loss[loss=0.1718, simple_loss=0.2655, pruned_loss=0.03912, over 16566.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2648, pruned_loss=0.03911, over 3076931.70 frames. ], batch size: 68, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:04:36,363 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4760, 2.0308, 1.7274, 1.7016, 2.3322, 1.8856, 1.9468, 2.3543], device='cuda:2'), covar=tensor([0.0158, 0.0391, 0.0496, 0.0489, 0.0253, 0.0379, 0.0197, 0.0257], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0220, 0.0212, 0.0212, 0.0219, 0.0217, 0.0214, 0.0207], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:05:20,542 INFO [optim.py:368] (2/8) 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 15:05:38,178 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 17, batch 9650, loss[loss=0.1544, simple_loss=0.2513, pruned_loss=0.0288, over 16481.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2665, pruned_loss=0.03941, over 3050675.13 frames. ], batch size: 68, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:07:32,480 INFO [train.py:904] (2/8) Epoch 17, batch 9700, loss[loss=0.1751, simple_loss=0.2702, pruned_loss=0.03996, over 12477.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2657, pruned_loss=0.0394, over 3045389.04 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:08:05,467 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172118.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:08:57,398 INFO [optim.py:368] (2/8) 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 15:09:14,191 INFO [train.py:904] (2/8) Epoch 17, batch 9750, loss[loss=0.1689, simple_loss=0.2636, pruned_loss=0.03713, over 17013.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2644, pruned_loss=0.03995, over 3033703.40 frames. ], batch size: 109, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:09:18,291 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:10:09,710 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:10:49,756 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 15:10:54,093 INFO [train.py:904] (2/8) Epoch 17, batch 9800, loss[loss=0.1761, simple_loss=0.2799, pruned_loss=0.0361, over 16579.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2652, pruned_loss=0.03895, over 3051684.26 frames. ], batch size: 68, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:10:54,780 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:11:21,841 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5791, 3.5665, 3.4991, 2.8600, 3.4495, 2.0144, 3.2060, 2.9588], device='cuda:2'), covar=tensor([0.0107, 0.0099, 0.0147, 0.0163, 0.0075, 0.2305, 0.0106, 0.0216], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0134, 0.0177, 0.0160, 0.0154, 0.0192, 0.0166, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:12:17,139 INFO [optim.py:368] (2/8) 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 15:12:39,001 INFO [train.py:904] (2/8) Epoch 17, batch 9850, loss[loss=0.1572, simple_loss=0.2592, pruned_loss=0.02756, over 16113.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2661, pruned_loss=0.03836, over 3070110.10 frames. ], batch size: 165, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:14:05,900 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9098, 3.1144, 2.8025, 5.0894, 3.8803, 4.5309, 1.6535, 3.2566], device='cuda:2'), covar=tensor([0.1325, 0.0675, 0.1133, 0.0128, 0.0157, 0.0283, 0.1566, 0.0672], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0164, 0.0185, 0.0166, 0.0191, 0.0205, 0.0190, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:2') 2023-04-30 15:14:13,936 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:14:32,002 INFO [train.py:904] (2/8) Epoch 17, batch 9900, loss[loss=0.1719, simple_loss=0.2705, pruned_loss=0.03661, over 15157.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2658, pruned_loss=0.03794, over 3066337.32 frames. ], batch size: 190, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:14:35,939 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 15:16:10,689 INFO [optim.py:368] (2/8) 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 15:16:31,646 INFO [train.py:904] (2/8) Epoch 17, batch 9950, loss[loss=0.1783, simple_loss=0.2763, pruned_loss=0.04019, over 15191.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.267, pruned_loss=0.03837, over 3043589.67 frames. ], batch size: 190, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:18:33,055 INFO [train.py:904] (2/8) Epoch 17, batch 10000, loss[loss=0.1702, simple_loss=0.2688, pruned_loss=0.03586, over 16713.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2661, pruned_loss=0.03809, over 3060341.20 frames. ], batch size: 134, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:19:56,011 INFO [optim.py:368] (2/8) 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 15:20:13,519 INFO [train.py:904] (2/8) Epoch 17, batch 10050, loss[loss=0.1766, simple_loss=0.2696, pruned_loss=0.04173, over 16714.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2667, pruned_loss=0.03819, over 3074274.93 frames. ], batch size: 134, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:20:58,137 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:21:47,150 INFO [train.py:904] (2/8) Epoch 17, batch 10100, loss[loss=0.1666, simple_loss=0.2604, pruned_loss=0.03637, over 16200.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2663, pruned_loss=0.03822, over 3060506.49 frames. ], batch size: 165, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:22:57,894 INFO [optim.py:368] (2/8) 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 15:23:33,081 INFO [train.py:904] (2/8) Epoch 18, batch 0, loss[loss=0.1709, simple_loss=0.2525, pruned_loss=0.04467, over 17207.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2525, pruned_loss=0.04467, over 17207.00 frames. ], batch size: 44, lr: 3.82e-03, grad_scale: 8.0 2023-04-30 15:23:33,081 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 15:23:40,336 INFO [train.py:938] (2/8) Epoch 18, validation: loss=0.148, simple_loss=0.2515, pruned_loss=0.02223, over 944034.00 frames. 2023-04-30 15:23:40,337 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 15:24:36,549 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172593.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:44,181 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:50,173 INFO [train.py:904] (2/8) Epoch 18, batch 50, loss[loss=0.1539, simple_loss=0.2415, pruned_loss=0.03321, over 17025.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2699, pruned_loss=0.05059, over 748006.26 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:25:21,086 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:22,856 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0296, 3.9844, 4.3633, 4.3654, 4.4257, 4.1163, 4.1734, 4.1060], device='cuda:2'), covar=tensor([0.0345, 0.0694, 0.0474, 0.0458, 0.0438, 0.0465, 0.0764, 0.0542], device='cuda:2'), in_proj_covar=tensor([0.0360, 0.0392, 0.0385, 0.0365, 0.0428, 0.0403, 0.0490, 0.0321], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 15:25:42,961 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172641.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:48,587 INFO [optim.py:368] (2/8) 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 15:25:56,033 INFO [train.py:904] (2/8) Epoch 18, batch 100, loss[loss=0.2211, simple_loss=0.2947, pruned_loss=0.07374, over 16715.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.268, pruned_loss=0.04837, over 1322320.47 frames. ], batch size: 124, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:26:05,934 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:26:44,018 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:27:02,727 INFO [train.py:904] (2/8) Epoch 18, batch 150, loss[loss=0.2054, simple_loss=0.306, pruned_loss=0.05243, over 16644.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2672, pruned_loss=0.04831, over 1764573.23 frames. ], batch size: 62, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:27:24,378 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0738, 3.9702, 4.5624, 2.1910, 4.7000, 4.7315, 3.4296, 3.6068], device='cuda:2'), covar=tensor([0.0715, 0.0239, 0.0155, 0.1229, 0.0056, 0.0157, 0.0407, 0.0399], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0103, 0.0088, 0.0135, 0.0072, 0.0115, 0.0122, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 15:27:46,305 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:27:47,788 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 15:28:01,451 INFO [optim.py:368] (2/8) 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 15:28:09,983 INFO [train.py:904] (2/8) Epoch 18, batch 200, loss[loss=0.1605, simple_loss=0.2574, pruned_loss=0.03178, over 17113.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2665, pruned_loss=0.04792, over 2111021.73 frames. ], batch size: 47, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:28:20,122 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9740, 4.4737, 3.2196, 2.3393, 2.8450, 2.4299, 4.8011, 3.7251], device='cuda:2'), covar=tensor([0.2662, 0.0590, 0.1699, 0.2906, 0.2808, 0.2173, 0.0343, 0.1361], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0257, 0.0296, 0.0297, 0.0280, 0.0242, 0.0281, 0.0317], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 15:28:29,965 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5552, 4.5120, 4.9587, 4.9581, 5.0092, 4.6473, 4.6199, 4.5069], device='cuda:2'), covar=tensor([0.0459, 0.1134, 0.0613, 0.0635, 0.0602, 0.0703, 0.1328, 0.0707], device='cuda:2'), in_proj_covar=tensor([0.0372, 0.0403, 0.0397, 0.0375, 0.0441, 0.0416, 0.0506, 0.0331], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 15:28:41,356 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:10,028 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:18,640 INFO [train.py:904] (2/8) Epoch 18, batch 250, loss[loss=0.139, simple_loss=0.2223, pruned_loss=0.02781, over 15834.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.264, pruned_loss=0.04724, over 2384336.11 frames. ], batch size: 35, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:29:47,728 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:18,768 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 15:30:28,992 INFO [train.py:904] (2/8) Epoch 18, batch 300, loss[loss=0.177, simple_loss=0.2552, pruned_loss=0.04945, over 16743.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2616, pruned_loss=0.04631, over 2592663.48 frames. ], batch size: 89, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:39,798 INFO [train.py:904] (2/8) Epoch 18, batch 350, loss[loss=0.1884, simple_loss=0.2617, pruned_loss=0.0576, over 16892.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2596, pruned_loss=0.04546, over 2749681.05 frames. ], batch size: 116, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:45,390 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172906.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:31:46,884 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 15:32:34,579 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8514, 4.6114, 4.9213, 5.0835, 5.2642, 4.6177, 5.3102, 5.2700], device='cuda:2'), covar=tensor([0.1824, 0.1370, 0.1749, 0.0750, 0.0598, 0.0975, 0.0536, 0.0604], device='cuda:2'), in_proj_covar=tensor([0.0593, 0.0732, 0.0861, 0.0743, 0.0557, 0.0589, 0.0605, 0.0700], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:32:40,483 INFO [optim.py:368] (2/8) 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 15:32:47,897 INFO [train.py:904] (2/8) Epoch 18, batch 400, loss[loss=0.191, simple_loss=0.2651, pruned_loss=0.05849, over 16433.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2583, pruned_loss=0.04493, over 2884866.96 frames. ], batch size: 75, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:32:51,112 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:30,366 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:59,530 INFO [train.py:904] (2/8) Epoch 18, batch 450, loss[loss=0.1549, simple_loss=0.2382, pruned_loss=0.0358, over 16993.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2562, pruned_loss=0.04392, over 2981002.55 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:34:23,097 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0733, 3.0813, 3.1057, 2.1533, 2.9842, 3.1778, 3.0375, 1.9663], device='cuda:2'), covar=tensor([0.0512, 0.0131, 0.0068, 0.0427, 0.0115, 0.0112, 0.0103, 0.0459], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0079, 0.0078, 0.0134, 0.0092, 0.0102, 0.0090, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 15:34:38,911 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9806, 2.0431, 2.5921, 2.9103, 2.7374, 3.3175, 2.3472, 3.1827], device='cuda:2'), covar=tensor([0.0221, 0.0479, 0.0294, 0.0293, 0.0312, 0.0190, 0.0427, 0.0170], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0172, 0.0182, 0.0141, 0.0187, 0.0135], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:34:47,853 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 15:35:00,082 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 18, batch 500, loss[loss=0.1619, simple_loss=0.2592, pruned_loss=0.03234, over 17118.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2556, pruned_loss=0.0432, over 3051540.50 frames. ], batch size: 47, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:35:59,675 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:36:14,708 INFO [train.py:904] (2/8) Epoch 18, batch 550, loss[loss=0.1404, simple_loss=0.2302, pruned_loss=0.02525, over 17234.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2553, pruned_loss=0.04297, over 3115207.46 frames. ], batch size: 45, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:37:14,572 INFO [optim.py:368] (2/8) 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 15:37:20,884 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173150.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:37:22,686 INFO [train.py:904] (2/8) Epoch 18, batch 600, loss[loss=0.1944, simple_loss=0.2604, pruned_loss=0.06417, over 16713.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2553, pruned_loss=0.04375, over 3163748.16 frames. ], batch size: 124, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:37:51,280 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8412, 2.8852, 2.7059, 4.8870, 3.9292, 4.3691, 1.6453, 3.1922], device='cuda:2'), covar=tensor([0.1413, 0.0778, 0.1194, 0.0260, 0.0260, 0.0431, 0.1649, 0.0814], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0167, 0.0188, 0.0173, 0.0197, 0.0211, 0.0193, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 15:38:30,597 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173201.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:38:31,376 INFO [train.py:904] (2/8) Epoch 18, batch 650, loss[loss=0.176, simple_loss=0.2551, pruned_loss=0.04841, over 15265.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.255, pruned_loss=0.04342, over 3205718.97 frames. ], batch size: 190, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:38:37,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5154, 4.5141, 4.9136, 4.9113, 4.9319, 4.5982, 4.5736, 4.4833], device='cuda:2'), covar=tensor([0.0350, 0.0802, 0.0442, 0.0431, 0.0472, 0.0472, 0.0903, 0.0552], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0422, 0.0413, 0.0389, 0.0456, 0.0435, 0.0528, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 15:38:45,524 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:38:58,866 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9437, 4.7022, 4.9485, 5.1641, 5.4393, 4.6610, 5.3955, 5.3751], device='cuda:2'), covar=tensor([0.1982, 0.1410, 0.1930, 0.0861, 0.0553, 0.1026, 0.0564, 0.0706], device='cuda:2'), in_proj_covar=tensor([0.0603, 0.0741, 0.0877, 0.0752, 0.0562, 0.0598, 0.0613, 0.0713], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:39:32,444 INFO [optim.py:368] (2/8) 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 15:39:40,833 INFO [train.py:904] (2/8) Epoch 18, batch 700, loss[loss=0.1648, simple_loss=0.2485, pruned_loss=0.04053, over 17237.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2544, pruned_loss=0.04269, over 3232746.56 frames. ], batch size: 45, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:39:44,111 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173254.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:22,721 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:46,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0302, 3.0739, 1.8901, 3.1860, 2.4234, 3.2725, 2.0305, 2.5756], device='cuda:2'), covar=tensor([0.0298, 0.0437, 0.1560, 0.0332, 0.0770, 0.0644, 0.1503, 0.0737], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0173, 0.0195, 0.0155, 0.0174, 0.0213, 0.0203, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 15:40:50,345 INFO [train.py:904] (2/8) Epoch 18, batch 750, loss[loss=0.1745, simple_loss=0.2537, pruned_loss=0.04761, over 16276.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2539, pruned_loss=0.04249, over 3252374.38 frames. ], batch size: 165, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:40:51,630 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173302.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:00,074 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:26,654 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:50,586 INFO [optim.py:368] (2/8) 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 15:41:58,280 INFO [train.py:904] (2/8) Epoch 18, batch 800, loss[loss=0.1954, simple_loss=0.2685, pruned_loss=0.06111, over 11866.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2537, pruned_loss=0.04295, over 3259005.29 frames. ], batch size: 246, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:42:22,673 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:42:30,202 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7918, 2.8700, 2.4729, 2.7976, 3.1621, 2.9981, 3.5182, 3.4410], device='cuda:2'), covar=tensor([0.0119, 0.0375, 0.0469, 0.0389, 0.0265, 0.0349, 0.0250, 0.0214], device='cuda:2'), in_proj_covar=tensor([0.0193, 0.0231, 0.0221, 0.0223, 0.0231, 0.0230, 0.0231, 0.0221], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:42:43,047 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 15:42:49,270 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:43:05,168 INFO [train.py:904] (2/8) Epoch 18, batch 850, loss[loss=0.1722, simple_loss=0.2637, pruned_loss=0.04034, over 17150.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2537, pruned_loss=0.04277, over 3278132.22 frames. ], batch size: 48, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:43:55,000 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:44:03,700 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1720, 2.1444, 2.2689, 3.7639, 2.2055, 2.4636, 2.2069, 2.2842], device='cuda:2'), covar=tensor([0.1352, 0.3689, 0.2907, 0.0633, 0.3780, 0.2592, 0.3839, 0.3030], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0426, 0.0355, 0.0322, 0.0427, 0.0490, 0.0396, 0.0498], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:44:07,463 INFO [optim.py:368] (2/8) 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 15:44:15,656 INFO [train.py:904] (2/8) Epoch 18, batch 900, loss[loss=0.1564, simple_loss=0.2369, pruned_loss=0.03799, over 16946.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2528, pruned_loss=0.04188, over 3291056.26 frames. ], batch size: 41, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:44:40,118 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8247, 2.8741, 3.0152, 5.1270, 4.2714, 4.5484, 1.8007, 3.3017], device='cuda:2'), covar=tensor([0.1349, 0.0730, 0.1003, 0.0160, 0.0221, 0.0332, 0.1561, 0.0701], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0167, 0.0189, 0.0175, 0.0198, 0.0213, 0.0194, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 15:45:11,347 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173491.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 15:45:24,069 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:45:25,648 INFO [train.py:904] (2/8) Epoch 18, batch 950, loss[loss=0.1596, simple_loss=0.2504, pruned_loss=0.03444, over 16875.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2531, pruned_loss=0.04216, over 3304065.68 frames. ], batch size: 42, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:45:31,768 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:45:58,307 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5180, 2.5629, 2.2964, 2.3064, 2.9074, 2.6785, 3.1988, 3.1328], device='cuda:2'), covar=tensor([0.0157, 0.0441, 0.0501, 0.0472, 0.0283, 0.0373, 0.0267, 0.0272], device='cuda:2'), in_proj_covar=tensor([0.0196, 0.0234, 0.0224, 0.0225, 0.0234, 0.0232, 0.0233, 0.0224], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:46:26,059 INFO [optim.py:368] (2/8) 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 15:46:31,273 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173549.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:46:35,238 INFO [train.py:904] (2/8) Epoch 18, batch 1000, loss[loss=0.1733, simple_loss=0.2718, pruned_loss=0.03738, over 17239.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2521, pruned_loss=0.04212, over 3309401.63 frames. ], batch size: 52, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:46:35,688 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173552.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:47:44,518 INFO [train.py:904] (2/8) Epoch 18, batch 1050, loss[loss=0.1525, simple_loss=0.2345, pruned_loss=0.03532, over 16844.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2519, pruned_loss=0.04188, over 3312250.00 frames. ], batch size: 39, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:48:02,422 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:48:32,962 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-04-30 15:48:46,276 INFO [optim.py:368] (2/8) 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 15:48:54,961 INFO [train.py:904] (2/8) Epoch 18, batch 1100, loss[loss=0.1577, simple_loss=0.244, pruned_loss=0.03567, over 17230.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2519, pruned_loss=0.04189, over 3306440.02 frames. ], batch size: 43, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:49:12,480 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:49:26,678 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173675.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:49:33,280 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0132, 2.0404, 2.5339, 2.8386, 2.7470, 3.0093, 2.2430, 3.1061], device='cuda:2'), covar=tensor([0.0178, 0.0432, 0.0285, 0.0266, 0.0279, 0.0224, 0.0447, 0.0145], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0187, 0.0172, 0.0176, 0.0184, 0.0144, 0.0190, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:49:34,242 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0215, 5.5647, 5.7278, 5.3859, 5.4696, 6.0766, 5.5973, 5.3254], device='cuda:2'), covar=tensor([0.0934, 0.1891, 0.2064, 0.2200, 0.2706, 0.0971, 0.1416, 0.2315], device='cuda:2'), in_proj_covar=tensor([0.0393, 0.0577, 0.0630, 0.0477, 0.0646, 0.0661, 0.0498, 0.0645], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 15:50:01,540 INFO [train.py:904] (2/8) Epoch 18, batch 1150, loss[loss=0.1466, simple_loss=0.2323, pruned_loss=0.03041, over 16957.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2509, pruned_loss=0.04127, over 3315258.45 frames. ], batch size: 41, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:50:34,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7593, 3.8821, 2.3354, 4.4316, 2.9792, 4.3905, 2.6811, 3.1795], device='cuda:2'), covar=tensor([0.0309, 0.0376, 0.1706, 0.0305, 0.0871, 0.0557, 0.1358, 0.0738], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0174, 0.0196, 0.0157, 0.0175, 0.0215, 0.0204, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 15:51:00,427 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8951, 4.0111, 2.3829, 4.4963, 3.0636, 4.4371, 2.4753, 3.2717], device='cuda:2'), covar=tensor([0.0261, 0.0315, 0.1550, 0.0230, 0.0779, 0.0503, 0.1469, 0.0648], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0174, 0.0196, 0.0157, 0.0175, 0.0215, 0.0204, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 15:51:02,212 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 18, batch 1200, loss[loss=0.1488, simple_loss=0.2298, pruned_loss=0.03394, over 16793.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2496, pruned_loss=0.04031, over 3320047.81 frames. ], batch size: 39, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:52:19,926 INFO [train.py:904] (2/8) Epoch 18, batch 1250, loss[loss=0.161, simple_loss=0.2413, pruned_loss=0.0403, over 12402.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2502, pruned_loss=0.04054, over 3321510.84 frames. ], batch size: 246, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:52:24,765 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:53:21,075 INFO [optim.py:368] (2/8) 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 15:53:22,576 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173847.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:53:28,943 INFO [train.py:904] (2/8) Epoch 18, batch 1300, loss[loss=0.1751, simple_loss=0.2614, pruned_loss=0.04439, over 17094.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2502, pruned_loss=0.04048, over 3319570.73 frames. ], batch size: 53, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:53:32,735 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:54:37,695 INFO [train.py:904] (2/8) Epoch 18, batch 1350, loss[loss=0.1555, simple_loss=0.2409, pruned_loss=0.03505, over 16723.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2507, pruned_loss=0.04067, over 3324535.60 frames. ], batch size: 83, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:54:44,079 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8067, 2.4517, 2.5186, 4.6087, 2.5041, 2.7607, 2.5545, 2.6882], device='cuda:2'), covar=tensor([0.1073, 0.3545, 0.2874, 0.0424, 0.3959, 0.2775, 0.3231, 0.3545], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0429, 0.0358, 0.0325, 0.0428, 0.0493, 0.0399, 0.0501], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:55:14,228 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7715, 4.1719, 4.2023, 2.9353, 3.6087, 4.1960, 3.7536, 2.4751], device='cuda:2'), covar=tensor([0.0467, 0.0067, 0.0047, 0.0363, 0.0111, 0.0094, 0.0093, 0.0417], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0079, 0.0079, 0.0133, 0.0093, 0.0103, 0.0091, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 15:55:37,173 INFO [optim.py:368] (2/8) 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 15:55:44,243 INFO [train.py:904] (2/8) Epoch 18, batch 1400, loss[loss=0.1597, simple_loss=0.2516, pruned_loss=0.03386, over 17098.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2508, pruned_loss=0.04051, over 3326324.91 frames. ], batch size: 47, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:56:04,277 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:56:09,764 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173970.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:56:26,357 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7903, 2.6951, 2.6710, 4.1660, 3.4897, 4.1369, 1.6225, 2.8637], device='cuda:2'), covar=tensor([0.1306, 0.0694, 0.1028, 0.0189, 0.0140, 0.0351, 0.1478, 0.0836], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0167, 0.0187, 0.0175, 0.0197, 0.0212, 0.0192, 0.0188], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 15:56:38,000 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 15:56:56,559 INFO [train.py:904] (2/8) Epoch 18, batch 1450, loss[loss=0.1596, simple_loss=0.2517, pruned_loss=0.03376, over 17222.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2503, pruned_loss=0.04028, over 3319774.92 frames. ], batch size: 46, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:57:11,462 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:14,078 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:16,923 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:25,534 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5940, 4.4532, 4.4879, 4.1941, 4.1866, 4.5258, 4.2978, 4.3141], device='cuda:2'), covar=tensor([0.0568, 0.0811, 0.0294, 0.0283, 0.0824, 0.0450, 0.0480, 0.0612], device='cuda:2'), in_proj_covar=tensor([0.0288, 0.0407, 0.0337, 0.0327, 0.0350, 0.0381, 0.0232, 0.0406], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 15:57:37,558 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 15:57:54,754 INFO [optim.py:368] (2/8) 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 15:58:03,183 INFO [train.py:904] (2/8) Epoch 18, batch 1500, loss[loss=0.1595, simple_loss=0.2606, pruned_loss=0.02926, over 17128.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2511, pruned_loss=0.04068, over 3324173.21 frames. ], batch size: 48, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 15:58:17,543 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 15:58:30,504 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 15:58:36,161 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174076.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:58:38,859 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174078.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:59:10,942 INFO [train.py:904] (2/8) Epoch 18, batch 1550, loss[loss=0.1587, simple_loss=0.2394, pruned_loss=0.03895, over 16658.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2514, pruned_loss=0.04114, over 3313774.59 frames. ], batch size: 37, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:00:00,545 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:00:09,837 INFO [optim.py:368] (2/8) 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 16:00:11,998 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174147.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:00:17,426 INFO [train.py:904] (2/8) Epoch 18, batch 1600, loss[loss=0.167, simple_loss=0.2461, pruned_loss=0.04396, over 16842.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2546, pruned_loss=0.04275, over 3309302.44 frames. ], batch size: 96, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:16,761 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174195.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:01:23,744 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:01:25,621 INFO [train.py:904] (2/8) Epoch 18, batch 1650, loss[loss=0.1538, simple_loss=0.2375, pruned_loss=0.03505, over 16789.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2561, pruned_loss=0.04348, over 3310684.01 frames. ], batch size: 102, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:40,433 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7023, 1.8255, 2.2775, 2.5976, 2.5906, 2.6522, 1.9143, 2.8576], device='cuda:2'), covar=tensor([0.0159, 0.0440, 0.0314, 0.0251, 0.0257, 0.0247, 0.0471, 0.0151], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0190, 0.0176, 0.0180, 0.0188, 0.0147, 0.0193, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:02:23,837 INFO [optim.py:368] (2/8) 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 16:02:32,758 INFO [train.py:904] (2/8) Epoch 18, batch 1700, loss[loss=0.1851, simple_loss=0.261, pruned_loss=0.05458, over 16873.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.258, pruned_loss=0.04409, over 3309476.94 frames. ], batch size: 109, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:02:57,319 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174270.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:03:12,401 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-30 16:03:15,598 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:24,421 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:24,503 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8626, 2.7007, 2.5760, 3.9723, 3.2040, 4.0487, 1.5071, 2.9240], device='cuda:2'), covar=tensor([0.1266, 0.0641, 0.1069, 0.0177, 0.0141, 0.0343, 0.1523, 0.0750], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0168, 0.0189, 0.0177, 0.0199, 0.0214, 0.0193, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 16:03:28,375 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8172, 2.5037, 1.8987, 2.2463, 2.8654, 2.6004, 2.8838, 2.9351], device='cuda:2'), covar=tensor([0.0204, 0.0388, 0.0558, 0.0465, 0.0216, 0.0334, 0.0220, 0.0281], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0233, 0.0223, 0.0225, 0.0233, 0.0232, 0.0235, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:03:40,505 INFO [train.py:904] (2/8) Epoch 18, batch 1750, loss[loss=0.1768, simple_loss=0.2585, pruned_loss=0.04753, over 16426.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2582, pruned_loss=0.04379, over 3311661.00 frames. ], batch size: 146, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:03:56,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0607, 2.1382, 2.6192, 3.1880, 2.8691, 3.5137, 2.5862, 3.4604], device='cuda:2'), covar=tensor([0.0213, 0.0439, 0.0289, 0.0232, 0.0264, 0.0148, 0.0382, 0.0149], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0189, 0.0175, 0.0179, 0.0187, 0.0147, 0.0192, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:04:01,306 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:36,730 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:38,537 INFO [optim.py:368] (2/8) 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 16:04:46,157 INFO [train.py:904] (2/8) Epoch 18, batch 1800, loss[loss=0.1799, simple_loss=0.2763, pruned_loss=0.04173, over 17107.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2589, pruned_loss=0.04307, over 3319864.52 frames. ], batch size: 55, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:46,504 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174352.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:05:00,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8625, 3.9870, 4.2751, 4.2528, 4.2861, 4.0064, 4.0511, 3.9716], device='cuda:2'), covar=tensor([0.0411, 0.0684, 0.0445, 0.0436, 0.0489, 0.0499, 0.0825, 0.0626], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0439, 0.0426, 0.0399, 0.0473, 0.0449, 0.0544, 0.0355], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 16:05:10,951 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174371.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:05:11,506 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 16:05:13,910 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174373.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:33,392 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:51,522 INFO [train.py:904] (2/8) Epoch 18, batch 1850, loss[loss=0.1928, simple_loss=0.2834, pruned_loss=0.05115, over 16504.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2597, pruned_loss=0.04334, over 3316610.59 frames. ], batch size: 68, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:06:18,869 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3333, 5.6865, 5.4458, 5.5233, 5.1581, 5.1162, 5.1498, 5.8146], device='cuda:2'), covar=tensor([0.1189, 0.0894, 0.0958, 0.0791, 0.0851, 0.0697, 0.1141, 0.0823], device='cuda:2'), in_proj_covar=tensor([0.0650, 0.0801, 0.0643, 0.0594, 0.0504, 0.0505, 0.0665, 0.0617], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:06:32,178 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 16:06:37,443 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 16:06:51,028 INFO [optim.py:368] (2/8) 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 16:06:55,574 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174449.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:06:58,376 INFO [train.py:904] (2/8) Epoch 18, batch 1900, loss[loss=0.1443, simple_loss=0.2271, pruned_loss=0.03078, over 16800.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2584, pruned_loss=0.04278, over 3320506.06 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:07:20,675 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 16:07:45,543 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 16:07:57,451 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:08:06,709 INFO [train.py:904] (2/8) Epoch 18, batch 1950, loss[loss=0.1801, simple_loss=0.2765, pruned_loss=0.0418, over 17124.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2582, pruned_loss=0.04257, over 3323175.95 frames. ], batch size: 48, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:08:28,784 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3386, 3.9833, 4.4490, 2.1895, 4.7131, 4.6652, 3.4962, 3.7382], device='cuda:2'), covar=tensor([0.0599, 0.0244, 0.0227, 0.1098, 0.0053, 0.0140, 0.0370, 0.0324], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0107, 0.0094, 0.0140, 0.0076, 0.0122, 0.0126, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 16:08:56,373 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7342, 4.6821, 4.5974, 4.1179, 4.6413, 1.9783, 4.3778, 4.4163], device='cuda:2'), covar=tensor([0.0139, 0.0129, 0.0195, 0.0365, 0.0133, 0.2474, 0.0166, 0.0214], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0145, 0.0191, 0.0173, 0.0166, 0.0202, 0.0181, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:09:05,465 INFO [optim.py:368] (2/8) 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 16:09:14,567 INFO [train.py:904] (2/8) Epoch 18, batch 2000, loss[loss=0.1783, simple_loss=0.2537, pruned_loss=0.05141, over 16809.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2576, pruned_loss=0.04259, over 3334148.50 frames. ], batch size: 83, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:09:16,189 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7233, 3.9036, 2.8692, 2.2111, 2.5264, 2.2427, 3.8689, 3.2732], device='cuda:2'), covar=tensor([0.2635, 0.0595, 0.1858, 0.2816, 0.2731, 0.2138, 0.0536, 0.1439], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0263, 0.0299, 0.0300, 0.0288, 0.0246, 0.0285, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:09:52,952 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0553, 4.8720, 5.1277, 5.3091, 5.5161, 4.8502, 5.4273, 5.4591], device='cuda:2'), covar=tensor([0.1869, 0.1211, 0.1552, 0.0728, 0.0529, 0.0862, 0.0528, 0.0532], device='cuda:2'), in_proj_covar=tensor([0.0624, 0.0773, 0.0907, 0.0784, 0.0581, 0.0621, 0.0635, 0.0740], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:10:23,407 INFO [train.py:904] (2/8) Epoch 18, batch 2050, loss[loss=0.1579, simple_loss=0.2444, pruned_loss=0.03573, over 16834.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2571, pruned_loss=0.04295, over 3339207.82 frames. ], batch size: 42, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:10:35,782 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 16:11:17,516 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:11:26,651 INFO [optim.py:368] (2/8) 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 16:11:28,707 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174647.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:11:33,672 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3696, 5.3142, 5.2580, 4.7898, 4.8976, 5.3163, 5.1599, 4.9426], device='cuda:2'), covar=tensor([0.0583, 0.0407, 0.0237, 0.0310, 0.0967, 0.0408, 0.0278, 0.0657], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0412, 0.0341, 0.0331, 0.0354, 0.0385, 0.0236, 0.0411], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:11:35,057 INFO [train.py:904] (2/8) Epoch 18, batch 2100, loss[loss=0.147, simple_loss=0.234, pruned_loss=0.02996, over 17238.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2579, pruned_loss=0.0437, over 3326709.89 frames. ], batch size: 45, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:12:02,035 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174671.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:12:06,238 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174673.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:12:45,803 INFO [train.py:904] (2/8) Epoch 18, batch 2150, loss[loss=0.1526, simple_loss=0.2491, pruned_loss=0.02801, over 17263.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2592, pruned_loss=0.0446, over 3320164.86 frames. ], batch size: 45, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:12:50,324 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174705.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:12:52,099 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-30 16:13:10,638 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:12,890 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:44,830 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 16:13:45,421 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174744.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:49,508 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 18, batch 2200, loss[loss=0.1792, simple_loss=0.2684, pruned_loss=0.04497, over 16464.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2596, pruned_loss=0.04439, over 3321052.47 frames. ], batch size: 75, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:14:16,317 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:14:56,019 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:15:06,365 INFO [train.py:904] (2/8) Epoch 18, batch 2250, loss[loss=0.2205, simple_loss=0.3006, pruned_loss=0.0702, over 12110.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2607, pruned_loss=0.04512, over 3318693.49 frames. ], batch size: 246, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:15:10,970 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0636, 5.0426, 4.8176, 4.3645, 4.9657, 1.9754, 4.6368, 4.7106], device='cuda:2'), covar=tensor([0.0102, 0.0084, 0.0208, 0.0316, 0.0098, 0.2618, 0.0139, 0.0190], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0146, 0.0192, 0.0173, 0.0167, 0.0202, 0.0181, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:15:25,051 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2540, 5.2344, 5.1455, 4.6868, 4.6937, 5.1890, 5.0740, 4.8034], device='cuda:2'), covar=tensor([0.0606, 0.0526, 0.0276, 0.0306, 0.1075, 0.0426, 0.0351, 0.0699], device='cuda:2'), in_proj_covar=tensor([0.0291, 0.0411, 0.0340, 0.0331, 0.0354, 0.0383, 0.0236, 0.0410], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:16:02,771 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:16:07,686 INFO [optim.py:368] (2/8) 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 16:16:14,117 INFO [train.py:904] (2/8) Epoch 18, batch 2300, loss[loss=0.1814, simple_loss=0.2761, pruned_loss=0.04332, over 17025.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2609, pruned_loss=0.04507, over 3317048.42 frames. ], batch size: 55, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:17:22,134 INFO [train.py:904] (2/8) Epoch 18, batch 2350, loss[loss=0.171, simple_loss=0.2629, pruned_loss=0.03957, over 17125.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2616, pruned_loss=0.04559, over 3316150.24 frames. ], batch size: 47, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:18:14,610 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:18:23,326 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 16:18:25,281 INFO [optim.py:368] (2/8) 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 16:18:25,680 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174947.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:18:32,052 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1173, 5.0441, 4.8019, 4.0658, 4.9392, 1.5149, 4.5932, 4.6376], device='cuda:2'), covar=tensor([0.0112, 0.0098, 0.0254, 0.0512, 0.0128, 0.3503, 0.0192, 0.0285], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0147, 0.0194, 0.0175, 0.0169, 0.0204, 0.0183, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:18:32,764 INFO [train.py:904] (2/8) Epoch 18, batch 2400, loss[loss=0.171, simple_loss=0.2712, pruned_loss=0.03539, over 17255.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2621, pruned_loss=0.0457, over 3319474.97 frames. ], batch size: 52, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:19:20,826 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:33,333 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174995.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:42,204 INFO [train.py:904] (2/8) Epoch 18, batch 2450, loss[loss=0.1769, simple_loss=0.2667, pruned_loss=0.04356, over 16015.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2633, pruned_loss=0.04592, over 3307697.18 frames. ], batch size: 35, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:20:12,745 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0962, 2.0748, 2.6851, 3.1163, 2.8877, 3.5787, 2.4805, 3.5831], device='cuda:2'), covar=tensor([0.0245, 0.0489, 0.0302, 0.0303, 0.0299, 0.0172, 0.0436, 0.0132], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:20:24,398 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1252, 5.0392, 5.0001, 4.5289, 4.6122, 5.0332, 4.9420, 4.6779], device='cuda:2'), covar=tensor([0.0582, 0.0583, 0.0270, 0.0333, 0.1029, 0.0477, 0.0353, 0.0658], device='cuda:2'), in_proj_covar=tensor([0.0291, 0.0413, 0.0341, 0.0333, 0.0355, 0.0385, 0.0236, 0.0411], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:20:40,462 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:20:41,719 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6347, 4.5877, 4.5561, 4.0571, 4.6002, 1.7193, 4.3048, 4.2513], device='cuda:2'), covar=tensor([0.0140, 0.0104, 0.0197, 0.0311, 0.0105, 0.2747, 0.0171, 0.0215], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0147, 0.0193, 0.0175, 0.0168, 0.0203, 0.0183, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:20:43,743 INFO [optim.py:368] (2/8) 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 16:20:44,166 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4498, 5.3514, 5.3206, 4.8181, 4.8930, 5.3311, 5.3011, 4.9805], device='cuda:2'), covar=tensor([0.0557, 0.0598, 0.0286, 0.0327, 0.1078, 0.0496, 0.0272, 0.0684], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0414, 0.0342, 0.0334, 0.0356, 0.0386, 0.0236, 0.0412], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:20:51,501 INFO [train.py:904] (2/8) Epoch 18, batch 2500, loss[loss=0.1426, simple_loss=0.2296, pruned_loss=0.02783, over 16780.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2627, pruned_loss=0.04547, over 3317388.86 frames. ], batch size: 39, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:20:59,431 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 16:21:04,687 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175061.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:21:36,586 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0450, 4.1458, 2.6962, 4.7212, 3.2185, 4.6741, 2.6534, 3.4292], device='cuda:2'), covar=tensor([0.0247, 0.0328, 0.1291, 0.0217, 0.0712, 0.0424, 0.1402, 0.0617], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0178, 0.0198, 0.0162, 0.0177, 0.0222, 0.0206, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 16:21:46,586 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175092.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:22:00,357 INFO [train.py:904] (2/8) Epoch 18, batch 2550, loss[loss=0.1647, simple_loss=0.2483, pruned_loss=0.04057, over 16795.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2633, pruned_loss=0.04577, over 3323932.31 frames. ], batch size: 102, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:22:01,909 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9857, 4.0970, 4.3842, 4.3589, 4.4022, 4.0918, 4.1489, 4.0352], device='cuda:2'), covar=tensor([0.0413, 0.0627, 0.0452, 0.0440, 0.0521, 0.0482, 0.0871, 0.0665], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0433, 0.0422, 0.0395, 0.0469, 0.0444, 0.0539, 0.0350], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 16:22:49,222 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8641, 3.9411, 2.1993, 4.5127, 2.9626, 4.4637, 2.3220, 3.2463], device='cuda:2'), covar=tensor([0.0275, 0.0331, 0.1701, 0.0260, 0.0796, 0.0515, 0.1609, 0.0666], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0161, 0.0176, 0.0221, 0.0205, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 16:22:51,971 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 16:23:09,304 INFO [train.py:904] (2/8) Epoch 18, batch 2600, loss[loss=0.1814, simple_loss=0.2616, pruned_loss=0.05061, over 16865.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2633, pruned_loss=0.0452, over 3318498.70 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:17,657 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:24:19,525 INFO [train.py:904] (2/8) Epoch 18, batch 2650, loss[loss=0.1613, simple_loss=0.2506, pruned_loss=0.03597, over 17222.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2642, pruned_loss=0.04515, over 3309453.53 frames. ], batch size: 45, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:27,563 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:25:22,161 INFO [optim.py:368] (2/8) 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 16:25:27,538 INFO [train.py:904] (2/8) Epoch 18, batch 2700, loss[loss=0.1745, simple_loss=0.2548, pruned_loss=0.04712, over 16208.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2644, pruned_loss=0.04468, over 3313896.62 frames. ], batch size: 164, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:25:42,411 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0659, 4.8227, 4.8942, 5.2650, 5.3945, 4.7743, 5.4699, 5.4608], device='cuda:2'), covar=tensor([0.1925, 0.1325, 0.2237, 0.0977, 0.0785, 0.1079, 0.0759, 0.0783], device='cuda:2'), in_proj_covar=tensor([0.0626, 0.0775, 0.0911, 0.0792, 0.0586, 0.0627, 0.0639, 0.0743], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:25:50,082 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175268.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:26:31,990 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3124, 4.0291, 4.4654, 2.2114, 4.7290, 4.7246, 3.4537, 3.6183], device='cuda:2'), covar=tensor([0.0592, 0.0203, 0.0179, 0.1100, 0.0058, 0.0153, 0.0372, 0.0367], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0138, 0.0077, 0.0122, 0.0125, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 16:26:36,861 INFO [train.py:904] (2/8) Epoch 18, batch 2750, loss[loss=0.1576, simple_loss=0.245, pruned_loss=0.03514, over 16797.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2633, pruned_loss=0.04384, over 3317492.43 frames. ], batch size: 83, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:27:22,308 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3834, 1.6817, 2.0677, 2.2025, 2.4144, 2.4028, 1.7349, 2.5096], device='cuda:2'), covar=tensor([0.0190, 0.0433, 0.0280, 0.0290, 0.0270, 0.0249, 0.0474, 0.0127], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:27:40,025 INFO [optim.py:368] (2/8) 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 16:27:45,214 INFO [train.py:904] (2/8) Epoch 18, batch 2800, loss[loss=0.1772, simple_loss=0.2631, pruned_loss=0.04567, over 16130.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2623, pruned_loss=0.04358, over 3324042.19 frames. ], batch size: 35, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:27:58,706 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:11,758 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:45,677 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:48,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2314, 4.2069, 4.3416, 4.1357, 4.2538, 4.7853, 4.3768, 4.0530], device='cuda:2'), covar=tensor([0.1817, 0.2210, 0.2244, 0.2562, 0.2845, 0.1247, 0.1469, 0.2648], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0585, 0.0643, 0.0488, 0.0658, 0.0669, 0.0506, 0.0654], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:28:56,413 INFO [train.py:904] (2/8) Epoch 18, batch 2850, loss[loss=0.1849, simple_loss=0.2671, pruned_loss=0.05129, over 15548.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2623, pruned_loss=0.04372, over 3318807.56 frames. ], batch size: 190, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:29:05,423 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175409.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:15,102 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8526, 2.7710, 2.4059, 2.5802, 3.0692, 2.7995, 3.5199, 3.3589], device='cuda:2'), covar=tensor([0.0114, 0.0363, 0.0477, 0.0402, 0.0271, 0.0387, 0.0227, 0.0242], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0235, 0.0224, 0.0224, 0.0235, 0.0233, 0.0238, 0.0227], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:29:35,275 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175431.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:46,845 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1586, 2.6208, 2.0858, 2.4042, 2.9847, 2.7336, 3.0998, 3.0984], device='cuda:2'), covar=tensor([0.0211, 0.0371, 0.0533, 0.0477, 0.0236, 0.0353, 0.0241, 0.0282], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0235, 0.0225, 0.0225, 0.0236, 0.0234, 0.0240, 0.0228], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:29:58,776 INFO [optim.py:368] (2/8) 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 16:30:04,917 INFO [train.py:904] (2/8) Epoch 18, batch 2900, loss[loss=0.1644, simple_loss=0.2401, pruned_loss=0.0444, over 16901.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2613, pruned_loss=0.04419, over 3317760.05 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:30:08,939 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175455.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:45,640 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:50,715 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4946, 5.3542, 5.3742, 4.9677, 5.0202, 5.3955, 5.3451, 5.0965], device='cuda:2'), covar=tensor([0.0582, 0.0455, 0.0276, 0.0308, 0.1027, 0.0452, 0.0267, 0.0639], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0420, 0.0347, 0.0339, 0.0361, 0.0392, 0.0240, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:31:04,440 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:31:08,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4344, 2.3337, 2.4511, 4.2124, 2.3488, 2.7088, 2.4009, 2.5030], device='cuda:2'), covar=tensor([0.1255, 0.3695, 0.2709, 0.0567, 0.3779, 0.2519, 0.3624, 0.3072], device='cuda:2'), in_proj_covar=tensor([0.0392, 0.0432, 0.0359, 0.0327, 0.0430, 0.0499, 0.0401, 0.0505], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:31:13,712 INFO [train.py:904] (2/8) Epoch 18, batch 2950, loss[loss=0.1576, simple_loss=0.2459, pruned_loss=0.03465, over 17230.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2605, pruned_loss=0.04398, over 3324403.51 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:32:09,615 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:32:17,355 INFO [optim.py:368] (2/8) 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 16:32:23,314 INFO [train.py:904] (2/8) Epoch 18, batch 3000, loss[loss=0.1643, simple_loss=0.2557, pruned_loss=0.03645, over 16731.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2612, pruned_loss=0.04466, over 3324564.09 frames. ], batch size: 83, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:32:23,315 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 16:32:32,119 INFO [train.py:938] (2/8) Epoch 18, validation: loss=0.1359, simple_loss=0.2415, pruned_loss=0.01516, over 944034.00 frames. 2023-04-30 16:32:32,120 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 16:32:48,740 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175563.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:32:56,438 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 16:33:41,415 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5923, 6.0316, 5.7302, 5.7663, 5.3888, 5.3604, 5.4061, 6.1217], device='cuda:2'), covar=tensor([0.1366, 0.0942, 0.1012, 0.0862, 0.0948, 0.0698, 0.1125, 0.1030], device='cuda:2'), in_proj_covar=tensor([0.0664, 0.0817, 0.0658, 0.0603, 0.0511, 0.0515, 0.0678, 0.0631], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:33:42,215 INFO [train.py:904] (2/8) Epoch 18, batch 3050, loss[loss=0.1524, simple_loss=0.2374, pruned_loss=0.0337, over 17209.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2614, pruned_loss=0.04495, over 3318281.23 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:34:46,185 INFO [optim.py:368] (2/8) 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 16:34:51,893 INFO [train.py:904] (2/8) Epoch 18, batch 3100, loss[loss=0.1454, simple_loss=0.2337, pruned_loss=0.02861, over 16842.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2601, pruned_loss=0.0445, over 3319600.53 frames. ], batch size: 42, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:01,371 INFO [train.py:904] (2/8) Epoch 18, batch 3150, loss[loss=0.159, simple_loss=0.2388, pruned_loss=0.03957, over 16511.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2593, pruned_loss=0.04455, over 3320834.18 frames. ], batch size: 75, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:13,049 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175710.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:36:36,339 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:36:40,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3510, 5.3563, 5.1224, 4.5055, 5.2243, 2.0631, 4.8860, 5.1106], device='cuda:2'), covar=tensor([0.0105, 0.0093, 0.0196, 0.0430, 0.0109, 0.2449, 0.0152, 0.0185], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0147, 0.0195, 0.0176, 0.0169, 0.0203, 0.0185, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:37:05,952 INFO [optim.py:368] (2/8) 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 16:37:09,220 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175750.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:37:11,226 INFO [train.py:904] (2/8) Epoch 18, batch 3200, loss[loss=0.1904, simple_loss=0.272, pruned_loss=0.05444, over 12499.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2586, pruned_loss=0.04414, over 3319060.74 frames. ], batch size: 246, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:37:37,874 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175771.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:11,058 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:20,674 INFO [train.py:904] (2/8) Epoch 18, batch 3250, loss[loss=0.1572, simple_loss=0.2481, pruned_loss=0.03314, over 17240.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2586, pruned_loss=0.04391, over 3321010.60 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:38:45,661 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-30 16:39:09,517 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:17,730 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:23,828 INFO [optim.py:368] (2/8) 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 16:39:29,873 INFO [train.py:904] (2/8) Epoch 18, batch 3300, loss[loss=0.177, simple_loss=0.2655, pruned_loss=0.04419, over 16841.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2595, pruned_loss=0.04421, over 3322389.81 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:39:45,183 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8999, 5.0931, 4.8226, 4.5122, 4.0587, 5.0754, 5.0766, 4.6074], device='cuda:2'), covar=tensor([0.0959, 0.0584, 0.0462, 0.0431, 0.2011, 0.0472, 0.0312, 0.0690], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0420, 0.0347, 0.0339, 0.0363, 0.0393, 0.0239, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:39:45,194 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:40:10,979 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:40:38,820 INFO [train.py:904] (2/8) Epoch 18, batch 3350, loss[loss=0.1617, simple_loss=0.2566, pruned_loss=0.03343, over 17209.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2603, pruned_loss=0.04441, over 3329341.82 frames. ], batch size: 46, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:40:50,864 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:08,449 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-04-30 16:41:34,538 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:38,095 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 16:41:47,202 INFO [train.py:904] (2/8) Epoch 18, batch 3400, loss[loss=0.1728, simple_loss=0.2548, pruned_loss=0.04539, over 16510.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.26, pruned_loss=0.0447, over 3337742.99 frames. ], batch size: 75, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:41:52,779 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 16:41:56,149 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 16:42:08,489 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1725, 5.1310, 5.0275, 4.5437, 4.6749, 5.0916, 5.0793, 4.7261], device='cuda:2'), covar=tensor([0.0644, 0.0515, 0.0329, 0.0369, 0.1081, 0.0453, 0.0338, 0.0852], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0420, 0.0347, 0.0338, 0.0362, 0.0393, 0.0239, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:42:50,702 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 16:43:01,863 INFO [train.py:904] (2/8) Epoch 18, batch 3450, loss[loss=0.1531, simple_loss=0.2356, pruned_loss=0.03528, over 16966.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2586, pruned_loss=0.04383, over 3329994.91 frames. ], batch size: 41, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:43:04,387 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7989, 2.7944, 2.6139, 4.2238, 3.3796, 4.1795, 1.8430, 2.9192], device='cuda:2'), covar=tensor([0.1448, 0.0772, 0.1191, 0.0201, 0.0175, 0.0399, 0.1495, 0.0856], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0169, 0.0188, 0.0181, 0.0203, 0.0215, 0.0193, 0.0188], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 16:43:07,976 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176006.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:43:35,383 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:06,111 INFO [optim.py:368] (2/8) 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 16:44:09,517 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:11,308 INFO [train.py:904] (2/8) Epoch 18, batch 3500, loss[loss=0.1875, simple_loss=0.2836, pruned_loss=0.04565, over 17030.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2577, pruned_loss=0.04339, over 3335487.56 frames. ], batch size: 53, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:44:31,797 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:41,807 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176074.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:14,938 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176098.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:19,352 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2677, 3.2783, 3.4797, 2.4490, 3.1910, 3.5587, 3.2406, 2.0126], device='cuda:2'), covar=tensor([0.0450, 0.0108, 0.0058, 0.0349, 0.0097, 0.0084, 0.0105, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0080, 0.0080, 0.0133, 0.0094, 0.0104, 0.0091, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 16:45:20,520 INFO [train.py:904] (2/8) Epoch 18, batch 3550, loss[loss=0.1719, simple_loss=0.25, pruned_loss=0.04695, over 16505.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2571, pruned_loss=0.04309, over 3327969.65 frames. ], batch size: 146, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:46:10,540 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:46:26,481 INFO [optim.py:368] (2/8) 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 16:46:32,324 INFO [train.py:904] (2/8) Epoch 18, batch 3600, loss[loss=0.164, simple_loss=0.2543, pruned_loss=0.03684, over 17121.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2557, pruned_loss=0.04213, over 3332902.77 frames. ], batch size: 53, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:47:18,726 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:47:43,325 INFO [train.py:904] (2/8) Epoch 18, batch 3650, loss[loss=0.1561, simple_loss=0.2534, pruned_loss=0.0294, over 17124.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2543, pruned_loss=0.04176, over 3340005.07 frames. ], batch size: 47, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:48:35,858 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176237.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:37,607 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 16:48:43,964 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:53,037 INFO [optim.py:368] (2/8) 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 16:48:58,528 INFO [train.py:904] (2/8) Epoch 18, batch 3700, loss[loss=0.1725, simple_loss=0.2436, pruned_loss=0.05076, over 16712.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2531, pruned_loss=0.04328, over 3322591.78 frames. ], batch size: 124, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:49:06,344 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:10,766 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176301.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:11,703 INFO [train.py:904] (2/8) Epoch 18, batch 3750, loss[loss=0.1939, simple_loss=0.2852, pruned_loss=0.05131, over 17031.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2544, pruned_loss=0.04509, over 3307257.42 frames. ], batch size: 41, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:50:13,738 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:35,190 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:51:17,331 INFO [optim.py:368] (2/8) 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 16:51:23,247 INFO [train.py:904] (2/8) Epoch 18, batch 3800, loss[loss=0.184, simple_loss=0.2663, pruned_loss=0.05087, over 16250.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2561, pruned_loss=0.04663, over 3292805.25 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:51:44,098 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:52:35,500 INFO [train.py:904] (2/8) Epoch 18, batch 3850, loss[loss=0.1882, simple_loss=0.2602, pruned_loss=0.05809, over 16867.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2563, pruned_loss=0.04746, over 3290091.68 frames. ], batch size: 116, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:52:53,510 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 16:53:49,449 INFO [train.py:904] (2/8) Epoch 18, batch 3900, loss[loss=0.1824, simple_loss=0.2645, pruned_loss=0.05011, over 16562.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2556, pruned_loss=0.048, over 3282253.46 frames. ], batch size: 62, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:54:11,237 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:00,904 INFO [train.py:904] (2/8) Epoch 18, batch 3950, loss[loss=0.1595, simple_loss=0.2328, pruned_loss=0.04309, over 16827.00 frames. ], tot_loss[loss=0.176, simple_loss=0.255, pruned_loss=0.04849, over 3282663.28 frames. ], batch size: 83, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:55:35,234 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176527.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:50,449 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176537.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:56:05,763 INFO [optim.py:368] (2/8) 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 16:56:12,321 INFO [train.py:904] (2/8) Epoch 18, batch 4000, loss[loss=0.1767, simple_loss=0.2542, pruned_loss=0.0496, over 16712.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2555, pruned_loss=0.04887, over 3279020.87 frames. ], batch size: 83, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:56:56,156 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3527, 3.1646, 3.5280, 1.7049, 3.7514, 3.7444, 2.9685, 2.6872], device='cuda:2'), covar=tensor([0.0848, 0.0304, 0.0202, 0.1423, 0.0083, 0.0150, 0.0412, 0.0507], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0138, 0.0077, 0.0123, 0.0125, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 16:57:00,439 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:18,433 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:22,920 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176601.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:24,387 INFO [train.py:904] (2/8) Epoch 18, batch 4050, loss[loss=0.172, simple_loss=0.2617, pruned_loss=0.04109, over 15427.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2562, pruned_loss=0.04843, over 3281052.71 frames. ], batch size: 190, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:57:32,908 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2595, 2.3092, 2.9338, 3.2440, 3.1271, 3.7552, 2.3585, 3.6889], device='cuda:2'), covar=tensor([0.0157, 0.0400, 0.0232, 0.0237, 0.0226, 0.0096, 0.0413, 0.0079], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0188, 0.0176, 0.0180, 0.0187, 0.0148, 0.0191, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:57:39,760 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:58,553 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:13,004 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2046, 5.5061, 5.2327, 5.2841, 5.0046, 4.9279, 4.8430, 5.6061], device='cuda:2'), covar=tensor([0.1122, 0.0832, 0.0944, 0.0806, 0.0790, 0.0778, 0.1049, 0.0881], device='cuda:2'), in_proj_covar=tensor([0.0650, 0.0803, 0.0651, 0.0599, 0.0503, 0.0512, 0.0668, 0.0622], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:58:30,484 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176649.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:36,778 INFO [train.py:904] (2/8) Epoch 18, batch 4100, loss[loss=0.1961, simple_loss=0.2809, pruned_loss=0.05566, over 16843.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2578, pruned_loss=0.04805, over 3272734.77 frames. ], batch size: 116, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:58:57,248 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0666, 5.0644, 4.8770, 4.2361, 4.9968, 1.9665, 4.7486, 4.6211], device='cuda:2'), covar=tensor([0.0060, 0.0055, 0.0144, 0.0303, 0.0058, 0.2558, 0.0097, 0.0179], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0148, 0.0196, 0.0178, 0.0170, 0.0204, 0.0186, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 16:58:57,272 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:59,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9943, 2.4861, 2.3946, 2.6332, 2.1633, 3.1910, 1.9232, 2.6793], device='cuda:2'), covar=tensor([0.1093, 0.0574, 0.1019, 0.0155, 0.0152, 0.0374, 0.1268, 0.0706], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0182, 0.0205, 0.0215, 0.0195, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 16:59:28,845 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:59:51,168 INFO [train.py:904] (2/8) Epoch 18, batch 4150, loss[loss=0.2034, simple_loss=0.2998, pruned_loss=0.05356, over 16148.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2652, pruned_loss=0.0506, over 3232879.93 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:00:27,857 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:01:00,609 INFO [optim.py:368] (2/8) 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 17:01:06,246 INFO [train.py:904] (2/8) Epoch 18, batch 4200, loss[loss=0.227, simple_loss=0.3204, pruned_loss=0.06681, over 16503.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2717, pruned_loss=0.05179, over 3226978.12 frames. ], batch size: 68, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:01:06,875 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0575, 2.7511, 2.8569, 2.1244, 2.6882, 2.1840, 2.6229, 2.9275], device='cuda:2'), covar=tensor([0.0348, 0.0800, 0.0469, 0.1559, 0.0785, 0.0810, 0.0711, 0.0741], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0160, 0.0164, 0.0150, 0.0142, 0.0126, 0.0142, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:02:21,462 INFO [train.py:904] (2/8) Epoch 18, batch 4250, loss[loss=0.1796, simple_loss=0.269, pruned_loss=0.04512, over 16685.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2756, pruned_loss=0.05238, over 3195806.90 frames. ], batch size: 57, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:02:51,595 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:03:15,969 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3105, 3.3286, 1.8868, 3.7607, 2.5077, 3.7721, 2.2123, 2.6696], device='cuda:2'), covar=tensor([0.0296, 0.0392, 0.1825, 0.0162, 0.0909, 0.0464, 0.1552, 0.0844], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0175, 0.0192, 0.0156, 0.0174, 0.0216, 0.0200, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:03:30,835 INFO [optim.py:368] (2/8) 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 17:03:35,707 INFO [train.py:904] (2/8) Epoch 18, batch 4300, loss[loss=0.1725, simple_loss=0.2699, pruned_loss=0.0376, over 17242.00 frames. ], tot_loss[loss=0.19, simple_loss=0.277, pruned_loss=0.05151, over 3190704.64 frames. ], batch size: 52, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:03:37,831 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:44,707 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176898.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:49,704 INFO [train.py:904] (2/8) Epoch 18, batch 4350, loss[loss=0.2111, simple_loss=0.292, pruned_loss=0.06513, over 16733.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2803, pruned_loss=0.05262, over 3189665.88 frames. ], batch size: 57, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:05:06,451 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176913.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:07,740 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176914.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:05:48,322 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3363, 2.1801, 2.2885, 3.9970, 2.0146, 2.5204, 2.3179, 2.2925], device='cuda:2'), covar=tensor([0.1425, 0.3857, 0.2993, 0.0602, 0.5075, 0.2772, 0.3348, 0.3915], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0435, 0.0358, 0.0326, 0.0429, 0.0502, 0.0403, 0.0507], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:05:55,203 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:57,196 INFO [optim.py:368] (2/8) 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 17:06:03,297 INFO [train.py:904] (2/8) Epoch 18, batch 4400, loss[loss=0.1827, simple_loss=0.2713, pruned_loss=0.04707, over 17105.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2822, pruned_loss=0.0539, over 3192796.85 frames. ], batch size: 47, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:06:06,933 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5304, 2.4797, 2.5115, 4.2247, 3.2536, 3.8845, 1.4353, 2.8784], device='cuda:2'), covar=tensor([0.1408, 0.0884, 0.1206, 0.0138, 0.0267, 0.0397, 0.1703, 0.0833], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0181, 0.0204, 0.0213, 0.0194, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:06:17,358 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:06:45,569 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:07:03,505 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 17:07:16,202 INFO [train.py:904] (2/8) Epoch 18, batch 4450, loss[loss=0.2264, simple_loss=0.3095, pruned_loss=0.07164, over 16643.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2854, pruned_loss=0.05527, over 3198719.32 frames. ], batch size: 57, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:07:38,628 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3284, 4.6528, 4.4464, 4.4473, 4.1378, 4.1084, 4.1641, 4.7017], device='cuda:2'), covar=tensor([0.1198, 0.0849, 0.0905, 0.0765, 0.0805, 0.1553, 0.1129, 0.0873], device='cuda:2'), in_proj_covar=tensor([0.0631, 0.0779, 0.0634, 0.0581, 0.0490, 0.0498, 0.0647, 0.0604], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:07:43,889 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:08:14,872 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8841, 4.8477, 4.6803, 3.0834, 4.0728, 4.5958, 4.0234, 2.8157], device='cuda:2'), covar=tensor([0.0445, 0.0016, 0.0030, 0.0360, 0.0072, 0.0072, 0.0070, 0.0351], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0078, 0.0079, 0.0132, 0.0094, 0.0105, 0.0092, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 17:08:24,607 INFO [optim.py:368] (2/8) 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 17:08:30,150 INFO [train.py:904] (2/8) Epoch 18, batch 4500, loss[loss=0.2007, simple_loss=0.282, pruned_loss=0.05964, over 15414.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2859, pruned_loss=0.05566, over 3199419.37 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:43,612 INFO [train.py:904] (2/8) Epoch 18, batch 4550, loss[loss=0.1722, simple_loss=0.2674, pruned_loss=0.03848, over 16674.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2858, pruned_loss=0.05608, over 3194050.63 frames. ], batch size: 76, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:53,008 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177109.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:09:55,557 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:11,676 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:31,732 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6772, 5.9803, 5.7325, 5.8045, 5.4424, 5.1902, 5.3671, 6.1368], device='cuda:2'), covar=tensor([0.1195, 0.0762, 0.1075, 0.0784, 0.0813, 0.0630, 0.1052, 0.0787], device='cuda:2'), in_proj_covar=tensor([0.0627, 0.0773, 0.0631, 0.0578, 0.0487, 0.0495, 0.0642, 0.0601], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:10:44,763 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-30 17:10:48,535 INFO [optim.py:368] (2/8) 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 17:10:53,841 INFO [train.py:904] (2/8) Epoch 18, batch 4600, loss[loss=0.2012, simple_loss=0.2897, pruned_loss=0.05635, over 15385.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2868, pruned_loss=0.05632, over 3197927.44 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:11:19,982 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:20,126 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:24,023 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:37,383 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7951, 1.3794, 1.7156, 1.6944, 1.7727, 1.9690, 1.6174, 1.7874], device='cuda:2'), covar=tensor([0.0215, 0.0340, 0.0185, 0.0226, 0.0245, 0.0159, 0.0378, 0.0116], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0189, 0.0177, 0.0181, 0.0188, 0.0148, 0.0192, 0.0142], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:12:05,776 INFO [train.py:904] (2/8) Epoch 18, batch 4650, loss[loss=0.18, simple_loss=0.2712, pruned_loss=0.04438, over 16891.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2862, pruned_loss=0.05637, over 3200973.29 frames. ], batch size: 96, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:12:07,405 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5733, 1.7474, 2.2004, 2.4982, 2.5061, 2.7958, 1.7700, 2.7048], device='cuda:2'), covar=tensor([0.0184, 0.0476, 0.0289, 0.0270, 0.0285, 0.0180, 0.0552, 0.0121], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0190, 0.0178, 0.0181, 0.0189, 0.0148, 0.0193, 0.0142], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:12:16,075 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177209.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:13:10,933 INFO [optim.py:368] (2/8) 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 17:13:16,215 INFO [train.py:904] (2/8) Epoch 18, batch 4700, loss[loss=0.2291, simple_loss=0.3047, pruned_loss=0.07674, over 11391.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2834, pruned_loss=0.05484, over 3205457.47 frames. ], batch size: 246, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:13:44,396 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2229, 5.5711, 5.3103, 5.3677, 5.0633, 4.8848, 4.9608, 5.6719], device='cuda:2'), covar=tensor([0.1133, 0.0715, 0.0858, 0.0700, 0.0780, 0.0774, 0.1026, 0.0767], device='cuda:2'), in_proj_covar=tensor([0.0623, 0.0766, 0.0626, 0.0572, 0.0484, 0.0492, 0.0638, 0.0598], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:13:59,445 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:14:28,060 INFO [train.py:904] (2/8) Epoch 18, batch 4750, loss[loss=0.1678, simple_loss=0.2588, pruned_loss=0.03839, over 16695.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2797, pruned_loss=0.05305, over 3202193.42 frames. ], batch size: 134, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:14:56,095 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:15:34,632 INFO [optim.py:368] (2/8) 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 17:15:40,552 INFO [train.py:904] (2/8) Epoch 18, batch 4800, loss[loss=0.2183, simple_loss=0.2981, pruned_loss=0.06919, over 16644.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2759, pruned_loss=0.05078, over 3216154.84 frames. ], batch size: 62, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:16:00,911 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8479, 5.1771, 5.3372, 5.1071, 5.0955, 5.7038, 5.0915, 4.8231], device='cuda:2'), covar=tensor([0.0877, 0.1554, 0.1358, 0.1744, 0.2203, 0.0776, 0.1448, 0.2237], device='cuda:2'), in_proj_covar=tensor([0.0392, 0.0565, 0.0616, 0.0474, 0.0635, 0.0647, 0.0489, 0.0637], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 17:16:01,101 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7589, 1.8665, 2.3860, 2.7946, 2.6825, 3.1012, 2.0693, 3.0661], device='cuda:2'), covar=tensor([0.0193, 0.0449, 0.0296, 0.0264, 0.0262, 0.0167, 0.0458, 0.0131], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0188, 0.0177, 0.0181, 0.0188, 0.0147, 0.0192, 0.0142], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:16:06,854 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177369.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:16:13,107 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3961, 3.3630, 3.4308, 3.5240, 3.5649, 3.3206, 3.5489, 3.6156], device='cuda:2'), covar=tensor([0.1190, 0.0873, 0.0977, 0.0588, 0.0581, 0.2186, 0.0884, 0.0674], device='cuda:2'), in_proj_covar=tensor([0.0600, 0.0741, 0.0870, 0.0757, 0.0558, 0.0598, 0.0610, 0.0706], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:16:57,032 INFO [train.py:904] (2/8) Epoch 18, batch 4850, loss[loss=0.1749, simple_loss=0.2662, pruned_loss=0.04184, over 16902.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2754, pruned_loss=0.04955, over 3207802.35 frames. ], batch size: 109, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:17:40,498 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6493, 1.7859, 1.5892, 1.4849, 1.9460, 1.6409, 1.5569, 1.9208], device='cuda:2'), covar=tensor([0.0170, 0.0303, 0.0420, 0.0387, 0.0234, 0.0287, 0.0202, 0.0224], device='cuda:2'), in_proj_covar=tensor([0.0196, 0.0229, 0.0220, 0.0220, 0.0230, 0.0229, 0.0231, 0.0224], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:18:05,879 INFO [optim.py:368] (2/8) 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 17:18:11,208 INFO [train.py:904] (2/8) Epoch 18, batch 4900, loss[loss=0.1842, simple_loss=0.2762, pruned_loss=0.04612, over 16852.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2752, pruned_loss=0.04877, over 3187278.81 frames. ], batch size: 116, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:18:20,330 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 17:18:30,791 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:18:33,653 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177467.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:19:25,469 INFO [train.py:904] (2/8) Epoch 18, batch 4950, loss[loss=0.1816, simple_loss=0.2836, pruned_loss=0.03982, over 16906.00 frames. ], tot_loss[loss=0.186, simple_loss=0.275, pruned_loss=0.04851, over 3193091.48 frames. ], batch size: 102, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:19:34,712 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177509.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:20:14,904 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1013, 4.9012, 5.1714, 5.3374, 5.5610, 4.9201, 5.5467, 5.5084], device='cuda:2'), covar=tensor([0.1795, 0.1337, 0.1708, 0.0811, 0.0513, 0.0725, 0.0449, 0.0585], device='cuda:2'), in_proj_covar=tensor([0.0600, 0.0743, 0.0872, 0.0761, 0.0558, 0.0598, 0.0611, 0.0706], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:20:30,520 INFO [optim.py:368] (2/8) 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 17:20:35,824 INFO [train.py:904] (2/8) Epoch 18, batch 5000, loss[loss=0.2122, simple_loss=0.3023, pruned_loss=0.06107, over 16297.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.277, pruned_loss=0.04906, over 3184054.02 frames. ], batch size: 165, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:20:37,374 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177553.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:20:41,849 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 18, batch 5050, loss[loss=0.198, simple_loss=0.2888, pruned_loss=0.05356, over 15485.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2775, pruned_loss=0.04874, over 3187777.11 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:21:49,826 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1119, 5.6399, 5.8230, 5.4806, 5.5547, 6.1686, 5.6399, 5.3269], device='cuda:2'), covar=tensor([0.0779, 0.1586, 0.1706, 0.1965, 0.2461, 0.0834, 0.1183, 0.2244], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0562, 0.0615, 0.0474, 0.0633, 0.0645, 0.0485, 0.0635], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 17:22:04,624 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:22:54,211 INFO [optim.py:368] (2/8) 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 17:22:58,430 INFO [train.py:904] (2/8) Epoch 18, batch 5100, loss[loss=0.1714, simple_loss=0.2572, pruned_loss=0.0428, over 16584.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.276, pruned_loss=0.04817, over 3192784.39 frames. ], batch size: 68, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:23:16,210 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177664.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:23:40,936 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4563, 4.5331, 4.3293, 3.9846, 3.9923, 4.4140, 4.2231, 4.1811], device='cuda:2'), covar=tensor([0.0605, 0.0353, 0.0333, 0.0327, 0.1019, 0.0436, 0.0503, 0.0623], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0395, 0.0329, 0.0319, 0.0341, 0.0371, 0.0224, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:24:05,157 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 17:24:10,476 INFO [train.py:904] (2/8) Epoch 18, batch 5150, loss[loss=0.1732, simple_loss=0.2809, pruned_loss=0.03275, over 16818.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2756, pruned_loss=0.04696, over 3203972.06 frames. ], batch size: 102, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:24:38,693 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2308, 4.2864, 4.5939, 4.5617, 4.5438, 4.2613, 4.2554, 4.1475], device='cuda:2'), covar=tensor([0.0299, 0.0488, 0.0325, 0.0346, 0.0456, 0.0389, 0.0884, 0.0493], device='cuda:2'), in_proj_covar=tensor([0.0379, 0.0413, 0.0407, 0.0380, 0.0453, 0.0427, 0.0522, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 17:24:46,063 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:04,660 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7186, 4.7553, 5.1118, 5.0765, 5.0797, 4.7382, 4.6920, 4.5158], device='cuda:2'), covar=tensor([0.0285, 0.0526, 0.0324, 0.0371, 0.0444, 0.0362, 0.0926, 0.0441], device='cuda:2'), in_proj_covar=tensor([0.0380, 0.0413, 0.0408, 0.0381, 0.0453, 0.0427, 0.0522, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 17:25:20,503 INFO [optim.py:368] (2/8) 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 17:25:24,673 INFO [train.py:904] (2/8) Epoch 18, batch 5200, loss[loss=0.1986, simple_loss=0.2766, pruned_loss=0.06032, over 16992.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2742, pruned_loss=0.04642, over 3215642.64 frames. ], batch size: 55, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:25:44,158 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:47,825 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:06,678 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:07,063 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 17:26:21,186 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 17:26:31,594 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:39,158 INFO [train.py:904] (2/8) Epoch 18, batch 5250, loss[loss=0.1732, simple_loss=0.2605, pruned_loss=0.04294, over 16543.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2716, pruned_loss=0.04584, over 3218782.44 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:26:55,360 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:57,997 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6811, 4.7487, 4.5361, 4.2134, 4.1795, 4.6503, 4.5084, 4.3784], device='cuda:2'), covar=tensor([0.0603, 0.0397, 0.0322, 0.0315, 0.1041, 0.0454, 0.0370, 0.0592], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0395, 0.0329, 0.0320, 0.0340, 0.0370, 0.0224, 0.0389], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:26:58,850 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:37,104 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8144, 3.7869, 3.9629, 3.7267, 3.9410, 4.3035, 4.0106, 3.7124], device='cuda:2'), covar=tensor([0.2181, 0.2416, 0.2079, 0.2484, 0.2713, 0.1481, 0.1410, 0.2410], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0561, 0.0614, 0.0473, 0.0632, 0.0643, 0.0485, 0.0633], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 17:27:37,230 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:48,908 INFO [optim.py:368] (2/8) 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 17:27:49,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1802, 3.1879, 2.0007, 3.4621, 2.4467, 3.4662, 2.1073, 2.5851], device='cuda:2'), covar=tensor([0.0294, 0.0375, 0.1613, 0.0148, 0.0842, 0.0500, 0.1559, 0.0751], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0151, 0.0173, 0.0211, 0.0199, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:27:52,847 INFO [train.py:904] (2/8) Epoch 18, batch 5300, loss[loss=0.1757, simple_loss=0.2593, pruned_loss=0.046, over 16478.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2681, pruned_loss=0.04489, over 3214481.96 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:28:02,071 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177858.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:28:47,965 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1646, 5.4840, 5.1627, 5.2597, 4.9835, 4.9096, 4.8548, 5.5649], device='cuda:2'), covar=tensor([0.1149, 0.0801, 0.0984, 0.0714, 0.0772, 0.0744, 0.1058, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0624, 0.0770, 0.0631, 0.0574, 0.0486, 0.0493, 0.0642, 0.0600], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:29:05,781 INFO [train.py:904] (2/8) Epoch 18, batch 5350, loss[loss=0.1758, simple_loss=0.2713, pruned_loss=0.04017, over 16518.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2667, pruned_loss=0.04422, over 3217250.37 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:29:17,510 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:30:15,239 INFO [optim.py:368] (2/8) 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 17:30:19,770 INFO [train.py:904] (2/8) Epoch 18, batch 5400, loss[loss=0.1965, simple_loss=0.2793, pruned_loss=0.05683, over 17071.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2691, pruned_loss=0.045, over 3213218.99 frames. ], batch size: 50, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:30:21,407 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9670, 2.3217, 2.3684, 2.7855, 1.9792, 3.2557, 1.7559, 2.6794], device='cuda:2'), covar=tensor([0.1194, 0.0671, 0.1054, 0.0145, 0.0116, 0.0339, 0.1467, 0.0743], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0180, 0.0204, 0.0214, 0.0196, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:31:39,766 INFO [train.py:904] (2/8) Epoch 18, batch 5450, loss[loss=0.1986, simple_loss=0.2876, pruned_loss=0.0548, over 16804.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2718, pruned_loss=0.0464, over 3206923.74 frames. ], batch size: 83, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:31:52,876 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6793, 2.7287, 2.4630, 3.8319, 2.6121, 3.8872, 1.6119, 2.8405], device='cuda:2'), covar=tensor([0.1418, 0.0725, 0.1208, 0.0224, 0.0210, 0.0418, 0.1673, 0.0798], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0169, 0.0191, 0.0179, 0.0203, 0.0213, 0.0195, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:31:55,174 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-30 17:32:08,122 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 18, batch 5500, loss[loss=0.2652, simple_loss=0.3288, pruned_loss=0.1008, over 12059.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2797, pruned_loss=0.05133, over 3163855.88 frames. ], batch size: 248, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:33:23,722 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 17:34:14,385 INFO [train.py:904] (2/8) Epoch 18, batch 5550, loss[loss=0.3138, simple_loss=0.365, pruned_loss=0.1313, over 11147.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2866, pruned_loss=0.05619, over 3144149.94 frames. ], batch size: 247, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:05,342 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-04-30 17:35:09,026 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:35:31,150 INFO [optim.py:368] (2/8) 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 17:35:34,161 INFO [train.py:904] (2/8) Epoch 18, batch 5600, loss[loss=0.2959, simple_loss=0.3462, pruned_loss=0.1229, over 11136.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2921, pruned_loss=0.06114, over 3098186.15 frames. ], batch size: 248, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:36,602 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:36:18,292 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2051, 3.1996, 3.4935, 1.6738, 3.7032, 3.7022, 2.8730, 2.7170], device='cuda:2'), covar=tensor([0.0844, 0.0264, 0.0191, 0.1210, 0.0073, 0.0179, 0.0425, 0.0454], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0105, 0.0094, 0.0137, 0.0076, 0.0121, 0.0124, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 17:36:58,587 INFO [train.py:904] (2/8) Epoch 18, batch 5650, loss[loss=0.2149, simple_loss=0.2883, pruned_loss=0.0707, over 16627.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.296, pruned_loss=0.06377, over 3108350.95 frames. ], batch size: 62, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:37:09,705 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:38:16,449 INFO [optim.py:368] (2/8) 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 17:38:17,793 INFO [train.py:904] (2/8) Epoch 18, batch 5700, loss[loss=0.2068, simple_loss=0.2965, pruned_loss=0.0585, over 16202.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2981, pruned_loss=0.06585, over 3097497.77 frames. ], batch size: 165, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:38:20,131 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9484, 3.2362, 3.1431, 2.0165, 2.9677, 3.2161, 3.0112, 1.8511], device='cuda:2'), covar=tensor([0.0540, 0.0056, 0.0081, 0.0469, 0.0119, 0.0114, 0.0112, 0.0477], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0130, 0.0093, 0.0103, 0.0090, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 17:38:25,290 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:39:39,091 INFO [train.py:904] (2/8) Epoch 18, batch 5750, loss[loss=0.2216, simple_loss=0.3089, pruned_loss=0.0671, over 16255.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.3008, pruned_loss=0.06721, over 3075121.94 frames. ], batch size: 165, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:40:08,195 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:40:56,719 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 17:40:59,315 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 18, batch 5800, loss[loss=0.2194, simple_loss=0.2925, pruned_loss=0.07313, over 12013.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3013, pruned_loss=0.06734, over 3044902.14 frames. ], batch size: 250, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:41:28,156 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:42:19,744 INFO [train.py:904] (2/8) Epoch 18, batch 5850, loss[loss=0.2084, simple_loss=0.3137, pruned_loss=0.05153, over 16675.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2986, pruned_loss=0.065, over 3043729.56 frames. ], batch size: 89, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:16,099 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:39,873 INFO [optim.py:368] (2/8) 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 17:43:41,465 INFO [train.py:904] (2/8) Epoch 18, batch 5900, loss[loss=0.2241, simple_loss=0.2988, pruned_loss=0.07467, over 15505.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2986, pruned_loss=0.06539, over 3037748.76 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:42,241 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:44,119 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178484.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:51,634 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6656, 4.2439, 4.2691, 2.9350, 3.8632, 4.3195, 3.9033, 2.3363], device='cuda:2'), covar=tensor([0.0480, 0.0044, 0.0043, 0.0360, 0.0078, 0.0096, 0.0071, 0.0434], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0079, 0.0080, 0.0132, 0.0094, 0.0105, 0.0091, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 17:44:58,646 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:59,547 INFO [train.py:904] (2/8) Epoch 18, batch 5950, loss[loss=0.2523, simple_loss=0.321, pruned_loss=0.09182, over 11553.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2993, pruned_loss=0.0638, over 3054076.05 frames. ], batch size: 246, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:45:18,619 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:45:21,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2662, 3.2615, 2.0263, 3.5640, 2.5771, 3.5887, 2.1681, 2.7012], device='cuda:2'), covar=tensor([0.0263, 0.0382, 0.1586, 0.0217, 0.0791, 0.0626, 0.1438, 0.0674], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0171, 0.0190, 0.0151, 0.0173, 0.0211, 0.0198, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:45:29,059 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1897, 5.9109, 6.1137, 5.7492, 5.8417, 6.3960, 5.8070, 5.6430], device='cuda:2'), covar=tensor([0.0904, 0.1631, 0.1890, 0.1889, 0.2318, 0.0790, 0.1498, 0.2293], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0561, 0.0616, 0.0471, 0.0631, 0.0642, 0.0488, 0.0636], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 17:46:17,848 INFO [optim.py:368] (2/8) 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 17:46:19,086 INFO [train.py:904] (2/8) Epoch 18, batch 6000, loss[loss=0.1876, simple_loss=0.2754, pruned_loss=0.0499, over 16746.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2982, pruned_loss=0.06335, over 3058786.66 frames. ], batch size: 83, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:46:19,086 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 17:46:29,944 INFO [train.py:938] (2/8) Epoch 18, validation: loss=0.1531, simple_loss=0.2661, pruned_loss=0.02007, over 944034.00 frames. 2023-04-30 17:46:29,945 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 17:47:39,556 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 17:47:48,044 INFO [train.py:904] (2/8) Epoch 18, batch 6050, loss[loss=0.2089, simple_loss=0.3053, pruned_loss=0.05621, over 16580.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2966, pruned_loss=0.06277, over 3059093.88 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:48:14,287 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178619.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:48:21,605 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7958, 1.8677, 2.3766, 2.6712, 2.6012, 3.0015, 1.9334, 3.0018], device='cuda:2'), covar=tensor([0.0188, 0.0469, 0.0295, 0.0288, 0.0296, 0.0169, 0.0552, 0.0142], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0187, 0.0174, 0.0178, 0.0186, 0.0145, 0.0191, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:49:00,524 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-30 17:49:06,483 INFO [optim.py:368] (2/8) 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 17:49:06,498 INFO [train.py:904] (2/8) Epoch 18, batch 6100, loss[loss=0.1966, simple_loss=0.2936, pruned_loss=0.04975, over 16720.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2958, pruned_loss=0.06129, over 3084816.97 frames. ], batch size: 83, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:49:51,352 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:12,320 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:16,871 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178697.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:50:23,988 INFO [train.py:904] (2/8) Epoch 18, batch 6150, loss[loss=0.2025, simple_loss=0.2844, pruned_loss=0.06029, over 16713.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2939, pruned_loss=0.06101, over 3081291.74 frames. ], batch size: 124, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:39,657 INFO [optim.py:368] (2/8) 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 17:51:39,672 INFO [train.py:904] (2/8) Epoch 18, batch 6200, loss[loss=0.1975, simple_loss=0.2925, pruned_loss=0.05127, over 16775.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2914, pruned_loss=0.05997, over 3093444.06 frames. ], batch size: 83, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:45,190 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178755.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:51:49,740 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178758.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 17:52:09,005 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:52:14,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4645, 2.6202, 2.1744, 2.3479, 3.0489, 2.6404, 3.0976, 3.1992], device='cuda:2'), covar=tensor([0.0112, 0.0370, 0.0484, 0.0400, 0.0238, 0.0331, 0.0246, 0.0213], device='cuda:2'), in_proj_covar=tensor([0.0190, 0.0224, 0.0218, 0.0217, 0.0226, 0.0224, 0.0227, 0.0220], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:52:14,306 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2241, 3.5101, 3.7444, 2.0910, 3.1473, 2.4371, 3.7115, 3.7381], device='cuda:2'), covar=tensor([0.0240, 0.0738, 0.0522, 0.1901, 0.0784, 0.0910, 0.0558, 0.0832], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0161, 0.0166, 0.0150, 0.0143, 0.0128, 0.0143, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 17:52:52,892 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3110, 4.2835, 4.2119, 3.4832, 4.2593, 1.6944, 4.0254, 3.8337], device='cuda:2'), covar=tensor([0.0134, 0.0119, 0.0185, 0.0343, 0.0112, 0.2778, 0.0175, 0.0242], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0142, 0.0187, 0.0172, 0.0162, 0.0198, 0.0177, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 17:52:56,568 INFO [train.py:904] (2/8) Epoch 18, batch 6250, loss[loss=0.2185, simple_loss=0.3094, pruned_loss=0.06382, over 15265.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2912, pruned_loss=0.05961, over 3110057.19 frames. ], batch size: 190, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:53:08,009 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:53:41,400 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:54:15,682 INFO [train.py:904] (2/8) Epoch 18, batch 6300, loss[loss=0.2102, simple_loss=0.2964, pruned_loss=0.06203, over 16265.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2905, pruned_loss=0.05831, over 3136120.20 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:54:17,523 INFO [optim.py:368] (2/8) 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 17:55:22,167 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178894.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:55:34,230 INFO [train.py:904] (2/8) Epoch 18, batch 6350, loss[loss=0.1991, simple_loss=0.2801, pruned_loss=0.05908, over 16624.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2907, pruned_loss=0.05895, over 3134344.34 frames. ], batch size: 57, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:56:52,036 INFO [train.py:904] (2/8) Epoch 18, batch 6400, loss[loss=0.2856, simple_loss=0.3535, pruned_loss=0.1088, over 11219.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2907, pruned_loss=0.06006, over 3115750.16 frames. ], batch size: 247, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:56:53,837 INFO [optim.py:368] (2/8) 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 17:56:57,553 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178955.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:57:27,447 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178975.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:58:07,507 INFO [train.py:904] (2/8) Epoch 18, batch 6450, loss[loss=0.1991, simple_loss=0.2871, pruned_loss=0.05558, over 16430.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2911, pruned_loss=0.06012, over 3107246.53 frames. ], batch size: 146, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:24,077 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179050.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:59:26,157 INFO [train.py:904] (2/8) Epoch 18, batch 6500, loss[loss=0.1865, simple_loss=0.279, pruned_loss=0.04701, over 16794.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2897, pruned_loss=0.05964, over 3121462.25 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:27,317 INFO [optim.py:368] (2/8) 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 17:59:28,315 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179053.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:00:44,329 INFO [train.py:904] (2/8) Epoch 18, batch 6550, loss[loss=0.1978, simple_loss=0.2855, pruned_loss=0.05506, over 16809.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2928, pruned_loss=0.06092, over 3113000.31 frames. ], batch size: 39, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:00:54,328 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179108.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:20,178 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179126.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:58,890 INFO [train.py:904] (2/8) Epoch 18, batch 6600, loss[loss=0.2205, simple_loss=0.3122, pruned_loss=0.06437, over 16612.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2951, pruned_loss=0.06131, over 3108438.05 frames. ], batch size: 57, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:02:00,668 INFO [optim.py:368] (2/8) 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 18:02:02,473 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4567, 2.3874, 2.3989, 4.3318, 2.2896, 2.7594, 2.4405, 2.5700], device='cuda:2'), covar=tensor([0.1161, 0.3371, 0.2580, 0.0414, 0.3704, 0.2299, 0.3145, 0.3022], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0427, 0.0352, 0.0319, 0.0427, 0.0493, 0.0397, 0.0500], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:02:05,099 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179156.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:03:06,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8417, 4.9127, 5.2768, 5.2027, 5.2609, 4.9047, 4.8685, 4.5551], device='cuda:2'), covar=tensor([0.0301, 0.0492, 0.0328, 0.0418, 0.0433, 0.0364, 0.0921, 0.0525], device='cuda:2'), in_proj_covar=tensor([0.0384, 0.0419, 0.0409, 0.0386, 0.0458, 0.0430, 0.0527, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 18:03:17,513 INFO [train.py:904] (2/8) Epoch 18, batch 6650, loss[loss=0.2074, simple_loss=0.2903, pruned_loss=0.06222, over 16892.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2946, pruned_loss=0.06203, over 3102810.64 frames. ], batch size: 116, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:07,405 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4945, 4.5351, 4.8749, 4.8117, 4.8631, 4.5278, 4.5337, 4.3002], device='cuda:2'), covar=tensor([0.0348, 0.0557, 0.0367, 0.0443, 0.0443, 0.0396, 0.1009, 0.0574], device='cuda:2'), in_proj_covar=tensor([0.0387, 0.0422, 0.0411, 0.0389, 0.0461, 0.0433, 0.0531, 0.0348], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 18:04:30,598 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179250.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:04:32,473 INFO [train.py:904] (2/8) Epoch 18, batch 6700, loss[loss=0.2092, simple_loss=0.295, pruned_loss=0.06177, over 16735.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2934, pruned_loss=0.06218, over 3106887.50 frames. ], batch size: 124, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:34,183 INFO [optim.py:368] (2/8) 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 18:05:08,631 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:05:24,220 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8772, 2.7951, 2.6056, 4.7992, 3.5508, 4.1739, 1.6870, 3.0408], device='cuda:2'), covar=tensor([0.1379, 0.0861, 0.1317, 0.0178, 0.0421, 0.0466, 0.1706, 0.0878], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0167, 0.0189, 0.0176, 0.0204, 0.0212, 0.0194, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:05:48,863 INFO [train.py:904] (2/8) Epoch 18, batch 6750, loss[loss=0.2049, simple_loss=0.2905, pruned_loss=0.0597, over 16752.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2929, pruned_loss=0.06247, over 3091302.86 frames. ], batch size: 124, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:06:19,477 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:06:21,447 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3525, 4.1707, 4.3563, 4.5337, 4.6775, 4.2900, 4.6078, 4.6726], device='cuda:2'), covar=tensor([0.1755, 0.1298, 0.1586, 0.0733, 0.0579, 0.1047, 0.0741, 0.0657], device='cuda:2'), in_proj_covar=tensor([0.0600, 0.0735, 0.0866, 0.0751, 0.0562, 0.0593, 0.0609, 0.0699], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:06:25,197 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6129, 2.5696, 1.8790, 2.6693, 2.0935, 2.7398, 2.0668, 2.3382], device='cuda:2'), covar=tensor([0.0370, 0.0370, 0.1317, 0.0237, 0.0659, 0.0498, 0.1392, 0.0674], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0153, 0.0174, 0.0213, 0.0200, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:06:36,837 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1053, 5.0927, 4.8981, 4.1911, 5.0012, 1.8383, 4.7601, 4.6970], device='cuda:2'), covar=tensor([0.0083, 0.0078, 0.0185, 0.0410, 0.0092, 0.2843, 0.0129, 0.0196], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0142, 0.0188, 0.0172, 0.0163, 0.0199, 0.0178, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:07:00,948 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:03,095 INFO [train.py:904] (2/8) Epoch 18, batch 6800, loss[loss=0.201, simple_loss=0.289, pruned_loss=0.05652, over 16921.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2928, pruned_loss=0.06236, over 3097444.46 frames. ], batch size: 96, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:07:04,927 INFO [optim.py:368] (2/8) 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 18:07:05,952 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179353.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:07:18,277 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:15,639 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179398.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:20,457 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179401.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:08:21,233 INFO [train.py:904] (2/8) Epoch 18, batch 6850, loss[loss=0.2004, simple_loss=0.3054, pruned_loss=0.04767, over 16778.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2943, pruned_loss=0.06279, over 3104234.33 frames. ], batch size: 83, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:08:47,266 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 18:08:50,193 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:56,269 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:09:23,524 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179444.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:09:34,752 INFO [train.py:904] (2/8) Epoch 18, batch 6900, loss[loss=0.2025, simple_loss=0.2945, pruned_loss=0.0552, over 16657.00 frames. ], tot_loss[loss=0.209, simple_loss=0.296, pruned_loss=0.06099, over 3137582.11 frames. ], batch size: 134, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:09:38,469 INFO [optim.py:368] (2/8) 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 18:09:47,284 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7143, 4.6951, 5.0526, 5.0316, 5.0289, 4.7236, 4.7024, 4.5126], device='cuda:2'), covar=tensor([0.0330, 0.0654, 0.0474, 0.0448, 0.0444, 0.0444, 0.0904, 0.0530], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0425, 0.0414, 0.0391, 0.0464, 0.0436, 0.0534, 0.0350], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 18:10:10,141 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:10:48,208 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7977, 3.8778, 4.1414, 4.1073, 4.1095, 3.8499, 3.8796, 3.8692], device='cuda:2'), covar=tensor([0.0371, 0.0658, 0.0422, 0.0409, 0.0470, 0.0501, 0.0894, 0.0523], device='cuda:2'), in_proj_covar=tensor([0.0393, 0.0429, 0.0417, 0.0394, 0.0468, 0.0440, 0.0539, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 18:10:53,469 INFO [train.py:904] (2/8) Epoch 18, batch 6950, loss[loss=0.2519, simple_loss=0.3107, pruned_loss=0.09653, over 10986.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2978, pruned_loss=0.06275, over 3120477.62 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:10:58,856 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179505.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:11:34,351 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5624, 2.5977, 2.2403, 3.7984, 2.5741, 3.8357, 1.3853, 2.6316], device='cuda:2'), covar=tensor([0.1532, 0.0839, 0.1431, 0.0209, 0.0272, 0.0494, 0.1882, 0.0961], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0176, 0.0203, 0.0211, 0.0194, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:12:07,079 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179550.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:12:09,864 INFO [train.py:904] (2/8) Epoch 18, batch 7000, loss[loss=0.1998, simple_loss=0.2977, pruned_loss=0.05098, over 16464.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2973, pruned_loss=0.06172, over 3119100.87 frames. ], batch size: 35, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:12:12,176 INFO [optim.py:368] (2/8) 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 18:12:55,826 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 18:13:16,939 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179598.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:13:21,679 INFO [train.py:904] (2/8) Epoch 18, batch 7050, loss[loss=0.2691, simple_loss=0.3261, pruned_loss=0.106, over 11310.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2978, pruned_loss=0.06152, over 3108710.44 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:13:35,915 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5659, 3.4932, 2.7548, 2.2044, 2.4376, 2.2877, 3.7512, 3.2846], device='cuda:2'), covar=tensor([0.2769, 0.0797, 0.1778, 0.2579, 0.2505, 0.2079, 0.0466, 0.1245], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0265, 0.0300, 0.0306, 0.0292, 0.0249, 0.0287, 0.0329], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:14:12,910 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179636.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:14:37,717 INFO [train.py:904] (2/8) Epoch 18, batch 7100, loss[loss=0.1941, simple_loss=0.289, pruned_loss=0.04955, over 16871.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2962, pruned_loss=0.06117, over 3116043.08 frames. ], batch size: 96, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:14:40,300 INFO [optim.py:368] (2/8) 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 18:15:33,609 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0231, 1.9455, 2.5783, 2.8993, 2.8118, 3.3709, 2.1003, 3.3538], device='cuda:2'), covar=tensor([0.0207, 0.0509, 0.0312, 0.0288, 0.0286, 0.0147, 0.0558, 0.0134], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0189, 0.0175, 0.0178, 0.0187, 0.0145, 0.0191, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:15:46,888 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:15:53,678 INFO [train.py:904] (2/8) Epoch 18, batch 7150, loss[loss=0.2559, simple_loss=0.3094, pruned_loss=0.1012, over 11298.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2953, pruned_loss=0.06252, over 3075417.11 frames. ], batch size: 247, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:15:55,541 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-30 18:16:08,990 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4998, 3.4409, 2.6510, 2.1696, 2.2624, 2.2268, 3.5748, 3.1312], device='cuda:2'), covar=tensor([0.2902, 0.0718, 0.1836, 0.2680, 0.2734, 0.2163, 0.0509, 0.1286], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0292, 0.0249, 0.0287, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:16:16,359 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179717.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:16:33,656 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4502, 5.7365, 5.4307, 5.5027, 5.1354, 5.0530, 5.1890, 5.8322], device='cuda:2'), covar=tensor([0.1083, 0.0681, 0.0970, 0.0793, 0.0821, 0.0696, 0.0981, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0764, 0.0624, 0.0572, 0.0478, 0.0491, 0.0636, 0.0593], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:16:36,589 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9543, 4.2002, 4.0197, 4.0436, 3.7354, 3.8305, 3.8984, 4.1912], device='cuda:2'), covar=tensor([0.1038, 0.0810, 0.0888, 0.0792, 0.0731, 0.1435, 0.0844, 0.0964], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0764, 0.0624, 0.0572, 0.0478, 0.0491, 0.0636, 0.0593], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:17:08,205 INFO [train.py:904] (2/8) Epoch 18, batch 7200, loss[loss=0.1793, simple_loss=0.2724, pruned_loss=0.04308, over 16417.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2928, pruned_loss=0.06061, over 3076376.47 frames. ], batch size: 146, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:17:10,635 INFO [optim.py:368] (2/8) 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 18:18:17,283 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:18:20,084 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-04-30 18:18:24,741 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179800.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:18:27,049 INFO [train.py:904] (2/8) Epoch 18, batch 7250, loss[loss=0.1926, simple_loss=0.2784, pruned_loss=0.05336, over 16264.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2909, pruned_loss=0.05955, over 3058146.59 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:19:03,757 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 18:19:41,473 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3907, 2.4887, 2.1131, 2.2326, 2.8582, 2.4606, 2.9999, 3.0412], device='cuda:2'), covar=tensor([0.0089, 0.0367, 0.0471, 0.0439, 0.0230, 0.0361, 0.0222, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0220, 0.0216, 0.0216, 0.0223, 0.0221, 0.0223, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:19:42,150 INFO [train.py:904] (2/8) Epoch 18, batch 7300, loss[loss=0.2059, simple_loss=0.2965, pruned_loss=0.05767, over 16735.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2902, pruned_loss=0.05935, over 3067501.23 frames. ], batch size: 76, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:19:45,258 INFO [optim.py:368] (2/8) 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 18:19:50,429 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:20:20,435 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-30 18:20:58,393 INFO [train.py:904] (2/8) Epoch 18, batch 7350, loss[loss=0.2161, simple_loss=0.2997, pruned_loss=0.06626, over 16889.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2906, pruned_loss=0.05947, over 3079586.64 frames. ], batch size: 109, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:21:04,877 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179906.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:21:09,811 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9191, 2.8073, 2.4677, 2.6279, 3.1682, 2.8068, 3.4948, 3.3922], device='cuda:2'), covar=tensor([0.0083, 0.0386, 0.0456, 0.0415, 0.0241, 0.0352, 0.0201, 0.0225], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0221, 0.0216, 0.0216, 0.0224, 0.0221, 0.0224, 0.0217], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:21:21,457 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1265, 3.5194, 3.2200, 1.6657, 2.8185, 1.9200, 3.3853, 3.6952], device='cuda:2'), covar=tensor([0.0270, 0.0729, 0.0728, 0.2528, 0.1065, 0.1220, 0.0760, 0.0977], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0160, 0.0166, 0.0150, 0.0143, 0.0127, 0.0141, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:21:32,267 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:22:01,644 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6826, 2.4111, 2.2803, 3.3380, 2.2735, 3.6397, 1.4500, 2.6972], device='cuda:2'), covar=tensor([0.1407, 0.0806, 0.1292, 0.0193, 0.0167, 0.0414, 0.1711, 0.0833], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0177, 0.0205, 0.0213, 0.0195, 0.0188], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:22:16,601 INFO [train.py:904] (2/8) Epoch 18, batch 7400, loss[loss=0.2349, simple_loss=0.3089, pruned_loss=0.08047, over 11271.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2915, pruned_loss=0.06034, over 3085645.85 frames. ], batch size: 247, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:22:19,965 INFO [optim.py:368] (2/8) 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 18:22:41,164 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179967.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:23:09,709 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:21,124 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:39,778 INFO [train.py:904] (2/8) Epoch 18, batch 7450, loss[loss=0.2001, simple_loss=0.2908, pruned_loss=0.0547, over 16775.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2924, pruned_loss=0.06118, over 3093162.42 frames. ], batch size: 83, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:24:03,323 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9063, 3.0630, 2.5569, 4.8704, 3.6012, 4.2782, 1.7226, 3.1451], device='cuda:2'), covar=tensor([0.1291, 0.0668, 0.1231, 0.0161, 0.0283, 0.0455, 0.1522, 0.0776], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0169, 0.0190, 0.0178, 0.0205, 0.0213, 0.0196, 0.0188], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:24:05,917 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:19,204 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:21,207 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1685, 1.9899, 2.6528, 3.1111, 2.9645, 3.5597, 2.0651, 3.6057], device='cuda:2'), covar=tensor([0.0179, 0.0497, 0.0301, 0.0241, 0.0242, 0.0149, 0.0525, 0.0115], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0175, 0.0184, 0.0143, 0.0189, 0.0139], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:25:01,488 INFO [train.py:904] (2/8) Epoch 18, batch 7500, loss[loss=0.1776, simple_loss=0.2736, pruned_loss=0.04076, over 16982.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2928, pruned_loss=0.06069, over 3087713.62 frames. ], batch size: 96, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:25:04,519 INFO [optim.py:368] (2/8) 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 18:25:22,815 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:25:26,329 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8326, 3.1953, 3.3980, 1.9130, 2.9600, 2.1238, 3.3438, 3.3963], device='cuda:2'), covar=tensor([0.0258, 0.0742, 0.0526, 0.2086, 0.0801, 0.0987, 0.0638, 0.0965], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0160, 0.0165, 0.0150, 0.0142, 0.0127, 0.0141, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:25:56,739 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180087.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:26:02,467 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2110, 4.4870, 4.7599, 4.6549, 4.6985, 4.4238, 4.1580, 4.2541], device='cuda:2'), covar=tensor([0.0645, 0.0891, 0.0564, 0.0832, 0.0762, 0.0686, 0.1685, 0.0690], device='cuda:2'), in_proj_covar=tensor([0.0383, 0.0418, 0.0408, 0.0386, 0.0459, 0.0430, 0.0525, 0.0346], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 18:26:06,120 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0112, 5.6332, 5.7780, 5.5093, 5.6111, 6.1327, 5.5703, 5.3702], device='cuda:2'), covar=tensor([0.0852, 0.1671, 0.2106, 0.1925, 0.2031, 0.0922, 0.1557, 0.2264], device='cuda:2'), in_proj_covar=tensor([0.0389, 0.0561, 0.0622, 0.0471, 0.0629, 0.0646, 0.0489, 0.0632], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:26:16,419 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7240, 5.0132, 5.1775, 4.9958, 5.0533, 5.5945, 5.0647, 4.8717], device='cuda:2'), covar=tensor([0.1038, 0.1854, 0.2281, 0.1868, 0.2210, 0.0907, 0.1557, 0.2288], device='cuda:2'), in_proj_covar=tensor([0.0389, 0.0560, 0.0621, 0.0470, 0.0628, 0.0645, 0.0488, 0.0631], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:26:16,448 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 18, batch 7550, loss[loss=0.1902, simple_loss=0.2726, pruned_loss=0.05393, over 16659.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2912, pruned_loss=0.06024, over 3102946.02 frames. ], batch size: 134, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:26:51,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8413, 1.3932, 1.7204, 1.6903, 1.8157, 1.9710, 1.6200, 1.7695], device='cuda:2'), covar=tensor([0.0224, 0.0342, 0.0196, 0.0250, 0.0229, 0.0160, 0.0360, 0.0112], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0186, 0.0172, 0.0175, 0.0184, 0.0143, 0.0188, 0.0139], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:27:00,525 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8386, 4.1161, 3.0456, 2.4962, 2.8032, 2.5732, 4.5256, 3.7011], device='cuda:2'), covar=tensor([0.2698, 0.0676, 0.1712, 0.2503, 0.2539, 0.1909, 0.0395, 0.1191], device='cuda:2'), in_proj_covar=tensor([0.0321, 0.0264, 0.0299, 0.0305, 0.0291, 0.0248, 0.0286, 0.0327], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:27:30,332 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180148.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:27:36,545 INFO [train.py:904] (2/8) Epoch 18, batch 7600, loss[loss=0.2173, simple_loss=0.2996, pruned_loss=0.06751, over 16200.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2907, pruned_loss=0.06054, over 3108372.90 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:27:36,917 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180152.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:27:39,406 INFO [optim.py:368] (2/8) 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 18:28:55,649 INFO [train.py:904] (2/8) Epoch 18, batch 7650, loss[loss=0.2041, simple_loss=0.2888, pruned_loss=0.05967, over 16699.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2926, pruned_loss=0.06233, over 3084705.32 frames. ], batch size: 62, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:29:00,764 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:29:46,120 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4350, 4.6912, 4.4717, 4.4801, 4.2273, 4.1810, 4.2132, 4.7267], device='cuda:2'), covar=tensor([0.1022, 0.0809, 0.0931, 0.0826, 0.0725, 0.1465, 0.1058, 0.0864], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0766, 0.0626, 0.0573, 0.0478, 0.0493, 0.0639, 0.0597], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:30:13,287 INFO [train.py:904] (2/8) Epoch 18, batch 7700, loss[loss=0.2226, simple_loss=0.3077, pruned_loss=0.06874, over 16467.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.293, pruned_loss=0.06315, over 3081629.10 frames. ], batch size: 35, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:30:18,205 INFO [optim.py:368] (2/8) 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 18:30:29,172 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180262.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:30:35,081 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:56,987 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:57,339 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 18:30:59,706 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9689, 2.1130, 2.2608, 3.5705, 2.0858, 2.4228, 2.2151, 2.2563], device='cuda:2'), covar=tensor([0.1367, 0.3397, 0.2741, 0.0565, 0.4137, 0.2536, 0.3400, 0.3279], device='cuda:2'), in_proj_covar=tensor([0.0384, 0.0428, 0.0352, 0.0320, 0.0428, 0.0492, 0.0398, 0.0498], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:31:06,800 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7811, 4.6027, 4.8219, 4.9793, 5.1525, 4.6377, 5.1379, 5.1378], device='cuda:2'), covar=tensor([0.1945, 0.1184, 0.1551, 0.0724, 0.0562, 0.0931, 0.0629, 0.0627], device='cuda:2'), in_proj_covar=tensor([0.0597, 0.0731, 0.0862, 0.0747, 0.0558, 0.0590, 0.0612, 0.0701], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:31:06,906 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5342, 2.5701, 2.0520, 2.2963, 2.9816, 2.5820, 3.1487, 3.2090], device='cuda:2'), covar=tensor([0.0098, 0.0403, 0.0543, 0.0451, 0.0236, 0.0390, 0.0199, 0.0231], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0224, 0.0218, 0.0219, 0.0227, 0.0223, 0.0226, 0.0219], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:31:15,912 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:30,817 INFO [train.py:904] (2/8) Epoch 18, batch 7750, loss[loss=0.1914, simple_loss=0.2808, pruned_loss=0.05104, over 17161.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2927, pruned_loss=0.06275, over 3090239.20 frames. ], batch size: 46, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:23,863 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180336.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:29,342 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180340.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:42,450 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 18:32:46,606 INFO [train.py:904] (2/8) Epoch 18, batch 7800, loss[loss=0.2133, simple_loss=0.299, pruned_loss=0.06382, over 16167.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2938, pruned_loss=0.06312, over 3102567.08 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:51,024 INFO [optim.py:368] (2/8) 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 18:33:18,119 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4238, 2.9766, 3.0874, 1.9313, 2.8279, 2.1197, 3.0076, 3.2090], device='cuda:2'), covar=tensor([0.0285, 0.0774, 0.0574, 0.2017, 0.0780, 0.1024, 0.0684, 0.0859], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0159, 0.0164, 0.0150, 0.0142, 0.0127, 0.0141, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:33:33,628 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:33:48,236 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9054, 4.9535, 4.7506, 4.3955, 4.4131, 4.8318, 4.7500, 4.4920], device='cuda:2'), covar=tensor([0.0672, 0.0522, 0.0297, 0.0320, 0.1035, 0.0513, 0.0341, 0.0723], device='cuda:2'), in_proj_covar=tensor([0.0274, 0.0390, 0.0319, 0.0311, 0.0332, 0.0363, 0.0220, 0.0383], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:33:55,593 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180397.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:34:02,022 INFO [train.py:904] (2/8) Epoch 18, batch 7850, loss[loss=0.2205, simple_loss=0.3193, pruned_loss=0.06083, over 16431.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2941, pruned_loss=0.06214, over 3096612.47 frames. ], batch size: 68, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:34:05,172 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 18:34:20,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5094, 5.8233, 5.4970, 5.5914, 5.1885, 5.1801, 5.1818, 5.9122], device='cuda:2'), covar=tensor([0.1133, 0.0719, 0.1045, 0.0849, 0.0847, 0.0674, 0.1118, 0.0849], device='cuda:2'), in_proj_covar=tensor([0.0618, 0.0760, 0.0624, 0.0569, 0.0475, 0.0490, 0.0636, 0.0591], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:34:44,114 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6049, 2.5618, 2.4311, 4.4963, 2.4234, 2.9467, 2.4765, 2.6468], device='cuda:2'), covar=tensor([0.1111, 0.3256, 0.2616, 0.0398, 0.3593, 0.2181, 0.3191, 0.2962], device='cuda:2'), in_proj_covar=tensor([0.0385, 0.0428, 0.0353, 0.0319, 0.0429, 0.0492, 0.0398, 0.0500], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:35:16,217 INFO [train.py:904] (2/8) Epoch 18, batch 7900, loss[loss=0.2114, simple_loss=0.297, pruned_loss=0.06292, over 16893.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2934, pruned_loss=0.06163, over 3104844.34 frames. ], batch size: 116, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:35:17,228 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:35:20,361 INFO [optim.py:368] (2/8) 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 18:36:31,922 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 18, batch 7950, loss[loss=0.248, simple_loss=0.3124, pruned_loss=0.09176, over 11671.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2944, pruned_loss=0.06278, over 3090184.29 frames. ], batch size: 248, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:37:52,817 INFO [train.py:904] (2/8) Epoch 18, batch 8000, loss[loss=0.2231, simple_loss=0.305, pruned_loss=0.07058, over 16780.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.295, pruned_loss=0.06327, over 3099429.63 frames. ], batch size: 124, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:37:57,088 INFO [optim.py:368] (2/8) 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 18:38:07,606 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:08,941 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180562.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:38:21,343 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:30,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8062, 3.8680, 2.4340, 4.4448, 3.0339, 4.3637, 2.5702, 3.0816], device='cuda:2'), covar=tensor([0.0237, 0.0353, 0.1484, 0.0188, 0.0717, 0.0563, 0.1315, 0.0735], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0173, 0.0192, 0.0151, 0.0175, 0.0213, 0.0200, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:38:36,645 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:10,314 INFO [train.py:904] (2/8) Epoch 18, batch 8050, loss[loss=0.1973, simple_loss=0.2932, pruned_loss=0.05073, over 16174.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2943, pruned_loss=0.06255, over 3098063.28 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:39:23,082 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180610.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:39:50,350 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:54,571 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:40:08,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5220, 2.2146, 1.7204, 2.0244, 2.5417, 2.2793, 2.4224, 2.7332], device='cuda:2'), covar=tensor([0.0195, 0.0376, 0.0545, 0.0403, 0.0223, 0.0336, 0.0205, 0.0229], device='cuda:2'), in_proj_covar=tensor([0.0187, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0218], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:40:26,588 INFO [train.py:904] (2/8) Epoch 18, batch 8100, loss[loss=0.1957, simple_loss=0.2857, pruned_loss=0.05291, over 16767.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2931, pruned_loss=0.06152, over 3097576.81 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:40:32,030 INFO [optim.py:368] (2/8) 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 18:40:41,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1511, 5.1705, 4.9834, 4.2825, 5.0133, 1.8198, 4.7444, 4.7288], device='cuda:2'), covar=tensor([0.0074, 0.0068, 0.0181, 0.0380, 0.0084, 0.2620, 0.0120, 0.0194], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0139, 0.0185, 0.0170, 0.0160, 0.0196, 0.0174, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:41:00,873 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9150, 3.9471, 3.1516, 2.4825, 2.7875, 2.6085, 4.4004, 3.5412], device='cuda:2'), covar=tensor([0.2838, 0.0861, 0.1784, 0.2694, 0.2894, 0.1999, 0.0530, 0.1445], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0265, 0.0301, 0.0306, 0.0294, 0.0250, 0.0288, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:41:11,994 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:26,889 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:39,450 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5976, 3.6428, 2.1499, 4.1809, 2.7748, 4.1423, 2.2393, 2.9094], device='cuda:2'), covar=tensor([0.0311, 0.0475, 0.1765, 0.0235, 0.0890, 0.0533, 0.1707, 0.0847], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0175, 0.0193, 0.0152, 0.0176, 0.0215, 0.0202, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 18:41:41,226 INFO [train.py:904] (2/8) Epoch 18, batch 8150, loss[loss=0.2199, simple_loss=0.289, pruned_loss=0.07536, over 11587.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2916, pruned_loss=0.06089, over 3106743.14 frames. ], batch size: 247, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:42:24,162 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180730.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:26,727 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:56,054 INFO [train.py:904] (2/8) Epoch 18, batch 8200, loss[loss=0.2333, simple_loss=0.3053, pruned_loss=0.08068, over 11474.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2893, pruned_loss=0.06049, over 3103571.63 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:43:02,075 INFO [optim.py:368] (2/8) 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 18:44:01,810 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180793.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:44:14,881 INFO [train.py:904] (2/8) Epoch 18, batch 8250, loss[loss=0.2034, simple_loss=0.2947, pruned_loss=0.05606, over 16899.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2875, pruned_loss=0.05818, over 3050679.14 frames. ], batch size: 116, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:44:23,192 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6605, 3.7245, 2.9306, 2.1721, 2.2918, 2.2984, 3.8186, 3.2239], device='cuda:2'), covar=tensor([0.2685, 0.0541, 0.1647, 0.3044, 0.2900, 0.2108, 0.0432, 0.1303], device='cuda:2'), in_proj_covar=tensor([0.0320, 0.0262, 0.0298, 0.0304, 0.0292, 0.0248, 0.0286, 0.0325], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:44:32,922 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 18:44:53,331 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:45:37,335 INFO [train.py:904] (2/8) Epoch 18, batch 8300, loss[loss=0.1831, simple_loss=0.2629, pruned_loss=0.05167, over 12292.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2848, pruned_loss=0.05552, over 3041974.85 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:45:43,842 INFO [optim.py:368] (2/8) 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 18:45:52,358 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180861.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:27,097 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-04-30 18:46:32,620 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:34,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6433, 4.6181, 4.4576, 3.7812, 4.5223, 1.7021, 4.2493, 4.2813], device='cuda:2'), covar=tensor([0.0082, 0.0083, 0.0170, 0.0319, 0.0091, 0.2758, 0.0133, 0.0207], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0138, 0.0183, 0.0168, 0.0159, 0.0195, 0.0173, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 18:46:58,357 INFO [train.py:904] (2/8) Epoch 18, batch 8350, loss[loss=0.198, simple_loss=0.2931, pruned_loss=0.05147, over 16402.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2846, pruned_loss=0.05411, over 3036480.90 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:47:09,895 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:47:10,168 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6452, 3.7034, 2.9082, 2.1222, 2.3654, 2.3818, 3.8900, 3.3328], device='cuda:2'), covar=tensor([0.2654, 0.0584, 0.1661, 0.2889, 0.2686, 0.2096, 0.0431, 0.1184], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0259, 0.0294, 0.0301, 0.0289, 0.0245, 0.0283, 0.0321], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 18:47:29,541 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-30 18:47:35,947 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 18, batch 8400, loss[loss=0.1877, simple_loss=0.2831, pruned_loss=0.04616, over 16659.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2818, pruned_loss=0.05186, over 3026329.05 frames. ], batch size: 134, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:48:22,152 INFO [optim.py:368] (2/8) 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 18:49:16,483 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:49:31,282 INFO [train.py:904] (2/8) Epoch 18, batch 8450, loss[loss=0.1735, simple_loss=0.2654, pruned_loss=0.04082, over 16627.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2804, pruned_loss=0.04992, over 3049218.41 frames. ], batch size: 57, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:50:31,762 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181040.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:50:50,497 INFO [train.py:904] (2/8) Epoch 18, batch 8500, loss[loss=0.1745, simple_loss=0.2662, pruned_loss=0.04138, over 16713.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2763, pruned_loss=0.04733, over 3038280.30 frames. ], batch size: 89, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:50:58,701 INFO [optim.py:368] (2/8) 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 18:51:48,654 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:52:13,305 INFO [train.py:904] (2/8) Epoch 18, batch 8550, loss[loss=0.1884, simple_loss=0.2801, pruned_loss=0.04837, over 15357.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2742, pruned_loss=0.04642, over 3033860.09 frames. ], batch size: 191, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:53:50,706 INFO [train.py:904] (2/8) Epoch 18, batch 8600, loss[loss=0.1581, simple_loss=0.257, pruned_loss=0.02966, over 16704.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2739, pruned_loss=0.04519, over 3025286.52 frames. ], batch size: 89, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:54:01,221 INFO [optim.py:368] (2/8) 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 18:54:49,989 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181181.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:55:29,852 INFO [train.py:904] (2/8) Epoch 18, batch 8650, loss[loss=0.1581, simple_loss=0.2488, pruned_loss=0.03376, over 12146.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2719, pruned_loss=0.04369, over 3026394.87 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:56:25,648 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181226.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:57:15,663 INFO [train.py:904] (2/8) Epoch 18, batch 8700, loss[loss=0.1669, simple_loss=0.266, pruned_loss=0.03389, over 16167.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2699, pruned_loss=0.04252, over 3052061.39 frames. ], batch size: 165, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:57:25,074 INFO [optim.py:368] (2/8) 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 18:57:56,766 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:39,668 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181297.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:48,988 INFO [train.py:904] (2/8) Epoch 18, batch 8750, loss[loss=0.1781, simple_loss=0.2826, pruned_loss=0.03679, over 16476.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.27, pruned_loss=0.04214, over 3052894.04 frames. ], batch size: 68, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:59:12,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6862, 3.9693, 2.9530, 2.2810, 2.4457, 2.4556, 4.2233, 3.3069], device='cuda:2'), covar=tensor([0.2911, 0.0586, 0.1790, 0.2721, 0.2787, 0.2109, 0.0365, 0.1333], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0256, 0.0291, 0.0296, 0.0282, 0.0242, 0.0279, 0.0316], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 19:00:11,051 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9113, 4.6813, 4.9632, 5.0907, 5.2714, 4.6790, 5.2773, 5.2507], device='cuda:2'), covar=tensor([0.1805, 0.1186, 0.1496, 0.0681, 0.0450, 0.0779, 0.0514, 0.0505], device='cuda:2'), in_proj_covar=tensor([0.0577, 0.0712, 0.0833, 0.0729, 0.0542, 0.0575, 0.0590, 0.0683], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:00:41,584 INFO [train.py:904] (2/8) Epoch 18, batch 8800, loss[loss=0.1715, simple_loss=0.2656, pruned_loss=0.03871, over 16948.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2683, pruned_loss=0.04093, over 3070886.74 frames. ], batch size: 109, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:00:51,216 INFO [optim.py:368] (2/8) 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 19:00:54,101 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181358.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:01:02,307 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-30 19:01:11,751 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2023-04-30 19:01:57,973 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:02:27,190 INFO [train.py:904] (2/8) Epoch 18, batch 8850, loss[loss=0.1587, simple_loss=0.2488, pruned_loss=0.03427, over 12505.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2704, pruned_loss=0.04026, over 3062359.76 frames. ], batch size: 250, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:03:20,215 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7342, 5.0589, 5.2003, 5.0031, 5.1118, 5.6147, 5.0598, 4.8075], device='cuda:2'), covar=tensor([0.0955, 0.1663, 0.1682, 0.1838, 0.2033, 0.0729, 0.1495, 0.2069], device='cuda:2'), in_proj_covar=tensor([0.0368, 0.0533, 0.0590, 0.0448, 0.0595, 0.0619, 0.0468, 0.0600], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 19:03:44,882 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:03:52,194 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7165, 3.0650, 3.3921, 2.0144, 2.8581, 2.0471, 3.2468, 3.2667], device='cuda:2'), covar=tensor([0.0294, 0.0877, 0.0531, 0.2056, 0.0833, 0.1091, 0.0707, 0.0928], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0154, 0.0160, 0.0146, 0.0139, 0.0125, 0.0138, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 19:04:15,168 INFO [train.py:904] (2/8) Epoch 18, batch 8900, loss[loss=0.1582, simple_loss=0.2566, pruned_loss=0.02993, over 16877.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2702, pruned_loss=0.03945, over 3057066.96 frames. ], batch size: 96, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:04:23,425 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3821, 4.3748, 4.2175, 3.4413, 4.2620, 1.5589, 3.9686, 4.0127], device='cuda:2'), covar=tensor([0.0110, 0.0091, 0.0220, 0.0362, 0.0118, 0.2827, 0.0180, 0.0273], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0136, 0.0179, 0.0163, 0.0156, 0.0193, 0.0170, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:04:25,752 INFO [optim.py:368] (2/8) 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 19:05:23,168 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 18, batch 8950, loss[loss=0.1619, simple_loss=0.2573, pruned_loss=0.03326, over 16851.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2695, pruned_loss=0.03988, over 3056399.71 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:07:17,052 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:07:19,767 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181530.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:08:08,279 INFO [train.py:904] (2/8) Epoch 18, batch 9000, loss[loss=0.1612, simple_loss=0.2531, pruned_loss=0.03462, over 16746.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2669, pruned_loss=0.03839, over 3078758.61 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:08:08,280 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 19:08:17,828 INFO [train.py:938] (2/8) Epoch 18, validation: loss=0.1475, simple_loss=0.2516, pruned_loss=0.02169, over 944034.00 frames. 2023-04-30 19:08:17,829 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 19:08:27,944 INFO [optim.py:368] (2/8) 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 19:09:39,033 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181591.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:10:01,194 INFO [train.py:904] (2/8) Epoch 18, batch 9050, loss[loss=0.1828, simple_loss=0.2697, pruned_loss=0.04794, over 16761.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2681, pruned_loss=0.03903, over 3094613.36 frames. ], batch size: 83, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:10:43,939 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 19:10:51,746 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 19:11:46,766 INFO [train.py:904] (2/8) Epoch 18, batch 9100, loss[loss=0.1843, simple_loss=0.2726, pruned_loss=0.04801, over 12582.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.268, pruned_loss=0.03958, over 3103933.70 frames. ], batch size: 246, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:50,070 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181653.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:11:55,851 INFO [optim.py:368] (2/8) 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 19:13:32,137 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6290, 3.8067, 2.8564, 2.2701, 2.3569, 2.3656, 4.0432, 3.2967], device='cuda:2'), covar=tensor([0.2835, 0.0617, 0.1793, 0.2842, 0.2882, 0.2142, 0.0412, 0.1334], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0256, 0.0291, 0.0295, 0.0280, 0.0242, 0.0279, 0.0315], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:13:43,588 INFO [train.py:904] (2/8) Epoch 18, batch 9150, loss[loss=0.1429, simple_loss=0.2411, pruned_loss=0.02236, over 16879.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2679, pruned_loss=0.03909, over 3078537.81 frames. ], batch size: 96, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:15:27,777 INFO [train.py:904] (2/8) Epoch 18, batch 9200, loss[loss=0.1546, simple_loss=0.247, pruned_loss=0.03114, over 16834.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2632, pruned_loss=0.03809, over 3096757.39 frames. ], batch size: 83, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:15:36,945 INFO [optim.py:368] (2/8) 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 19:17:05,942 INFO [train.py:904] (2/8) Epoch 18, batch 9250, loss[loss=0.1353, simple_loss=0.2297, pruned_loss=0.02046, over 17016.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2627, pruned_loss=0.03798, over 3074457.60 frames. ], batch size: 50, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:18:33,637 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 19:18:57,424 INFO [train.py:904] (2/8) Epoch 18, batch 9300, loss[loss=0.1741, simple_loss=0.2585, pruned_loss=0.04485, over 16664.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.261, pruned_loss=0.03732, over 3090064.55 frames. ], batch size: 57, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:19:06,077 INFO [optim.py:368] (2/8) 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 19:19:51,054 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 19:20:12,886 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:20:41,424 INFO [train.py:904] (2/8) Epoch 18, batch 9350, loss[loss=0.1653, simple_loss=0.2635, pruned_loss=0.03357, over 16498.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.261, pruned_loss=0.0372, over 3096258.16 frames. ], batch size: 75, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:21:20,073 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2200, 2.0703, 2.0902, 3.7614, 2.0578, 2.4504, 2.2222, 2.2365], device='cuda:2'), covar=tensor([0.1119, 0.3793, 0.3171, 0.0523, 0.4370, 0.2556, 0.3632, 0.3489], device='cuda:2'), in_proj_covar=tensor([0.0378, 0.0421, 0.0350, 0.0312, 0.0423, 0.0481, 0.0390, 0.0489], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:22:15,192 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 19:22:24,027 INFO [train.py:904] (2/8) Epoch 18, batch 9400, loss[loss=0.1657, simple_loss=0.2706, pruned_loss=0.03037, over 16410.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2618, pruned_loss=0.03732, over 3095122.30 frames. ], batch size: 146, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:28,133 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:22:33,105 INFO [optim.py:368] (2/8) 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 19:22:36,233 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3869, 4.6841, 4.5163, 4.4637, 4.2400, 4.1710, 4.2190, 4.7286], device='cuda:2'), covar=tensor([0.1120, 0.0925, 0.0906, 0.0797, 0.0750, 0.1447, 0.1009, 0.0940], device='cuda:2'), in_proj_covar=tensor([0.0612, 0.0745, 0.0613, 0.0557, 0.0469, 0.0483, 0.0623, 0.0579], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:23:16,377 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:23:59,457 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 19:24:03,758 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9764, 3.9982, 4.3234, 4.2788, 4.3031, 4.0731, 4.0496, 4.0496], device='cuda:2'), covar=tensor([0.0375, 0.0728, 0.0478, 0.0667, 0.0533, 0.0587, 0.0912, 0.0465], device='cuda:2'), in_proj_covar=tensor([0.0366, 0.0397, 0.0389, 0.0367, 0.0434, 0.0409, 0.0497, 0.0330], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 19:24:05,165 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 18, batch 9450, loss[loss=0.1697, simple_loss=0.268, pruned_loss=0.03573, over 16390.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2635, pruned_loss=0.03756, over 3089485.16 frames. ], batch size: 146, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:24:09,800 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-04-30 19:24:28,871 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3248, 3.4833, 3.6937, 3.6466, 3.6658, 3.4774, 3.5155, 3.5548], device='cuda:2'), covar=tensor([0.0481, 0.0959, 0.0576, 0.0891, 0.0699, 0.0908, 0.0976, 0.0499], device='cuda:2'), in_proj_covar=tensor([0.0365, 0.0396, 0.0388, 0.0367, 0.0433, 0.0408, 0.0496, 0.0330], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 19:25:19,147 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182039.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:25:41,959 INFO [train.py:904] (2/8) Epoch 18, batch 9500, loss[loss=0.1564, simple_loss=0.2421, pruned_loss=0.03531, over 12790.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2627, pruned_loss=0.03736, over 3084762.45 frames. ], batch size: 250, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:55,086 INFO [optim.py:368] (2/8) 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 19:26:15,900 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-04-30 19:27:27,073 INFO [train.py:904] (2/8) Epoch 18, batch 9550, loss[loss=0.1538, simple_loss=0.2433, pruned_loss=0.03214, over 16334.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2627, pruned_loss=0.03751, over 3099942.99 frames. ], batch size: 35, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:28:08,138 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-30 19:29:06,816 INFO [train.py:904] (2/8) Epoch 18, batch 9600, loss[loss=0.1948, simple_loss=0.2898, pruned_loss=0.04993, over 16949.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2643, pruned_loss=0.03825, over 3095291.49 frames. ], batch size: 109, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:29:15,380 INFO [optim.py:368] (2/8) 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 19:30:14,228 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182186.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:30:22,143 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 19:30:52,030 INFO [train.py:904] (2/8) Epoch 18, batch 9650, loss[loss=0.1821, simple_loss=0.2838, pruned_loss=0.04024, over 16702.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2663, pruned_loss=0.03851, over 3104257.34 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:00,666 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182234.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:32:02,231 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0680, 4.0695, 4.4236, 4.3907, 4.4129, 4.1853, 4.1733, 4.1690], device='cuda:2'), covar=tensor([0.0314, 0.0675, 0.0426, 0.0428, 0.0455, 0.0417, 0.0780, 0.0401], device='cuda:2'), in_proj_covar=tensor([0.0363, 0.0395, 0.0385, 0.0367, 0.0432, 0.0407, 0.0494, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 19:32:37,722 INFO [train.py:904] (2/8) Epoch 18, batch 9700, loss[loss=0.165, simple_loss=0.2653, pruned_loss=0.03231, over 16861.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2651, pruned_loss=0.03834, over 3104026.60 frames. ], batch size: 102, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:45,794 INFO [optim.py:368] (2/8) 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 19:33:37,395 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:34:12,270 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 19:34:17,889 INFO [train.py:904] (2/8) Epoch 18, batch 9750, loss[loss=0.1706, simple_loss=0.2655, pruned_loss=0.03781, over 15365.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2632, pruned_loss=0.03827, over 3080263.14 frames. ], batch size: 191, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:35:24,353 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:39,000 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:56,047 INFO [train.py:904] (2/8) Epoch 18, batch 9800, loss[loss=0.1769, simple_loss=0.2785, pruned_loss=0.03767, over 16685.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2637, pruned_loss=0.03743, over 3097850.76 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:36:05,429 INFO [optim.py:368] (2/8) 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 19:36:47,338 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6074, 3.0276, 3.1231, 1.9480, 2.7757, 2.1527, 3.1244, 3.1281], device='cuda:2'), covar=tensor([0.0337, 0.0801, 0.0645, 0.2126, 0.0898, 0.1025, 0.0811, 0.1086], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0147, 0.0138, 0.0124, 0.0137, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 19:37:39,079 INFO [train.py:904] (2/8) Epoch 18, batch 9850, loss[loss=0.1755, simple_loss=0.2675, pruned_loss=0.0417, over 16757.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2646, pruned_loss=0.03702, over 3100689.14 frames. ], batch size: 124, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:29,928 INFO [train.py:904] (2/8) Epoch 18, batch 9900, loss[loss=0.1848, simple_loss=0.2803, pruned_loss=0.04468, over 15395.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2642, pruned_loss=0.03686, over 3079993.93 frames. ], batch size: 191, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:40,680 INFO [optim.py:368] (2/8) 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 19:41:00,287 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9880, 2.5105, 2.6094, 1.8545, 2.7380, 2.8344, 2.4605, 2.4038], device='cuda:2'), covar=tensor([0.0674, 0.0258, 0.0223, 0.1105, 0.0112, 0.0224, 0.0422, 0.0427], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0101, 0.0088, 0.0134, 0.0072, 0.0113, 0.0119, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 19:41:14,838 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0118, 1.8099, 1.6044, 1.4573, 1.9670, 1.6369, 1.5764, 1.8990], device='cuda:2'), covar=tensor([0.0170, 0.0314, 0.0429, 0.0394, 0.0240, 0.0293, 0.0159, 0.0204], device='cuda:2'), in_proj_covar=tensor([0.0183, 0.0223, 0.0217, 0.0217, 0.0224, 0.0222, 0.0220, 0.0214], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:41:27,898 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4637, 3.3931, 3.5233, 3.6006, 3.6070, 3.3024, 3.5967, 3.6559], device='cuda:2'), covar=tensor([0.1215, 0.0912, 0.0979, 0.0596, 0.0615, 0.2240, 0.0821, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0570, 0.0695, 0.0814, 0.0718, 0.0536, 0.0561, 0.0584, 0.0672], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:41:28,673 INFO [train.py:904] (2/8) Epoch 18, batch 9950, loss[loss=0.1798, simple_loss=0.265, pruned_loss=0.04728, over 12555.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2662, pruned_loss=0.03739, over 3057810.53 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:41:58,904 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:42:01,867 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1864, 2.0688, 2.5365, 3.1470, 2.8191, 3.5087, 2.2487, 3.4701], device='cuda:2'), covar=tensor([0.0168, 0.0427, 0.0353, 0.0211, 0.0277, 0.0133, 0.0440, 0.0146], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0180, 0.0168, 0.0169, 0.0180, 0.0138, 0.0184, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:42:48,020 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6684, 2.7053, 1.9273, 2.8089, 2.1640, 2.8281, 2.0622, 2.3496], device='cuda:2'), covar=tensor([0.0243, 0.0349, 0.1161, 0.0259, 0.0639, 0.0511, 0.1208, 0.0553], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0165, 0.0184, 0.0142, 0.0168, 0.0198, 0.0192, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-04-30 19:43:20,925 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 19:43:31,283 INFO [train.py:904] (2/8) Epoch 18, batch 10000, loss[loss=0.155, simple_loss=0.2485, pruned_loss=0.03076, over 17211.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2648, pruned_loss=0.03705, over 3056341.78 frames. ], batch size: 45, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:43:42,214 INFO [optim.py:368] (2/8) 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 19:44:16,783 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:45:02,377 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-04-30 19:45:14,019 INFO [train.py:904] (2/8) Epoch 18, batch 10050, loss[loss=0.1888, simple_loss=0.2871, pruned_loss=0.04529, over 16224.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2644, pruned_loss=0.03681, over 3053148.80 frames. ], batch size: 165, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:15,019 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:20,809 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182637.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:24,037 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 19:46:46,717 INFO [train.py:904] (2/8) Epoch 18, batch 10100, loss[loss=0.1494, simple_loss=0.2412, pruned_loss=0.02877, over 15301.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.265, pruned_loss=0.03724, over 3064202.52 frames. ], batch size: 190, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:55,747 INFO [optim.py:368] (2/8) 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 19:47:44,812 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:48:32,143 INFO [train.py:904] (2/8) Epoch 19, batch 0, loss[loss=0.1855, simple_loss=0.2761, pruned_loss=0.04746, over 17110.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2761, pruned_loss=0.04746, over 17110.00 frames. ], batch size: 48, lr: 3.61e-03, grad_scale: 8.0 2023-04-30 19:48:32,143 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 19:48:39,768 INFO [train.py:938] (2/8) Epoch 19, validation: loss=0.1468, simple_loss=0.2504, pruned_loss=0.0216, over 944034.00 frames. 2023-04-30 19:48:39,769 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 19:49:50,012 INFO [train.py:904] (2/8) Epoch 19, batch 50, loss[loss=0.2139, simple_loss=0.2844, pruned_loss=0.07174, over 16509.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2751, pruned_loss=0.05401, over 732708.91 frames. ], batch size: 146, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:49:59,029 INFO [optim.py:368] (2/8) 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 19:50:13,699 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2645, 3.3051, 3.3667, 2.1645, 2.9230, 2.3385, 3.7388, 3.6364], device='cuda:2'), covar=tensor([0.0214, 0.0853, 0.0583, 0.1901, 0.0857, 0.0999, 0.0504, 0.0874], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0147, 0.0138, 0.0124, 0.0137, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 19:50:25,155 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 19:50:34,123 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0708, 5.0744, 4.8325, 4.4410, 4.8947, 1.8733, 4.6870, 4.8106], device='cuda:2'), covar=tensor([0.0095, 0.0084, 0.0228, 0.0331, 0.0104, 0.2761, 0.0153, 0.0204], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0137, 0.0179, 0.0161, 0.0157, 0.0194, 0.0170, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:50:55,409 INFO [train.py:904] (2/8) Epoch 19, batch 100, loss[loss=0.2167, simple_loss=0.2885, pruned_loss=0.07241, over 15514.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2695, pruned_loss=0.04937, over 1309979.65 frames. ], batch size: 190, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:51:05,087 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6439, 4.7325, 4.8780, 4.7361, 4.7332, 5.3681, 4.8730, 4.5429], device='cuda:2'), covar=tensor([0.1425, 0.2274, 0.2296, 0.2467, 0.3082, 0.1216, 0.1793, 0.2809], device='cuda:2'), in_proj_covar=tensor([0.0376, 0.0545, 0.0607, 0.0458, 0.0611, 0.0636, 0.0478, 0.0614], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 19:52:03,140 INFO [train.py:904] (2/8) Epoch 19, batch 150, loss[loss=0.1509, simple_loss=0.2455, pruned_loss=0.02816, over 17196.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.27, pruned_loss=0.05015, over 1735263.50 frames. ], batch size: 46, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:09,651 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182856.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:15,180 INFO [optim.py:368] (2/8) 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 19:52:18,024 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:28,159 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:56,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8466, 4.3537, 3.0250, 2.3370, 2.7665, 2.4871, 4.6373, 3.6217], device='cuda:2'), covar=tensor([0.2766, 0.0533, 0.1782, 0.2788, 0.2702, 0.2084, 0.0374, 0.1230], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0256, 0.0292, 0.0296, 0.0280, 0.0242, 0.0280, 0.0317], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 19:53:13,450 INFO [train.py:904] (2/8) Epoch 19, batch 200, loss[loss=0.183, simple_loss=0.2615, pruned_loss=0.0523, over 16642.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2688, pruned_loss=0.0497, over 2090202.15 frames. ], batch size: 134, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:53:34,275 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:42,591 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:46,225 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6622, 2.8774, 2.7585, 4.9874, 4.1253, 4.4232, 1.6455, 3.1682], device='cuda:2'), covar=tensor([0.1489, 0.0798, 0.1178, 0.0191, 0.0192, 0.0375, 0.1687, 0.0766], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0168, 0.0189, 0.0175, 0.0196, 0.0210, 0.0194, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 19:54:00,497 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:06,773 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 19:54:21,114 INFO [train.py:904] (2/8) Epoch 19, batch 250, loss[loss=0.1649, simple_loss=0.2631, pruned_loss=0.03332, over 16731.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2663, pruned_loss=0.04898, over 2351947.16 frames. ], batch size: 57, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:54:32,985 INFO [optim.py:368] (2/8) 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 19:55:08,839 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:55:30,565 INFO [train.py:904] (2/8) Epoch 19, batch 300, loss[loss=0.16, simple_loss=0.248, pruned_loss=0.036, over 17238.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2622, pruned_loss=0.04666, over 2565641.87 frames. ], batch size: 45, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:08,287 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7634, 2.6574, 2.2224, 2.3697, 3.0160, 2.7910, 3.4794, 3.2800], device='cuda:2'), covar=tensor([0.0143, 0.0483, 0.0606, 0.0540, 0.0365, 0.0463, 0.0277, 0.0310], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0232, 0.0224, 0.0224, 0.0232, 0.0229, 0.0232, 0.0223], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:56:41,129 INFO [train.py:904] (2/8) Epoch 19, batch 350, loss[loss=0.1679, simple_loss=0.2673, pruned_loss=0.03421, over 17119.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2594, pruned_loss=0.04463, over 2738111.47 frames. ], batch size: 47, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:52,337 INFO [optim.py:368] (2/8) 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 19:56:58,695 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4962, 3.4394, 3.8290, 1.9891, 3.9153, 3.9325, 3.1636, 2.8738], device='cuda:2'), covar=tensor([0.0816, 0.0243, 0.0172, 0.1156, 0.0097, 0.0202, 0.0383, 0.0457], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0105, 0.0093, 0.0139, 0.0076, 0.0120, 0.0125, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 19:57:03,790 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 19:57:51,264 INFO [train.py:904] (2/8) Epoch 19, batch 400, loss[loss=0.1795, simple_loss=0.2542, pruned_loss=0.05241, over 16276.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2591, pruned_loss=0.04544, over 2863290.62 frames. ], batch size: 165, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:58:33,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7838, 4.5901, 4.8309, 5.0073, 5.1872, 4.4926, 5.1926, 5.1981], device='cuda:2'), covar=tensor([0.1939, 0.1333, 0.1532, 0.0781, 0.0568, 0.1048, 0.0557, 0.0613], device='cuda:2'), in_proj_covar=tensor([0.0601, 0.0739, 0.0862, 0.0756, 0.0562, 0.0592, 0.0615, 0.0708], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 19:59:03,157 INFO [train.py:904] (2/8) Epoch 19, batch 450, loss[loss=0.1829, simple_loss=0.2499, pruned_loss=0.05788, over 16210.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.257, pruned_loss=0.04376, over 2964920.77 frames. ], batch size: 165, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:59:14,099 INFO [optim.py:368] (2/8) 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 19:59:28,188 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:12,284 INFO [train.py:904] (2/8) Epoch 19, batch 500, loss[loss=0.1648, simple_loss=0.2565, pruned_loss=0.0366, over 17071.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.256, pruned_loss=0.04276, over 3049152.80 frames. ], batch size: 53, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:00:27,750 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,914 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,928 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:51,418 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2952, 2.2559, 2.8262, 3.1246, 2.9079, 3.5515, 2.5156, 3.5383], device='cuda:2'), covar=tensor([0.0202, 0.0448, 0.0307, 0.0303, 0.0315, 0.0186, 0.0426, 0.0180], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0188, 0.0175, 0.0177, 0.0187, 0.0145, 0.0191, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:01:23,245 INFO [train.py:904] (2/8) Epoch 19, batch 550, loss[loss=0.1852, simple_loss=0.276, pruned_loss=0.04719, over 17062.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.255, pruned_loss=0.04219, over 3109511.39 frames. ], batch size: 55, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:01:29,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8168, 3.8492, 4.1358, 4.1042, 4.1211, 3.8877, 3.9251, 3.8947], device='cuda:2'), covar=tensor([0.0432, 0.0787, 0.0448, 0.0443, 0.0503, 0.0536, 0.0787, 0.0565], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0427, 0.0416, 0.0392, 0.0463, 0.0439, 0.0529, 0.0350], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 20:01:34,916 INFO [optim.py:368] (2/8) 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 20:02:19,889 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5743, 1.7407, 2.2228, 2.3637, 2.5070, 2.5087, 1.7417, 2.6259], device='cuda:2'), covar=tensor([0.0145, 0.0475, 0.0298, 0.0273, 0.0276, 0.0271, 0.0514, 0.0171], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0188, 0.0175, 0.0177, 0.0187, 0.0145, 0.0191, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:02:32,735 INFO [train.py:904] (2/8) Epoch 19, batch 600, loss[loss=0.171, simple_loss=0.2466, pruned_loss=0.04771, over 16830.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2544, pruned_loss=0.04269, over 3153742.11 frames. ], batch size: 102, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:02:49,008 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-04-30 20:03:08,132 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 20:03:27,720 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0696, 2.5430, 2.0225, 2.2991, 2.9224, 2.6529, 2.9932, 3.0386], device='cuda:2'), covar=tensor([0.0212, 0.0378, 0.0540, 0.0437, 0.0246, 0.0341, 0.0244, 0.0251], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0232, 0.0224, 0.0224, 0.0232, 0.0231, 0.0234, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:03:42,736 INFO [train.py:904] (2/8) Epoch 19, batch 650, loss[loss=0.1538, simple_loss=0.2358, pruned_loss=0.03586, over 15976.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2532, pruned_loss=0.04278, over 3190561.48 frames. ], batch size: 35, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:54,610 INFO [optim.py:368] (2/8) 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 20:04:53,075 INFO [train.py:904] (2/8) Epoch 19, batch 700, loss[loss=0.1533, simple_loss=0.2382, pruned_loss=0.03414, over 16823.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2532, pruned_loss=0.04273, over 3222289.16 frames. ], batch size: 39, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:05:00,980 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:05:50,883 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 20:05:59,948 INFO [train.py:904] (2/8) Epoch 19, batch 750, loss[loss=0.1661, simple_loss=0.2466, pruned_loss=0.04283, over 16271.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2544, pruned_loss=0.04326, over 3230691.93 frames. ], batch size: 165, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:06:11,898 INFO [optim.py:368] (2/8) 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 20:06:25,260 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183469.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:06:39,724 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3990, 5.7645, 5.5334, 5.5635, 5.2172, 5.2011, 5.2207, 5.9124], device='cuda:2'), covar=tensor([0.1383, 0.1026, 0.0994, 0.0816, 0.0882, 0.0710, 0.1081, 0.0956], device='cuda:2'), in_proj_covar=tensor([0.0654, 0.0801, 0.0658, 0.0593, 0.0502, 0.0511, 0.0665, 0.0619], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:07:10,071 INFO [train.py:904] (2/8) Epoch 19, batch 800, loss[loss=0.1727, simple_loss=0.248, pruned_loss=0.0487, over 16266.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2537, pruned_loss=0.0432, over 3247849.92 frames. ], batch size: 165, lr: 3.61e-03, grad_scale: 4.0 2023-04-30 20:07:24,257 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:07:32,540 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:19,002 INFO [train.py:904] (2/8) Epoch 19, batch 850, loss[loss=0.1812, simple_loss=0.2532, pruned_loss=0.05458, over 16831.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2529, pruned_loss=0.04269, over 3272024.79 frames. ], batch size: 116, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:08:20,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6847, 2.8550, 2.7160, 1.7720, 2.4423, 1.8885, 3.0569, 3.1849], device='cuda:2'), covar=tensor([0.0246, 0.0905, 0.0718, 0.2363, 0.1119, 0.1244, 0.0696, 0.0878], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:08:29,786 INFO [optim.py:368] (2/8) 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 20:08:30,865 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183560.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:38,498 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:09:11,426 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8126, 2.6604, 2.7326, 4.7050, 2.5619, 3.0118, 2.7497, 2.7954], device='cuda:2'), covar=tensor([0.1107, 0.3248, 0.2649, 0.0417, 0.3929, 0.2414, 0.3080, 0.3343], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0434, 0.0360, 0.0323, 0.0433, 0.0498, 0.0405, 0.0507], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:09:28,459 INFO [train.py:904] (2/8) Epoch 19, batch 900, loss[loss=0.1908, simple_loss=0.2647, pruned_loss=0.05848, over 16921.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.252, pruned_loss=0.04228, over 3274163.35 frames. ], batch size: 109, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:09:44,199 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:10:35,713 INFO [train.py:904] (2/8) Epoch 19, batch 950, loss[loss=0.1803, simple_loss=0.274, pruned_loss=0.04332, over 17078.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2525, pruned_loss=0.0422, over 3288348.44 frames. ], batch size: 53, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:10:45,885 INFO [optim.py:368] (2/8) 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 20:11:08,225 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:11:09,462 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6773, 2.3794, 2.4702, 4.4873, 2.4325, 2.8054, 2.4992, 2.6011], device='cuda:2'), covar=tensor([0.1138, 0.3787, 0.2957, 0.0440, 0.4059, 0.2652, 0.3601, 0.3605], device='cuda:2'), in_proj_covar=tensor([0.0392, 0.0434, 0.0360, 0.0324, 0.0433, 0.0498, 0.0404, 0.0507], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:11:42,836 INFO [train.py:904] (2/8) Epoch 19, batch 1000, loss[loss=0.1722, simple_loss=0.2778, pruned_loss=0.03333, over 17275.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2512, pruned_loss=0.04167, over 3296307.78 frames. ], batch size: 52, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:12:50,597 INFO [train.py:904] (2/8) Epoch 19, batch 1050, loss[loss=0.1536, simple_loss=0.2515, pruned_loss=0.02789, over 17276.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2502, pruned_loss=0.04136, over 3303734.51 frames. ], batch size: 52, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:13:01,769 INFO [optim.py:368] (2/8) 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 20:13:06,574 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:14:00,540 INFO [train.py:904] (2/8) Epoch 19, batch 1100, loss[loss=0.1531, simple_loss=0.2466, pruned_loss=0.02977, over 17132.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2492, pruned_loss=0.04046, over 3310660.15 frames. ], batch size: 48, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:14:55,875 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 19, batch 1150, loss[loss=0.1499, simple_loss=0.2437, pruned_loss=0.02804, over 17112.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2483, pruned_loss=0.03998, over 3320100.36 frames. ], batch size: 47, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:15:20,304 INFO [optim.py:368] (2/8) 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 20:16:04,238 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2613, 4.1904, 4.1544, 3.9184, 3.9366, 4.2294, 3.9193, 4.0106], device='cuda:2'), covar=tensor([0.0632, 0.0744, 0.0334, 0.0267, 0.0727, 0.0452, 0.0888, 0.0631], device='cuda:2'), in_proj_covar=tensor([0.0291, 0.0416, 0.0342, 0.0331, 0.0352, 0.0387, 0.0235, 0.0406], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:16:18,543 INFO [train.py:904] (2/8) Epoch 19, batch 1200, loss[loss=0.1519, simple_loss=0.2322, pruned_loss=0.03581, over 16794.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2485, pruned_loss=0.03988, over 3310416.03 frames. ], batch size: 39, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:16:19,006 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0130, 2.6532, 2.0637, 2.3986, 2.9884, 2.8096, 3.0380, 3.0776], device='cuda:2'), covar=tensor([0.0226, 0.0349, 0.0512, 0.0417, 0.0250, 0.0300, 0.0282, 0.0260], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0235, 0.0226, 0.0227, 0.0236, 0.0234, 0.0238, 0.0229], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:16:20,046 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183903.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:17:25,280 INFO [train.py:904] (2/8) Epoch 19, batch 1250, loss[loss=0.1592, simple_loss=0.2438, pruned_loss=0.0373, over 16990.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2491, pruned_loss=0.04061, over 3316970.55 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:17:35,896 INFO [optim.py:368] (2/8) 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 20:17:50,866 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183970.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:18:37,490 INFO [train.py:904] (2/8) Epoch 19, batch 1300, loss[loss=0.1713, simple_loss=0.2582, pruned_loss=0.04219, over 17183.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2497, pruned_loss=0.04067, over 3317042.37 frames. ], batch size: 46, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:26,445 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 20:19:44,290 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 19, batch 1350, loss[loss=0.1592, simple_loss=0.2409, pruned_loss=0.03878, over 16502.00 frames. ], tot_loss[loss=0.165, simple_loss=0.25, pruned_loss=0.03999, over 3319783.59 frames. ], batch size: 146, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:55,438 INFO [optim.py:368] (2/8) 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 20:20:03,915 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:20:21,011 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-30 20:20:27,176 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 20:20:55,155 INFO [train.py:904] (2/8) Epoch 19, batch 1400, loss[loss=0.1393, simple_loss=0.2251, pruned_loss=0.02678, over 17007.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2506, pruned_loss=0.04033, over 3319570.15 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:21:09,587 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:10,485 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184112.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:11,881 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3155, 4.0726, 4.4694, 2.4791, 4.6692, 4.7816, 3.5046, 3.6769], device='cuda:2'), covar=tensor([0.0594, 0.0224, 0.0191, 0.0985, 0.0062, 0.0124, 0.0358, 0.0369], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0107, 0.0095, 0.0139, 0.0077, 0.0123, 0.0126, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 20:22:05,080 INFO [train.py:904] (2/8) Epoch 19, batch 1450, loss[loss=0.1673, simple_loss=0.2394, pruned_loss=0.04759, over 12110.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2495, pruned_loss=0.04027, over 3318132.51 frames. ], batch size: 248, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:22:15,457 INFO [optim.py:368] (2/8) 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 20:22:33,108 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8260, 3.9169, 3.0547, 2.3739, 2.6174, 2.4881, 4.1507, 3.4394], device='cuda:2'), covar=tensor([0.2616, 0.0635, 0.1701, 0.2885, 0.2719, 0.2080, 0.0489, 0.1462], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0268, 0.0302, 0.0306, 0.0294, 0.0253, 0.0290, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 20:22:40,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5641, 3.8413, 4.1953, 2.1654, 3.3167, 2.6948, 4.1520, 4.0244], device='cuda:2'), covar=tensor([0.0243, 0.0799, 0.0425, 0.1984, 0.0714, 0.0928, 0.0518, 0.0957], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0159, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:22:42,452 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1776, 4.1472, 4.1254, 3.3884, 4.1533, 1.7202, 3.8997, 3.6725], device='cuda:2'), covar=tensor([0.0138, 0.0117, 0.0187, 0.0317, 0.0114, 0.3022, 0.0151, 0.0282], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0176, 0.0170, 0.0205, 0.0184, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:23:08,679 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184198.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:23:13,288 INFO [train.py:904] (2/8) Epoch 19, batch 1500, loss[loss=0.1619, simple_loss=0.2533, pruned_loss=0.03526, over 17165.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2502, pruned_loss=0.04058, over 3324944.85 frames. ], batch size: 46, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:23:46,667 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-04-30 20:24:05,582 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-30 20:24:22,394 INFO [train.py:904] (2/8) Epoch 19, batch 1550, loss[loss=0.1889, simple_loss=0.278, pruned_loss=0.04988, over 15501.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2514, pruned_loss=0.04117, over 3329700.34 frames. ], batch size: 191, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:34,823 INFO [optim.py:368] (2/8) 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 20:24:47,472 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:25:31,173 INFO [train.py:904] (2/8) Epoch 19, batch 1600, loss[loss=0.188, simple_loss=0.2621, pruned_loss=0.05696, over 16766.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2535, pruned_loss=0.04227, over 3325762.79 frames. ], batch size: 102, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:25:53,350 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:15,652 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 20:26:39,948 INFO [train.py:904] (2/8) Epoch 19, batch 1650, loss[loss=0.1637, simple_loss=0.2548, pruned_loss=0.03635, over 17233.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2547, pruned_loss=0.04236, over 3320699.76 frames. ], batch size: 45, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:26:45,847 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5496, 3.6182, 2.2202, 3.8284, 2.8358, 3.7293, 2.2513, 2.8990], device='cuda:2'), covar=tensor([0.0272, 0.0410, 0.1450, 0.0339, 0.0737, 0.0888, 0.1426, 0.0686], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0177, 0.0195, 0.0160, 0.0175, 0.0215, 0.0202, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:26:50,237 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 20:27:45,803 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-04-30 20:27:49,670 INFO [train.py:904] (2/8) Epoch 19, batch 1700, loss[loss=0.2141, simple_loss=0.2829, pruned_loss=0.07265, over 16709.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2565, pruned_loss=0.04311, over 3308907.97 frames. ], batch size: 124, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:27:55,024 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:03,262 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9206, 4.9975, 5.3896, 5.3949, 5.4033, 5.0444, 5.0196, 4.7781], device='cuda:2'), covar=tensor([0.0387, 0.0522, 0.0510, 0.0501, 0.0599, 0.0546, 0.0947, 0.0486], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0440, 0.0427, 0.0402, 0.0476, 0.0450, 0.0542, 0.0358], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 20:28:14,044 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:32,385 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-04-30 20:28:58,451 INFO [train.py:904] (2/8) Epoch 19, batch 1750, loss[loss=0.1515, simple_loss=0.2449, pruned_loss=0.02901, over 17189.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2572, pruned_loss=0.04321, over 3321255.04 frames. ], batch size: 46, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:29:10,990 INFO [optim.py:368] (2/8) 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 20:29:45,738 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184486.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:00,556 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2331, 4.2929, 4.6010, 4.6069, 4.6204, 4.3442, 4.3421, 4.2282], device='cuda:2'), covar=tensor([0.0349, 0.0553, 0.0381, 0.0386, 0.0498, 0.0419, 0.0863, 0.0588], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0441, 0.0429, 0.0404, 0.0479, 0.0452, 0.0546, 0.0360], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 20:30:02,694 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:07,526 INFO [train.py:904] (2/8) Epoch 19, batch 1800, loss[loss=0.1555, simple_loss=0.2545, pruned_loss=0.02828, over 17278.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2594, pruned_loss=0.04376, over 3314805.64 frames. ], batch size: 52, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:30:47,152 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-04-30 20:31:07,262 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184546.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:09,234 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184547.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:16,693 INFO [train.py:904] (2/8) Epoch 19, batch 1850, loss[loss=0.178, simple_loss=0.2567, pruned_loss=0.04965, over 16926.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2592, pruned_loss=0.04339, over 3319675.07 frames. ], batch size: 90, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:31:20,762 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 20:31:29,852 INFO [optim.py:368] (2/8) 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 20:32:26,187 INFO [train.py:904] (2/8) Epoch 19, batch 1900, loss[loss=0.1834, simple_loss=0.2751, pruned_loss=0.04591, over 16692.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2585, pruned_loss=0.04302, over 3326814.30 frames. ], batch size: 62, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:32:53,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9911, 4.6132, 4.5297, 3.2292, 3.8248, 4.4532, 4.0038, 2.8433], device='cuda:2'), covar=tensor([0.0444, 0.0049, 0.0036, 0.0340, 0.0119, 0.0095, 0.0079, 0.0403], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0106, 0.0092, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 20:33:25,454 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7664, 4.8553, 5.0296, 4.8297, 4.8958, 5.4977, 4.9906, 4.6462], device='cuda:2'), covar=tensor([0.1359, 0.2159, 0.2359, 0.2327, 0.2845, 0.1107, 0.1653, 0.2772], device='cuda:2'), in_proj_covar=tensor([0.0406, 0.0582, 0.0646, 0.0490, 0.0655, 0.0684, 0.0504, 0.0656], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 20:33:35,922 INFO [train.py:904] (2/8) Epoch 19, batch 1950, loss[loss=0.1782, simple_loss=0.2522, pruned_loss=0.05204, over 16852.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2577, pruned_loss=0.04229, over 3327879.21 frames. ], batch size: 116, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:48,870 INFO [optim.py:368] (2/8) 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 20:34:14,601 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8969, 2.0785, 2.5855, 2.8254, 2.6441, 3.4207, 2.2536, 3.3652], device='cuda:2'), covar=tensor([0.0231, 0.0459, 0.0305, 0.0300, 0.0347, 0.0175, 0.0465, 0.0163], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0193, 0.0179, 0.0182, 0.0192, 0.0149, 0.0195, 0.0144], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:34:45,103 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5770, 4.5207, 4.4929, 4.2305, 4.2228, 4.5608, 4.3246, 4.3098], device='cuda:2'), covar=tensor([0.0676, 0.0784, 0.0301, 0.0276, 0.0833, 0.0487, 0.0521, 0.0664], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0422, 0.0348, 0.0337, 0.0357, 0.0393, 0.0237, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:34:47,042 INFO [train.py:904] (2/8) Epoch 19, batch 2000, loss[loss=0.2103, simple_loss=0.2807, pruned_loss=0.06998, over 16350.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2574, pruned_loss=0.04241, over 3319082.98 frames. ], batch size: 165, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:34:53,112 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184706.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:06,420 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184715.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:43,677 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4139, 4.6699, 4.6474, 3.4744, 3.8906, 4.5808, 4.0006, 2.8717], device='cuda:2'), covar=tensor([0.0356, 0.0054, 0.0037, 0.0306, 0.0112, 0.0075, 0.0077, 0.0404], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0105, 0.0092, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 20:35:56,927 INFO [train.py:904] (2/8) Epoch 19, batch 2050, loss[loss=0.1821, simple_loss=0.2616, pruned_loss=0.05125, over 16903.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2579, pruned_loss=0.04239, over 3323336.58 frames. ], batch size: 90, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:35:57,438 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7421, 4.0880, 2.8012, 2.2689, 2.7571, 2.5280, 4.5743, 3.5328], device='cuda:2'), covar=tensor([0.2855, 0.0698, 0.2027, 0.2816, 0.2713, 0.2033, 0.0349, 0.1399], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0267, 0.0301, 0.0304, 0.0293, 0.0251, 0.0289, 0.0331], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 20:35:59,926 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184754.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:36:10,010 INFO [optim.py:368] (2/8) 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 20:36:35,830 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8015, 4.3238, 4.3257, 3.1473, 3.6264, 4.2328, 3.8732, 2.2725], device='cuda:2'), covar=tensor([0.0449, 0.0059, 0.0041, 0.0317, 0.0129, 0.0089, 0.0087, 0.0502], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0105, 0.0092, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 20:37:10,296 INFO [train.py:904] (2/8) Epoch 19, batch 2100, loss[loss=0.1571, simple_loss=0.2493, pruned_loss=0.0325, over 17095.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2589, pruned_loss=0.04306, over 3317875.90 frames. ], batch size: 47, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:37:13,294 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 20:37:18,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6856, 4.7421, 5.1024, 5.0728, 5.1164, 4.7887, 4.7787, 4.5699], device='cuda:2'), covar=tensor([0.0317, 0.0501, 0.0350, 0.0409, 0.0430, 0.0376, 0.0759, 0.0496], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0437, 0.0426, 0.0402, 0.0473, 0.0451, 0.0542, 0.0357], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 20:38:05,009 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184842.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:38:18,977 INFO [train.py:904] (2/8) Epoch 19, batch 2150, loss[loss=0.1918, simple_loss=0.279, pruned_loss=0.05229, over 16696.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2601, pruned_loss=0.04345, over 3322806.21 frames. ], batch size: 62, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:38:30,739 INFO [optim.py:368] (2/8) 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 20:38:38,181 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184866.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:39:19,531 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-30 20:39:27,103 INFO [train.py:904] (2/8) Epoch 19, batch 2200, loss[loss=0.1625, simple_loss=0.2576, pruned_loss=0.03372, over 16552.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2605, pruned_loss=0.04395, over 3313623.73 frames. ], batch size: 68, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:01,196 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:17,870 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:35,942 INFO [train.py:904] (2/8) Epoch 19, batch 2250, loss[loss=0.1865, simple_loss=0.2745, pruned_loss=0.04928, over 16502.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2612, pruned_loss=0.04467, over 3316723.98 frames. ], batch size: 75, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:48,429 INFO [optim.py:368] (2/8) 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 20:41:32,216 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 20:41:43,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3060, 3.0067, 3.3141, 1.9082, 3.4431, 3.4365, 2.7949, 2.6396], device='cuda:2'), covar=tensor([0.0800, 0.0274, 0.0219, 0.1078, 0.0115, 0.0238, 0.0480, 0.0451], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0141, 0.0079, 0.0126, 0.0129, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:41:45,043 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185000.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:41:46,801 INFO [train.py:904] (2/8) Epoch 19, batch 2300, loss[loss=0.1703, simple_loss=0.2506, pruned_loss=0.04503, over 16869.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2619, pruned_loss=0.04527, over 3300255.18 frames. ], batch size: 96, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:42:05,906 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:42:56,744 INFO [train.py:904] (2/8) Epoch 19, batch 2350, loss[loss=0.1792, simple_loss=0.2689, pruned_loss=0.04474, over 17119.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2622, pruned_loss=0.04548, over 3297030.46 frames. ], batch size: 48, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:43:08,902 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:43:09,638 INFO [optim.py:368] (2/8) 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 20:43:12,101 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:43:45,435 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9500, 3.7833, 4.2713, 2.0858, 4.4958, 4.5526, 3.2634, 3.3906], device='cuda:2'), covar=tensor([0.0664, 0.0250, 0.0184, 0.1095, 0.0069, 0.0180, 0.0432, 0.0384], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0141, 0.0079, 0.0126, 0.0129, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:43:56,740 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3258, 5.2313, 5.1350, 4.6638, 4.7486, 5.1929, 5.1540, 4.7505], device='cuda:2'), covar=tensor([0.0546, 0.0505, 0.0281, 0.0372, 0.1083, 0.0529, 0.0282, 0.0827], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0425, 0.0350, 0.0340, 0.0359, 0.0395, 0.0239, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:44:06,534 INFO [train.py:904] (2/8) Epoch 19, batch 2400, loss[loss=0.1827, simple_loss=0.2694, pruned_loss=0.04803, over 16091.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2623, pruned_loss=0.0452, over 3311659.89 frames. ], batch size: 35, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:44:33,383 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185121.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 20:44:43,171 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:45:02,140 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 19, batch 2450, loss[loss=0.1865, simple_loss=0.2713, pruned_loss=0.05085, over 16237.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2633, pruned_loss=0.04514, over 3306267.66 frames. ], batch size: 165, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:45:27,049 INFO [optim.py:368] (2/8) 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 20:46:08,235 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:08,467 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:23,967 INFO [train.py:904] (2/8) Epoch 19, batch 2500, loss[loss=0.1663, simple_loss=0.2469, pruned_loss=0.04284, over 16786.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2629, pruned_loss=0.04539, over 3299572.26 frames. ], batch size: 102, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:46:51,192 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185222.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:47:32,266 INFO [train.py:904] (2/8) Epoch 19, batch 2550, loss[loss=0.1601, simple_loss=0.2375, pruned_loss=0.04132, over 16814.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2617, pruned_loss=0.04473, over 3315267.37 frames. ], batch size: 96, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:47:44,040 INFO [optim.py:368] (2/8) 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 20:48:31,909 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:48:40,163 INFO [train.py:904] (2/8) Epoch 19, batch 2600, loss[loss=0.1578, simple_loss=0.2463, pruned_loss=0.03465, over 17054.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2612, pruned_loss=0.04425, over 3322332.94 frames. ], batch size: 41, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:48:52,019 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7059, 3.8214, 2.3101, 4.2706, 2.8625, 4.1913, 2.4491, 3.0189], device='cuda:2'), covar=tensor([0.0287, 0.0391, 0.1535, 0.0310, 0.0833, 0.0695, 0.1468, 0.0775], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0178, 0.0196, 0.0163, 0.0177, 0.0219, 0.0203, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:48:59,137 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2167, 4.2234, 4.5655, 4.5433, 4.6033, 4.2853, 4.3274, 4.2474], device='cuda:2'), covar=tensor([0.0357, 0.0634, 0.0394, 0.0432, 0.0460, 0.0476, 0.0744, 0.0524], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0444, 0.0431, 0.0403, 0.0476, 0.0453, 0.0548, 0.0358], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 20:49:50,076 INFO [train.py:904] (2/8) Epoch 19, batch 2650, loss[loss=0.1622, simple_loss=0.2507, pruned_loss=0.03687, over 17225.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2605, pruned_loss=0.04359, over 3324747.41 frames. ], batch size: 45, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:50:03,485 INFO [optim.py:368] (2/8) 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 20:50:59,028 INFO [train.py:904] (2/8) Epoch 19, batch 2700, loss[loss=0.1658, simple_loss=0.2688, pruned_loss=0.03143, over 17092.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2611, pruned_loss=0.04336, over 3322597.41 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:51:04,741 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5904, 3.6553, 2.2761, 3.9682, 2.8934, 3.9191, 2.3094, 2.8815], device='cuda:2'), covar=tensor([0.0284, 0.0364, 0.1458, 0.0264, 0.0754, 0.0612, 0.1440, 0.0748], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0178, 0.0195, 0.0163, 0.0176, 0.0219, 0.0203, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:51:19,125 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185416.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 20:52:08,642 INFO [train.py:904] (2/8) Epoch 19, batch 2750, loss[loss=0.1526, simple_loss=0.2476, pruned_loss=0.02882, over 17194.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2613, pruned_loss=0.04332, over 3332359.09 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:52:20,528 INFO [optim.py:368] (2/8) 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 20:52:30,891 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:52:37,866 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 20:52:54,424 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185485.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:52:57,906 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1620, 4.9583, 5.1924, 5.3698, 5.5536, 4.7801, 5.5154, 5.5179], device='cuda:2'), covar=tensor([0.1792, 0.1334, 0.1591, 0.0779, 0.0531, 0.0981, 0.0718, 0.0668], device='cuda:2'), in_proj_covar=tensor([0.0646, 0.0800, 0.0938, 0.0815, 0.0606, 0.0640, 0.0654, 0.0764], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:53:18,344 INFO [train.py:904] (2/8) Epoch 19, batch 2800, loss[loss=0.1553, simple_loss=0.2534, pruned_loss=0.02854, over 17249.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2606, pruned_loss=0.04275, over 3333469.30 frames. ], batch size: 52, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:53:46,671 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185522.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:56,380 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:54:28,304 INFO [train.py:904] (2/8) Epoch 19, batch 2850, loss[loss=0.19, simple_loss=0.28, pruned_loss=0.05004, over 17090.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2605, pruned_loss=0.04273, over 3336897.22 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:54:29,977 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0899, 3.0946, 1.9546, 3.2601, 2.4406, 3.3074, 2.1351, 2.6017], device='cuda:2'), covar=tensor([0.0285, 0.0376, 0.1497, 0.0299, 0.0735, 0.0581, 0.1274, 0.0682], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0195, 0.0163, 0.0177, 0.0220, 0.0203, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 20:54:41,475 INFO [optim.py:368] (2/8) 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 20:54:53,419 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:21,034 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8338, 4.6287, 4.8298, 5.0481, 5.2320, 4.6135, 5.1820, 5.2268], device='cuda:2'), covar=tensor([0.1662, 0.1235, 0.1762, 0.0772, 0.0529, 0.0976, 0.0595, 0.0600], device='cuda:2'), in_proj_covar=tensor([0.0646, 0.0800, 0.0939, 0.0815, 0.0607, 0.0641, 0.0656, 0.0765], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 20:55:28,817 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185595.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:39,058 INFO [train.py:904] (2/8) Epoch 19, batch 2900, loss[loss=0.1769, simple_loss=0.2477, pruned_loss=0.05305, over 16864.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2602, pruned_loss=0.04322, over 3330285.35 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:56:00,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9201, 4.5203, 3.1126, 2.4257, 3.1416, 2.5907, 4.8375, 3.8447], device='cuda:2'), covar=tensor([0.2736, 0.0594, 0.1830, 0.2735, 0.2533, 0.1985, 0.0390, 0.1243], device='cuda:2'), in_proj_covar=tensor([0.0320, 0.0266, 0.0301, 0.0303, 0.0293, 0.0251, 0.0288, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 20:56:36,417 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:56:47,750 INFO [train.py:904] (2/8) Epoch 19, batch 2950, loss[loss=0.1792, simple_loss=0.256, pruned_loss=0.0512, over 16741.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2603, pruned_loss=0.0439, over 3312594.73 frames. ], batch size: 83, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:56:51,668 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-04-30 20:57:00,933 INFO [optim.py:368] (2/8) 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 20:57:58,854 INFO [train.py:904] (2/8) Epoch 19, batch 3000, loss[loss=0.1833, simple_loss=0.2743, pruned_loss=0.04616, over 17027.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2601, pruned_loss=0.04405, over 3314069.04 frames. ], batch size: 55, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:58,854 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 20:58:07,634 INFO [train.py:938] (2/8) Epoch 19, validation: loss=0.1362, simple_loss=0.2416, pruned_loss=0.01538, over 944034.00 frames. 2023-04-30 20:58:07,634 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 20:58:28,276 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185716.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 20:59:16,277 INFO [train.py:904] (2/8) Epoch 19, batch 3050, loss[loss=0.1645, simple_loss=0.2646, pruned_loss=0.03222, over 17230.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2603, pruned_loss=0.04428, over 3308061.54 frames. ], batch size: 52, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:59:28,088 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:59:29,997 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:03,972 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:23,294 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185799.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:00:26,865 INFO [train.py:904] (2/8) Epoch 19, batch 3100, loss[loss=0.1899, simple_loss=0.258, pruned_loss=0.06094, over 16354.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2599, pruned_loss=0.04449, over 3310039.34 frames. ], batch size: 146, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:00:51,361 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185820.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:00:51,555 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 21:00:56,478 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185824.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:09,059 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185833.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:18,887 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6359, 2.6912, 2.2989, 2.4564, 3.0061, 2.7938, 3.3179, 3.2468], device='cuda:2'), covar=tensor([0.0151, 0.0399, 0.0485, 0.0456, 0.0255, 0.0382, 0.0260, 0.0255], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0235, 0.0225, 0.0226, 0.0235, 0.0234, 0.0240, 0.0230], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:01:34,487 INFO [train.py:904] (2/8) Epoch 19, batch 3150, loss[loss=0.183, simple_loss=0.2615, pruned_loss=0.05227, over 16680.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2592, pruned_loss=0.04412, over 3313642.32 frames. ], batch size: 134, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:01:46,071 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185860.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:01:47,102 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 19, batch 3200, loss[loss=0.1681, simple_loss=0.2476, pruned_loss=0.04429, over 16702.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2579, pruned_loss=0.04319, over 3317976.00 frames. ], batch size: 89, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:03:43,192 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9271, 4.7130, 4.9151, 5.1234, 5.3033, 4.6495, 5.2744, 5.2966], device='cuda:2'), covar=tensor([0.1749, 0.1267, 0.1693, 0.0744, 0.0578, 0.0895, 0.0555, 0.0621], device='cuda:2'), in_proj_covar=tensor([0.0654, 0.0808, 0.0950, 0.0825, 0.0615, 0.0646, 0.0663, 0.0772], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:03:53,535 INFO [train.py:904] (2/8) Epoch 19, batch 3250, loss[loss=0.129, simple_loss=0.2159, pruned_loss=0.02105, over 17016.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2569, pruned_loss=0.04307, over 3322594.37 frames. ], batch size: 41, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:04:06,305 INFO [optim.py:368] (2/8) 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 21:05:07,473 INFO [train.py:904] (2/8) Epoch 19, batch 3300, loss[loss=0.1836, simple_loss=0.2684, pruned_loss=0.04938, over 16731.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2582, pruned_loss=0.0431, over 3323590.08 frames. ], batch size: 83, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:06:15,806 INFO [train.py:904] (2/8) Epoch 19, batch 3350, loss[loss=0.1781, simple_loss=0.2593, pruned_loss=0.04845, over 16895.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2585, pruned_loss=0.04265, over 3324822.70 frames. ], batch size: 116, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:06:27,998 INFO [optim.py:368] (2/8) 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 21:07:23,660 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 19, batch 3400, loss[loss=0.1707, simple_loss=0.2651, pruned_loss=0.03817, over 17037.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2587, pruned_loss=0.04265, over 3328616.44 frames. ], batch size: 53, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:07:33,736 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5294, 3.5021, 3.4483, 2.7920, 3.3369, 2.0820, 3.0720, 2.7618], device='cuda:2'), covar=tensor([0.0137, 0.0125, 0.0179, 0.0223, 0.0103, 0.2378, 0.0125, 0.0256], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0150, 0.0195, 0.0179, 0.0173, 0.0205, 0.0187, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:07:42,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7361, 3.8038, 2.3114, 4.1516, 2.9456, 4.1245, 2.5541, 2.9974], device='cuda:2'), covar=tensor([0.0264, 0.0346, 0.1547, 0.0342, 0.0806, 0.0616, 0.1363, 0.0691], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0180, 0.0197, 0.0165, 0.0178, 0.0222, 0.0204, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:07:44,668 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186115.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:07:57,631 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186124.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:34,722 INFO [train.py:904] (2/8) Epoch 19, batch 3450, loss[loss=0.1482, simple_loss=0.2384, pruned_loss=0.02899, over 17238.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2576, pruned_loss=0.04233, over 3333465.88 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:08:35,228 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5215, 3.8097, 4.1343, 2.2554, 3.3557, 2.6057, 4.0924, 3.9218], device='cuda:2'), covar=tensor([0.0284, 0.0887, 0.0439, 0.1881, 0.0732, 0.0936, 0.0517, 0.0941], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0162, 0.0165, 0.0151, 0.0143, 0.0128, 0.0143, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:08:38,532 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186155.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:08:44,179 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186158.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:47,841 INFO [optim.py:368] (2/8) 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 21:08:48,266 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:56,660 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6514, 2.3291, 1.7289, 2.0691, 2.6817, 2.5321, 2.8269, 2.8533], device='cuda:2'), covar=tensor([0.0224, 0.0498, 0.0728, 0.0561, 0.0314, 0.0391, 0.0257, 0.0317], device='cuda:2'), in_proj_covar=tensor([0.0206, 0.0236, 0.0226, 0.0227, 0.0237, 0.0235, 0.0243, 0.0231], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:09:02,955 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:45,135 INFO [train.py:904] (2/8) Epoch 19, batch 3500, loss[loss=0.1677, simple_loss=0.249, pruned_loss=0.04325, over 16820.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2572, pruned_loss=0.04218, over 3304138.82 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:10:08,921 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:10:38,839 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 21:10:49,035 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0093, 2.0343, 2.5567, 2.8964, 2.7649, 3.0830, 2.2007, 3.1964], device='cuda:2'), covar=tensor([0.0203, 0.0447, 0.0291, 0.0261, 0.0293, 0.0227, 0.0458, 0.0142], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0192, 0.0179, 0.0182, 0.0193, 0.0152, 0.0195, 0.0146], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:10:55,188 INFO [train.py:904] (2/8) Epoch 19, batch 3550, loss[loss=0.1495, simple_loss=0.2417, pruned_loss=0.02864, over 17195.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2566, pruned_loss=0.04207, over 3294699.10 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:11:00,699 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 21:11:05,154 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186259.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:11:06,978 INFO [optim.py:368] (2/8) 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 21:11:10,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1889, 2.1186, 2.2545, 3.9624, 2.1473, 2.5435, 2.2143, 2.3403], device='cuda:2'), covar=tensor([0.1437, 0.3980, 0.2880, 0.0542, 0.4028, 0.2575, 0.4080, 0.3100], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0439, 0.0362, 0.0327, 0.0434, 0.0506, 0.0409, 0.0513], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:12:03,397 INFO [train.py:904] (2/8) Epoch 19, batch 3600, loss[loss=0.1524, simple_loss=0.2377, pruned_loss=0.03359, over 16478.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2552, pruned_loss=0.04179, over 3295132.30 frames. ], batch size: 68, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:12:28,459 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:13:13,927 INFO [train.py:904] (2/8) Epoch 19, batch 3650, loss[loss=0.1562, simple_loss=0.2439, pruned_loss=0.03426, over 17247.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2547, pruned_loss=0.04221, over 3300355.46 frames. ], batch size: 43, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:13:27,661 INFO [optim.py:368] (2/8) 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 21:14:16,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6724, 3.7039, 2.9502, 2.2286, 2.3551, 2.3150, 3.7410, 3.2339], device='cuda:2'), covar=tensor([0.2516, 0.0583, 0.1545, 0.2737, 0.2701, 0.2104, 0.0516, 0.1363], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0268, 0.0302, 0.0307, 0.0298, 0.0254, 0.0291, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 21:14:28,525 INFO [train.py:904] (2/8) Epoch 19, batch 3700, loss[loss=0.1561, simple_loss=0.2383, pruned_loss=0.03698, over 16333.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2532, pruned_loss=0.04366, over 3278835.36 frames. ], batch size: 165, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:14:48,801 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186415.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:14:57,197 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:15:26,557 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5820, 3.4903, 2.7810, 2.2168, 2.2545, 2.2855, 3.5253, 3.1267], device='cuda:2'), covar=tensor([0.2634, 0.0658, 0.1699, 0.2870, 0.3019, 0.2076, 0.0560, 0.1385], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0268, 0.0301, 0.0307, 0.0298, 0.0254, 0.0291, 0.0334], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 21:15:41,467 INFO [train.py:904] (2/8) Epoch 19, batch 3750, loss[loss=0.1992, simple_loss=0.2729, pruned_loss=0.06277, over 11679.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2538, pruned_loss=0.04512, over 3265002.72 frames. ], batch size: 246, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:15:46,603 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186455.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:15:47,760 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:15:55,022 INFO [optim.py:368] (2/8) 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 21:15:57,662 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186463.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:15,914 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186475.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:26,552 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:42,301 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9202, 2.7742, 2.6104, 4.1621, 3.3906, 4.0870, 1.6483, 2.9168], device='cuda:2'), covar=tensor([0.1316, 0.0665, 0.1091, 0.0190, 0.0184, 0.0376, 0.1492, 0.0830], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0185, 0.0205, 0.0217, 0.0197, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:16:49,467 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9332, 5.3241, 5.4670, 5.2548, 5.2509, 5.8385, 5.3451, 5.0394], device='cuda:2'), covar=tensor([0.1108, 0.1781, 0.1773, 0.1823, 0.2288, 0.0874, 0.1382, 0.2530], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0586, 0.0647, 0.0490, 0.0654, 0.0683, 0.0507, 0.0657], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 21:16:53,853 INFO [train.py:904] (2/8) Epoch 19, batch 3800, loss[loss=0.1606, simple_loss=0.2355, pruned_loss=0.04285, over 16865.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.254, pruned_loss=0.0458, over 3268023.10 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:16:55,942 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186503.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:16:59,407 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 21:17:11,597 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:15,459 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:29,881 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9482, 5.0031, 5.3195, 5.3001, 5.3710, 4.9827, 4.9343, 4.6733], device='cuda:2'), covar=tensor([0.0330, 0.0450, 0.0412, 0.0419, 0.0412, 0.0369, 0.0979, 0.0508], device='cuda:2'), in_proj_covar=tensor([0.0405, 0.0446, 0.0431, 0.0403, 0.0475, 0.0455, 0.0549, 0.0363], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 21:17:43,644 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186536.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:47,922 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186539.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:04,512 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9561, 2.8776, 2.7511, 4.4822, 3.5637, 4.2335, 1.7104, 3.0782], device='cuda:2'), covar=tensor([0.1295, 0.0679, 0.1058, 0.0182, 0.0212, 0.0334, 0.1475, 0.0800], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0172, 0.0191, 0.0185, 0.0204, 0.0216, 0.0196, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:18:05,819 INFO [train.py:904] (2/8) Epoch 19, batch 3850, loss[loss=0.1557, simple_loss=0.2313, pruned_loss=0.04002, over 16746.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2545, pruned_loss=0.04652, over 3274002.74 frames. ], batch size: 83, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:18:22,164 INFO [optim.py:368] (2/8) 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 21:18:28,844 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:44,983 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:49,705 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7827, 4.7472, 4.7233, 4.1016, 4.7540, 1.7983, 4.5221, 4.3301], device='cuda:2'), covar=tensor([0.0117, 0.0101, 0.0163, 0.0327, 0.0082, 0.2704, 0.0144, 0.0225], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0150, 0.0196, 0.0179, 0.0173, 0.0205, 0.0187, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:18:56,169 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186586.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:11,498 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 21:19:16,616 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186600.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:18,739 INFO [train.py:904] (2/8) Epoch 19, batch 3900, loss[loss=0.1741, simple_loss=0.248, pruned_loss=0.05014, over 16873.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2538, pruned_loss=0.04704, over 3277720.34 frames. ], batch size: 116, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:19:37,489 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:55,073 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:01,757 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 21:20:15,282 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 21:20:20,735 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:28,339 INFO [train.py:904] (2/8) Epoch 19, batch 3950, loss[loss=0.184, simple_loss=0.2652, pruned_loss=0.05145, over 16309.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2538, pruned_loss=0.04765, over 3284120.71 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 4.0 2023-04-30 21:20:42,972 INFO [optim.py:368] (2/8) 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 21:21:39,186 INFO [train.py:904] (2/8) Epoch 19, batch 4000, loss[loss=0.1589, simple_loss=0.2468, pruned_loss=0.03548, over 16884.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2542, pruned_loss=0.04815, over 3281925.03 frames. ], batch size: 96, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:21:48,985 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186709.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:22:24,955 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3827, 2.4655, 2.4775, 4.2748, 2.2940, 2.8247, 2.5075, 2.6237], device='cuda:2'), covar=tensor([0.1177, 0.3116, 0.2470, 0.0396, 0.3533, 0.2206, 0.2988, 0.2830], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0441, 0.0362, 0.0328, 0.0434, 0.0509, 0.0410, 0.0516], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:22:49,933 INFO [train.py:904] (2/8) Epoch 19, batch 4050, loss[loss=0.1678, simple_loss=0.2518, pruned_loss=0.04184, over 16577.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2553, pruned_loss=0.04779, over 3280584.59 frames. ], batch size: 57, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:22:57,386 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:06,014 INFO [optim.py:368] (2/8) 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 21:23:16,236 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:26,483 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:03,619 INFO [train.py:904] (2/8) Epoch 19, batch 4100, loss[loss=0.2029, simple_loss=0.2774, pruned_loss=0.06422, over 16284.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2572, pruned_loss=0.04751, over 3272611.45 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:24:07,124 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:21,725 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:46,678 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:59,892 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186840.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:17,894 INFO [train.py:904] (2/8) Epoch 19, batch 4150, loss[loss=0.208, simple_loss=0.2882, pruned_loss=0.06392, over 16970.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2636, pruned_loss=0.04963, over 3231393.26 frames. ], batch size: 55, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:25:28,317 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8151, 3.7800, 3.9290, 3.7004, 3.8404, 4.2469, 3.9242, 3.5915], device='cuda:2'), covar=tensor([0.1808, 0.2077, 0.2020, 0.2192, 0.2457, 0.1524, 0.1387, 0.2335], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0582, 0.0641, 0.0486, 0.0648, 0.0679, 0.0502, 0.0648], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 21:25:33,650 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:34,923 INFO [optim.py:368] (2/8) 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 21:25:35,526 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8379, 3.2394, 3.1209, 2.0082, 3.0254, 3.1746, 2.9635, 1.8303], device='cuda:2'), covar=tensor([0.0575, 0.0058, 0.0072, 0.0444, 0.0106, 0.0126, 0.0101, 0.0512], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0080, 0.0081, 0.0132, 0.0095, 0.0105, 0.0092, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 21:25:49,712 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186873.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:10,336 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8780, 4.1512, 3.9943, 4.0372, 3.7106, 3.7809, 3.8226, 4.1489], device='cuda:2'), covar=tensor([0.1214, 0.0921, 0.0950, 0.0809, 0.0792, 0.1648, 0.0931, 0.1046], device='cuda:2'), in_proj_covar=tensor([0.0659, 0.0810, 0.0666, 0.0605, 0.0508, 0.0516, 0.0676, 0.0622], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:26:23,062 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:31,929 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186901.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:26:32,668 INFO [train.py:904] (2/8) Epoch 19, batch 4200, loss[loss=0.2206, simple_loss=0.3102, pruned_loss=0.06547, over 16766.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2705, pruned_loss=0.05131, over 3198665.29 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:26:47,058 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:53,457 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:05,449 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6860, 2.6731, 2.5788, 4.0943, 2.8231, 3.9050, 1.6928, 2.8104], device='cuda:2'), covar=tensor([0.1368, 0.0862, 0.1182, 0.0189, 0.0282, 0.0511, 0.1649, 0.0895], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0184, 0.0205, 0.0215, 0.0196, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:27:06,396 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:49,658 INFO [train.py:904] (2/8) Epoch 19, batch 4250, loss[loss=0.1843, simple_loss=0.2781, pruned_loss=0.04528, over 16721.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2731, pruned_loss=0.05038, over 3199856.88 frames. ], batch size: 134, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:28:05,619 INFO [optim.py:368] (2/8) 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 21:28:06,528 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:28:19,137 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:02,899 INFO [train.py:904] (2/8) Epoch 19, batch 4300, loss[loss=0.1949, simple_loss=0.2783, pruned_loss=0.05574, over 16290.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2749, pruned_loss=0.05008, over 3196136.53 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:29:37,738 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187025.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:56,651 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:10,226 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8838, 5.1353, 4.9138, 4.9134, 4.7028, 4.6125, 4.5586, 5.1888], device='cuda:2'), covar=tensor([0.0858, 0.0714, 0.0867, 0.0748, 0.0669, 0.0917, 0.1009, 0.0762], device='cuda:2'), in_proj_covar=tensor([0.0653, 0.0804, 0.0660, 0.0600, 0.0505, 0.0513, 0.0671, 0.0619], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:30:16,621 INFO [train.py:904] (2/8) Epoch 19, batch 4350, loss[loss=0.1863, simple_loss=0.2821, pruned_loss=0.0453, over 16554.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.278, pruned_loss=0.05117, over 3208543.42 frames. ], batch size: 68, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:30:30,633 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-30 21:30:32,930 INFO [optim.py:368] (2/8) 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 21:30:35,926 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:53,847 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:07,416 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187086.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:31:26,238 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187099.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:31,214 INFO [train.py:904] (2/8) Epoch 19, batch 4400, loss[loss=0.1888, simple_loss=0.2786, pruned_loss=0.04953, over 17190.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2793, pruned_loss=0.05176, over 3227814.32 frames. ], batch size: 44, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:31:42,562 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-04-30 21:32:05,655 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:13,521 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187131.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:13,656 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9933, 2.9736, 1.8915, 3.2743, 2.3095, 3.3407, 2.0583, 2.4587], device='cuda:2'), covar=tensor([0.0342, 0.0428, 0.1648, 0.0166, 0.0885, 0.0431, 0.1535, 0.0793], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0176, 0.0194, 0.0158, 0.0175, 0.0216, 0.0199, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:32:43,363 INFO [train.py:904] (2/8) Epoch 19, batch 4450, loss[loss=0.208, simple_loss=0.291, pruned_loss=0.06252, over 16339.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2828, pruned_loss=0.05314, over 3227708.36 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:32:47,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0982, 4.9429, 5.1628, 5.3033, 5.4342, 4.8109, 5.4356, 5.4797], device='cuda:2'), covar=tensor([0.1551, 0.1079, 0.1250, 0.0555, 0.0425, 0.0664, 0.0496, 0.0449], device='cuda:2'), in_proj_covar=tensor([0.0620, 0.0767, 0.0900, 0.0784, 0.0585, 0.0612, 0.0630, 0.0733], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:33:00,621 INFO [optim.py:368] (2/8) 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 21:33:15,024 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187173.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:24,885 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:35,455 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2349, 4.3410, 4.1215, 3.8606, 3.8710, 4.2531, 3.9034, 3.9900], device='cuda:2'), covar=tensor([0.0508, 0.0290, 0.0229, 0.0213, 0.0652, 0.0308, 0.0824, 0.0495], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0404, 0.0335, 0.0325, 0.0345, 0.0376, 0.0229, 0.0397], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:33:47,374 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187195.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:49,998 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187196.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:33:57,506 INFO [train.py:904] (2/8) Epoch 19, batch 4500, loss[loss=0.2022, simple_loss=0.2876, pruned_loss=0.05846, over 16580.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2835, pruned_loss=0.05399, over 3214658.46 frames. ], batch size: 62, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:34:24,641 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:27,135 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:54,212 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:55,420 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187243.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:08,499 INFO [train.py:904] (2/8) Epoch 19, batch 4550, loss[loss=0.2018, simple_loss=0.2886, pruned_loss=0.0575, over 16906.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2845, pruned_loss=0.05508, over 3227338.31 frames. ], batch size: 102, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:35:24,435 INFO [optim.py:368] (2/8) 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 21:35:30,230 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:36,697 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187271.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:43,265 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3295, 2.4738, 2.4495, 4.3628, 2.2560, 2.8270, 2.5057, 2.5717], device='cuda:2'), covar=tensor([0.1292, 0.3158, 0.2599, 0.0412, 0.4124, 0.2326, 0.3017, 0.3172], device='cuda:2'), in_proj_covar=tensor([0.0395, 0.0438, 0.0358, 0.0323, 0.0432, 0.0505, 0.0407, 0.0511], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:36:01,818 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-30 21:36:05,136 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:36:20,300 INFO [train.py:904] (2/8) Epoch 19, batch 4600, loss[loss=0.1981, simple_loss=0.2842, pruned_loss=0.05602, over 16921.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2856, pruned_loss=0.05537, over 3228350.28 frames. ], batch size: 109, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:33,043 INFO [train.py:904] (2/8) Epoch 19, batch 4650, loss[loss=0.1893, simple_loss=0.2707, pruned_loss=0.05394, over 16606.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2845, pruned_loss=0.05523, over 3224563.78 frames. ], batch size: 57, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:49,309 INFO [optim.py:368] (2/8) 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 21:37:52,239 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:01,609 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:14,877 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187381.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:38:33,294 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:43,262 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-30 21:38:44,846 INFO [train.py:904] (2/8) Epoch 19, batch 4700, loss[loss=0.1787, simple_loss=0.2751, pruned_loss=0.04117, over 15456.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2823, pruned_loss=0.05416, over 3207178.54 frames. ], batch size: 190, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:38:53,650 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3577, 5.3834, 5.1822, 4.4996, 5.3189, 1.8179, 5.0326, 4.8621], device='cuda:2'), covar=tensor([0.0100, 0.0097, 0.0139, 0.0416, 0.0095, 0.2835, 0.0112, 0.0245], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0146, 0.0191, 0.0175, 0.0168, 0.0201, 0.0183, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:38:59,543 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 21:39:01,715 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187413.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:29,992 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:34,155 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2711, 4.5435, 4.7114, 4.6812, 4.7009, 4.4394, 4.1212, 4.2884], device='cuda:2'), covar=tensor([0.0507, 0.0604, 0.0520, 0.0616, 0.0595, 0.0555, 0.1433, 0.0551], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0428, 0.0417, 0.0390, 0.0460, 0.0440, 0.0530, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 21:39:55,445 INFO [train.py:904] (2/8) Epoch 19, batch 4750, loss[loss=0.1731, simple_loss=0.258, pruned_loss=0.04404, over 16790.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2779, pruned_loss=0.05203, over 3200288.23 frames. ], batch size: 124, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:40:11,092 INFO [optim.py:368] (2/8) 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 21:40:57,686 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187496.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:05,659 INFO [train.py:904] (2/8) Epoch 19, batch 4800, loss[loss=0.1983, simple_loss=0.2913, pruned_loss=0.05262, over 16754.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.274, pruned_loss=0.04989, over 3209893.00 frames. ], batch size: 124, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:41:20,826 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:28,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9942, 2.7584, 2.8411, 2.0834, 2.6252, 2.1381, 2.6862, 2.9148], device='cuda:2'), covar=tensor([0.0280, 0.0722, 0.0513, 0.1718, 0.0820, 0.0899, 0.0617, 0.0747], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0150, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:41:30,313 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:37,742 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187523.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:40,230 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9860, 3.9741, 3.9611, 3.1435, 3.9160, 1.6862, 3.7266, 3.5322], device='cuda:2'), covar=tensor([0.0136, 0.0139, 0.0147, 0.0387, 0.0107, 0.2877, 0.0148, 0.0278], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0145, 0.0190, 0.0174, 0.0167, 0.0200, 0.0181, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:42:08,454 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187544.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:08,936 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 21:42:19,540 INFO [train.py:904] (2/8) Epoch 19, batch 4850, loss[loss=0.2408, simple_loss=0.312, pruned_loss=0.08477, over 11949.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2751, pruned_loss=0.04958, over 3181775.27 frames. ], batch size: 246, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:42:24,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9370, 3.8934, 4.4575, 2.0606, 4.7238, 4.6515, 3.3086, 3.1904], device='cuda:2'), covar=tensor([0.0769, 0.0231, 0.0127, 0.1212, 0.0036, 0.0093, 0.0371, 0.0504], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0107, 0.0096, 0.0137, 0.0077, 0.0122, 0.0126, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 21:42:36,205 INFO [optim.py:368] (2/8) 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 21:42:41,897 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:50,823 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:01,973 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:07,627 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:18,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8494, 3.2076, 3.2048, 2.0720, 2.8963, 3.1502, 3.0004, 1.8121], device='cuda:2'), covar=tensor([0.0549, 0.0059, 0.0053, 0.0400, 0.0109, 0.0109, 0.0107, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0078, 0.0079, 0.0131, 0.0094, 0.0104, 0.0091, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 21:43:32,172 INFO [train.py:904] (2/8) Epoch 19, batch 4900, loss[loss=0.1697, simple_loss=0.2601, pruned_loss=0.03961, over 17117.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.274, pruned_loss=0.04842, over 3160999.32 frames. ], batch size: 47, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:43:51,071 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:44:25,032 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6952, 3.8237, 2.3784, 4.3980, 2.8764, 4.2919, 2.4457, 2.9849], device='cuda:2'), covar=tensor([0.0267, 0.0321, 0.1544, 0.0122, 0.0853, 0.0456, 0.1492, 0.0749], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0173, 0.0190, 0.0154, 0.0172, 0.0211, 0.0196, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:44:43,108 INFO [train.py:904] (2/8) Epoch 19, batch 4950, loss[loss=0.1821, simple_loss=0.2813, pruned_loss=0.04145, over 16496.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.274, pruned_loss=0.04815, over 3168154.29 frames. ], batch size: 75, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:44:58,341 INFO [optim.py:368] (2/8) 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 21:45:24,181 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187681.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:45:39,671 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3506, 4.3726, 4.6404, 4.6177, 4.6360, 4.3291, 4.3140, 4.2272], device='cuda:2'), covar=tensor([0.0253, 0.0456, 0.0354, 0.0380, 0.0481, 0.0345, 0.0850, 0.0440], device='cuda:2'), in_proj_covar=tensor([0.0385, 0.0425, 0.0414, 0.0386, 0.0458, 0.0435, 0.0522, 0.0343], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 21:45:43,937 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:45:55,107 INFO [train.py:904] (2/8) Epoch 19, batch 5000, loss[loss=0.1986, simple_loss=0.2902, pruned_loss=0.05343, over 16840.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2754, pruned_loss=0.04765, over 3183543.11 frames. ], batch size: 116, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:46:32,203 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:33,446 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:52,594 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:47:07,084 INFO [train.py:904] (2/8) Epoch 19, batch 5050, loss[loss=0.1821, simple_loss=0.2726, pruned_loss=0.04581, over 16678.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2758, pruned_loss=0.04732, over 3202051.04 frames. ], batch size: 57, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:47:21,722 INFO [optim.py:368] (2/8) 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 21:47:34,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0313, 4.9703, 4.8812, 4.1055, 4.9000, 2.0515, 4.5808, 4.6672], device='cuda:2'), covar=tensor([0.0084, 0.0077, 0.0130, 0.0513, 0.0100, 0.2373, 0.0129, 0.0214], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0144, 0.0188, 0.0174, 0.0166, 0.0199, 0.0180, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:48:17,850 INFO [train.py:904] (2/8) Epoch 19, batch 5100, loss[loss=0.1862, simple_loss=0.2678, pruned_loss=0.05235, over 16766.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2738, pruned_loss=0.04647, over 3220336.26 frames. ], batch size: 134, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:06,687 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2657, 4.1512, 4.3252, 4.4724, 4.6272, 4.2147, 4.5621, 4.6382], device='cuda:2'), covar=tensor([0.1603, 0.1221, 0.1439, 0.0672, 0.0458, 0.1031, 0.0603, 0.0548], device='cuda:2'), in_proj_covar=tensor([0.0608, 0.0749, 0.0880, 0.0768, 0.0573, 0.0599, 0.0615, 0.0718], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:49:30,691 INFO [train.py:904] (2/8) Epoch 19, batch 5150, loss[loss=0.1886, simple_loss=0.2803, pruned_loss=0.04845, over 16922.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2742, pruned_loss=0.04584, over 3217180.43 frames. ], batch size: 109, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:47,461 INFO [optim.py:368] (2/8) 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 21:49:54,429 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187868.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:04,810 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187875.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:11,359 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187879.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:41,938 INFO [train.py:904] (2/8) Epoch 19, batch 5200, loss[loss=0.1696, simple_loss=0.2586, pruned_loss=0.04028, over 17125.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2721, pruned_loss=0.04525, over 3216854.45 frames. ], batch size: 47, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:51:12,272 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 21:51:44,545 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3045, 4.3557, 4.1788, 3.8969, 3.8287, 4.2540, 4.0263, 4.0292], device='cuda:2'), covar=tensor([0.0585, 0.0603, 0.0322, 0.0299, 0.0958, 0.0554, 0.0690, 0.0619], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0404, 0.0333, 0.0323, 0.0343, 0.0377, 0.0228, 0.0395], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:51:53,326 INFO [train.py:904] (2/8) Epoch 19, batch 5250, loss[loss=0.147, simple_loss=0.2459, pruned_loss=0.02405, over 16872.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2692, pruned_loss=0.04451, over 3230845.27 frames. ], batch size: 96, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:52:08,308 INFO [optim.py:368] (2/8) 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 21:53:08,598 INFO [train.py:904] (2/8) Epoch 19, batch 5300, loss[loss=0.1805, simple_loss=0.2614, pruned_loss=0.04978, over 12153.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2665, pruned_loss=0.04366, over 3220592.02 frames. ], batch size: 247, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:53:46,806 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:53:52,184 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7631, 4.6222, 4.8100, 5.0088, 5.2167, 4.6452, 5.1795, 5.2017], device='cuda:2'), covar=tensor([0.1798, 0.1296, 0.1818, 0.0764, 0.0485, 0.0814, 0.0515, 0.0584], device='cuda:2'), in_proj_covar=tensor([0.0618, 0.0762, 0.0897, 0.0782, 0.0583, 0.0610, 0.0625, 0.0730], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:54:04,398 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 21:54:21,271 INFO [train.py:904] (2/8) Epoch 19, batch 5350, loss[loss=0.162, simple_loss=0.2564, pruned_loss=0.03382, over 16973.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2654, pruned_loss=0.04342, over 3224765.86 frames. ], batch size: 55, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:54:24,802 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5387, 3.5018, 3.4715, 2.7799, 3.3789, 2.0628, 3.1703, 2.8129], device='cuda:2'), covar=tensor([0.0138, 0.0123, 0.0156, 0.0292, 0.0099, 0.2292, 0.0138, 0.0261], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0144, 0.0188, 0.0173, 0.0165, 0.0199, 0.0180, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:54:28,902 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4340, 3.4024, 2.5599, 2.1016, 2.3459, 2.1554, 3.4645, 3.1309], device='cuda:2'), covar=tensor([0.2875, 0.0588, 0.1899, 0.2651, 0.2472, 0.2082, 0.0534, 0.1098], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0263, 0.0297, 0.0302, 0.0290, 0.0248, 0.0286, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 21:54:38,014 INFO [optim.py:368] (2/8) 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 21:54:57,073 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:55:35,754 INFO [train.py:904] (2/8) Epoch 19, batch 5400, loss[loss=0.2107, simple_loss=0.292, pruned_loss=0.06476, over 12219.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2678, pruned_loss=0.04408, over 3210333.57 frames. ], batch size: 247, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:56:43,055 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 21:56:52,809 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7122, 2.4987, 2.3933, 3.6210, 2.5140, 3.7799, 1.4423, 2.7946], device='cuda:2'), covar=tensor([0.1292, 0.0754, 0.1209, 0.0167, 0.0221, 0.0425, 0.1674, 0.0773], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0182, 0.0203, 0.0213, 0.0196, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 21:56:54,047 INFO [train.py:904] (2/8) Epoch 19, batch 5450, loss[loss=0.2119, simple_loss=0.3013, pruned_loss=0.06122, over 16421.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2707, pruned_loss=0.04535, over 3204444.89 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:57:02,864 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1317, 4.0214, 4.1884, 4.3351, 4.4613, 4.0455, 4.3874, 4.4815], device='cuda:2'), covar=tensor([0.1799, 0.1219, 0.1546, 0.0725, 0.0555, 0.1317, 0.0783, 0.0645], device='cuda:2'), in_proj_covar=tensor([0.0617, 0.0759, 0.0894, 0.0780, 0.0581, 0.0609, 0.0624, 0.0728], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:57:11,919 INFO [optim.py:368] (2/8) 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 21:57:19,750 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188168.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:32,259 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188175.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:38,526 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:14,499 INFO [train.py:904] (2/8) Epoch 19, batch 5500, loss[loss=0.227, simple_loss=0.3093, pruned_loss=0.07234, over 16400.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2786, pruned_loss=0.0501, over 3184673.97 frames. ], batch size: 35, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:58:22,967 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3992, 3.3885, 3.4209, 3.5155, 3.5326, 3.3163, 3.5178, 3.5894], device='cuda:2'), covar=tensor([0.1244, 0.0881, 0.1090, 0.0610, 0.0670, 0.2213, 0.1027, 0.0753], device='cuda:2'), in_proj_covar=tensor([0.0617, 0.0759, 0.0895, 0.0779, 0.0582, 0.0610, 0.0623, 0.0729], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:58:34,642 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2708, 5.2504, 5.1463, 4.7575, 4.7550, 5.1461, 5.1366, 4.8919], device='cuda:2'), covar=tensor([0.0563, 0.0463, 0.0275, 0.0293, 0.1013, 0.0449, 0.0233, 0.0597], device='cuda:2'), in_proj_covar=tensor([0.0283, 0.0407, 0.0334, 0.0325, 0.0345, 0.0380, 0.0229, 0.0398], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-04-30 21:58:37,951 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:48,998 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:55,240 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:59:35,426 INFO [train.py:904] (2/8) Epoch 19, batch 5550, loss[loss=0.2072, simple_loss=0.2924, pruned_loss=0.06104, over 16581.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2857, pruned_loss=0.05555, over 3142026.51 frames. ], batch size: 57, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:59:53,552 INFO [optim.py:368] (2/8) 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 22:00:26,067 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:00:55,096 INFO [train.py:904] (2/8) Epoch 19, batch 5600, loss[loss=0.3335, simple_loss=0.3746, pruned_loss=0.1462, over 11129.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2911, pruned_loss=0.06023, over 3083874.76 frames. ], batch size: 247, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:01:32,940 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5945, 3.3996, 3.8705, 1.7245, 4.0714, 4.1162, 2.9962, 2.9132], device='cuda:2'), covar=tensor([0.0783, 0.0277, 0.0198, 0.1263, 0.0072, 0.0131, 0.0398, 0.0483], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0107, 0.0097, 0.0139, 0.0078, 0.0124, 0.0127, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:02:07,272 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:02:19,465 INFO [train.py:904] (2/8) Epoch 19, batch 5650, loss[loss=0.2404, simple_loss=0.3141, pruned_loss=0.08336, over 11558.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2956, pruned_loss=0.06412, over 3063762.00 frames. ], batch size: 247, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:02:36,991 INFO [optim.py:368] (2/8) 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 22:03:30,516 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-04-30 22:03:36,660 INFO [train.py:904] (2/8) Epoch 19, batch 5700, loss[loss=0.2555, simple_loss=0.313, pruned_loss=0.09894, over 11237.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2977, pruned_loss=0.06611, over 3052027.13 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:04:16,397 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 22:04:32,415 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 22:04:55,840 INFO [train.py:904] (2/8) Epoch 19, batch 5750, loss[loss=0.2157, simple_loss=0.2817, pruned_loss=0.07479, over 10905.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.2997, pruned_loss=0.06726, over 3035453.23 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:05:12,681 INFO [optim.py:368] (2/8) 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 22:05:45,287 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:06:00,145 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-30 22:06:03,568 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9297, 3.2543, 3.3066, 2.1487, 3.1262, 3.3514, 3.1991, 1.8776], device='cuda:2'), covar=tensor([0.0581, 0.0068, 0.0064, 0.0465, 0.0103, 0.0096, 0.0086, 0.0515], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0078, 0.0079, 0.0131, 0.0094, 0.0105, 0.0091, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:06:19,319 INFO [train.py:904] (2/8) Epoch 19, batch 5800, loss[loss=0.1994, simple_loss=0.2873, pruned_loss=0.05574, over 15381.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2998, pruned_loss=0.06699, over 3003230.91 frames. ], batch size: 191, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:08,182 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 22:07:24,060 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:07:38,950 INFO [train.py:904] (2/8) Epoch 19, batch 5850, loss[loss=0.2107, simple_loss=0.2995, pruned_loss=0.061, over 15419.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2972, pruned_loss=0.06489, over 3030874.74 frames. ], batch size: 190, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:54,837 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1997, 3.0504, 3.4193, 1.7277, 3.5151, 3.6040, 2.7597, 2.6647], device='cuda:2'), covar=tensor([0.0870, 0.0287, 0.0189, 0.1237, 0.0087, 0.0174, 0.0465, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0107, 0.0097, 0.0139, 0.0078, 0.0123, 0.0127, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:07:57,309 INFO [optim.py:368] (2/8) 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 22:08:59,601 INFO [train.py:904] (2/8) Epoch 19, batch 5900, loss[loss=0.2015, simple_loss=0.2841, pruned_loss=0.05951, over 16634.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2967, pruned_loss=0.06457, over 3033508.09 frames. ], batch size: 57, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:10:01,851 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:10:21,848 INFO [train.py:904] (2/8) Epoch 19, batch 5950, loss[loss=0.2162, simple_loss=0.3059, pruned_loss=0.06324, over 16640.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2976, pruned_loss=0.06322, over 3057004.47 frames. ], batch size: 76, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:10:40,535 INFO [optim.py:368] (2/8) 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 22:11:41,786 INFO [train.py:904] (2/8) Epoch 19, batch 6000, loss[loss=0.1943, simple_loss=0.2814, pruned_loss=0.05361, over 16433.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2962, pruned_loss=0.06234, over 3070376.95 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:11:41,786 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 22:11:52,551 INFO [train.py:938] (2/8) Epoch 19, validation: loss=0.1523, simple_loss=0.2653, pruned_loss=0.01968, over 944034.00 frames. 2023-04-30 22:11:52,552 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 22:13:13,860 INFO [train.py:904] (2/8) Epoch 19, batch 6050, loss[loss=0.1954, simple_loss=0.2863, pruned_loss=0.05222, over 16622.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2951, pruned_loss=0.06137, over 3093075.31 frames. ], batch size: 57, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:13:33,103 INFO [optim.py:368] (2/8) 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 22:14:32,939 INFO [train.py:904] (2/8) Epoch 19, batch 6100, loss[loss=0.2152, simple_loss=0.2976, pruned_loss=0.06645, over 16789.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2949, pruned_loss=0.06041, over 3098530.19 frames. ], batch size: 124, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:14:55,902 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0420, 2.3368, 2.5963, 1.9039, 2.6950, 2.8199, 2.4609, 2.4086], device='cuda:2'), covar=tensor([0.0716, 0.0267, 0.0228, 0.0993, 0.0114, 0.0266, 0.0445, 0.0461], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0107, 0.0096, 0.0139, 0.0078, 0.0123, 0.0127, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:14:57,464 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-30 22:15:21,075 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 22:15:33,782 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:15:57,515 INFO [train.py:904] (2/8) Epoch 19, batch 6150, loss[loss=0.2278, simple_loss=0.2999, pruned_loss=0.07782, over 11635.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2928, pruned_loss=0.05966, over 3096685.37 frames. ], batch size: 247, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:16:16,882 INFO [optim.py:368] (2/8) 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 22:17:16,331 INFO [train.py:904] (2/8) Epoch 19, batch 6200, loss[loss=0.1918, simple_loss=0.2724, pruned_loss=0.05557, over 17129.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2909, pruned_loss=0.05932, over 3095900.81 frames. ], batch size: 48, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:17:54,286 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4368, 2.9653, 3.0143, 1.9393, 2.6899, 2.1294, 3.0668, 3.2088], device='cuda:2'), covar=tensor([0.0290, 0.0774, 0.0628, 0.2002, 0.0895, 0.0982, 0.0670, 0.0857], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0144, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:17:58,828 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:01,793 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188931.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:14,982 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:34,122 INFO [train.py:904] (2/8) Epoch 19, batch 6250, loss[loss=0.1947, simple_loss=0.283, pruned_loss=0.05317, over 17009.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2901, pruned_loss=0.0591, over 3100549.27 frames. ], batch size: 55, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:18:52,889 INFO [optim.py:368] (2/8) 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 22:19:18,180 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1614, 3.8848, 3.9129, 4.2986, 4.4438, 4.0447, 4.3672, 4.4397], device='cuda:2'), covar=tensor([0.1627, 0.1525, 0.2328, 0.0952, 0.0865, 0.1736, 0.1179, 0.1019], device='cuda:2'), in_proj_covar=tensor([0.0611, 0.0751, 0.0881, 0.0769, 0.0576, 0.0604, 0.0619, 0.0722], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:19:28,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:32,259 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:36,042 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:51,018 INFO [train.py:904] (2/8) Epoch 19, batch 6300, loss[loss=0.1861, simple_loss=0.2747, pruned_loss=0.04878, over 16994.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2902, pruned_loss=0.05875, over 3101614.13 frames. ], batch size: 55, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:20:35,840 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189030.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:21:09,036 INFO [train.py:904] (2/8) Epoch 19, batch 6350, loss[loss=0.176, simple_loss=0.2708, pruned_loss=0.04065, over 16796.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2907, pruned_loss=0.05949, over 3098509.81 frames. ], batch size: 102, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:21:27,080 INFO [optim.py:368] (2/8) 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 22:21:28,351 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:21:51,832 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:03,999 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8985, 2.7968, 2.8822, 2.2010, 2.7306, 2.2222, 2.8272, 3.0014], device='cuda:2'), covar=tensor([0.0246, 0.0775, 0.0504, 0.1653, 0.0734, 0.1016, 0.0510, 0.0635], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0143, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:22:10,392 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:10,506 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7265, 3.8350, 2.8746, 2.3173, 2.6593, 2.4621, 4.1866, 3.5705], device='cuda:2'), covar=tensor([0.2660, 0.0660, 0.1777, 0.2529, 0.2516, 0.1952, 0.0425, 0.1147], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0265, 0.0299, 0.0305, 0.0293, 0.0250, 0.0289, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:22:26,883 INFO [train.py:904] (2/8) Epoch 19, batch 6400, loss[loss=0.178, simple_loss=0.2707, pruned_loss=0.0426, over 16730.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2908, pruned_loss=0.0606, over 3082412.05 frames. ], batch size: 76, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:23:02,679 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:21,428 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:25,616 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:43,274 INFO [train.py:904] (2/8) Epoch 19, batch 6450, loss[loss=0.1794, simple_loss=0.2759, pruned_loss=0.04142, over 16822.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2907, pruned_loss=0.05992, over 3070654.84 frames. ], batch size: 102, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:24:01,002 INFO [optim.py:368] (2/8) 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 22:24:10,967 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4890, 3.3011, 2.6692, 2.1324, 2.2263, 2.2539, 3.3657, 3.0989], device='cuda:2'), covar=tensor([0.2668, 0.0633, 0.1713, 0.2650, 0.2638, 0.2124, 0.0457, 0.1253], device='cuda:2'), in_proj_covar=tensor([0.0320, 0.0262, 0.0296, 0.0302, 0.0290, 0.0248, 0.0286, 0.0325], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:24:34,446 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:25:00,510 INFO [train.py:904] (2/8) Epoch 19, batch 6500, loss[loss=0.1954, simple_loss=0.2806, pruned_loss=0.0551, over 16915.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2884, pruned_loss=0.05882, over 3097371.17 frames. ], batch size: 116, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:25:43,746 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7588, 1.8504, 1.6037, 1.5117, 1.9681, 1.6011, 1.6821, 1.9434], device='cuda:2'), covar=tensor([0.0194, 0.0266, 0.0378, 0.0342, 0.0189, 0.0252, 0.0178, 0.0203], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0224, 0.0218, 0.0218, 0.0226, 0.0224, 0.0226, 0.0221], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:25:46,945 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7030, 3.9924, 2.9580, 2.2473, 2.7907, 2.4878, 4.2996, 3.5354], device='cuda:2'), covar=tensor([0.2921, 0.0660, 0.1781, 0.2802, 0.2397, 0.2039, 0.0459, 0.1249], device='cuda:2'), in_proj_covar=tensor([0.0321, 0.0263, 0.0297, 0.0303, 0.0291, 0.0249, 0.0287, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:26:14,550 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:26:19,680 INFO [train.py:904] (2/8) Epoch 19, batch 6550, loss[loss=0.2759, simple_loss=0.3248, pruned_loss=0.1135, over 11356.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2912, pruned_loss=0.05972, over 3097633.19 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:26:37,017 INFO [optim.py:368] (2/8) 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 22:27:03,287 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 22:27:09,602 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:12,559 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:34,958 INFO [train.py:904] (2/8) Epoch 19, batch 6600, loss[loss=0.2098, simple_loss=0.2933, pruned_loss=0.06313, over 16977.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2936, pruned_loss=0.06058, over 3083205.57 frames. ], batch size: 55, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:27:45,602 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:44,274 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:50,942 INFO [train.py:904] (2/8) Epoch 19, batch 6650, loss[loss=0.2688, simple_loss=0.3313, pruned_loss=0.1032, over 11285.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2938, pruned_loss=0.06097, over 3091813.98 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:29:08,220 INFO [optim.py:368] (2/8) 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 22:29:33,218 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 22:29:42,493 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189386.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:29:43,001 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 22:30:06,163 INFO [train.py:904] (2/8) Epoch 19, batch 6700, loss[loss=0.2104, simple_loss=0.2952, pruned_loss=0.0628, over 16944.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2919, pruned_loss=0.06058, over 3113996.61 frames. ], batch size: 109, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:30:15,364 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:33,827 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:56,600 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:31:21,911 INFO [train.py:904] (2/8) Epoch 19, batch 6750, loss[loss=0.2111, simple_loss=0.2915, pruned_loss=0.06538, over 16633.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2908, pruned_loss=0.06044, over 3119109.90 frames. ], batch size: 62, lr: 3.55e-03, grad_scale: 4.0 2023-04-30 22:31:42,236 INFO [optim.py:368] (2/8) 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 22:32:25,971 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9012, 4.1594, 3.9705, 4.0201, 3.7154, 3.7936, 3.8382, 4.1546], device='cuda:2'), covar=tensor([0.1167, 0.0850, 0.1030, 0.0790, 0.0772, 0.1514, 0.0910, 0.0926], device='cuda:2'), in_proj_covar=tensor([0.0643, 0.0784, 0.0648, 0.0586, 0.0492, 0.0504, 0.0652, 0.0605], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:32:38,317 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:32:39,012 INFO [train.py:904] (2/8) Epoch 19, batch 6800, loss[loss=0.2085, simple_loss=0.3007, pruned_loss=0.05816, over 16731.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2915, pruned_loss=0.06086, over 3117822.31 frames. ], batch size: 83, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:33:52,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7709, 3.8321, 3.9522, 3.7209, 3.8940, 4.2823, 3.9234, 3.6400], device='cuda:2'), covar=tensor([0.2309, 0.2182, 0.2428, 0.2550, 0.2606, 0.1663, 0.1722, 0.2667], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0571, 0.0631, 0.0478, 0.0635, 0.0661, 0.0493, 0.0641], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:33:52,628 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8175, 1.4176, 1.7360, 1.7312, 1.8577, 1.9304, 1.6194, 1.8452], device='cuda:2'), covar=tensor([0.0217, 0.0350, 0.0195, 0.0253, 0.0247, 0.0158, 0.0357, 0.0114], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0188, 0.0174, 0.0177, 0.0189, 0.0148, 0.0190, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:33:58,210 INFO [train.py:904] (2/8) Epoch 19, batch 6850, loss[loss=0.2235, simple_loss=0.3088, pruned_loss=0.06906, over 15537.00 frames. ], tot_loss[loss=0.207, simple_loss=0.292, pruned_loss=0.06101, over 3111521.38 frames. ], batch size: 191, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:34:13,594 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189562.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:17,024 INFO [optim.py:368] (2/8) 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 22:34:20,572 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8408, 3.1099, 3.0656, 1.9350, 2.9844, 3.1341, 3.0270, 1.6903], device='cuda:2'), covar=tensor([0.0600, 0.0094, 0.0098, 0.0511, 0.0129, 0.0153, 0.0125, 0.0613], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0079, 0.0079, 0.0133, 0.0095, 0.0106, 0.0091, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:34:47,197 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:50,390 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:35:03,297 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-04-30 22:35:12,995 INFO [train.py:904] (2/8) Epoch 19, batch 6900, loss[loss=0.2402, simple_loss=0.3101, pruned_loss=0.08516, over 11125.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2939, pruned_loss=0.05991, over 3128715.41 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:35:16,724 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:00,958 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189635.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:12,161 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189640.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:29,234 INFO [train.py:904] (2/8) Epoch 19, batch 6950, loss[loss=0.2113, simple_loss=0.2932, pruned_loss=0.0647, over 16729.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2951, pruned_loss=0.06116, over 3130703.17 frames. ], batch size: 134, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:36:48,878 INFO [optim.py:368] (2/8) 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 22:37:21,285 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:43,817 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:44,555 INFO [train.py:904] (2/8) Epoch 19, batch 7000, loss[loss=0.1964, simple_loss=0.2947, pruned_loss=0.04907, over 16442.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2956, pruned_loss=0.06104, over 3120709.14 frames. ], batch size: 146, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:37:47,748 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189703.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:49,134 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:04,476 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3445, 4.5028, 4.6244, 4.4279, 4.5202, 5.0306, 4.5041, 4.3058], device='cuda:2'), covar=tensor([0.1629, 0.1906, 0.2657, 0.2132, 0.2456, 0.1117, 0.1787, 0.2533], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0571, 0.0631, 0.0477, 0.0634, 0.0659, 0.0493, 0.0641], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:38:13,450 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:34,809 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:01,648 INFO [train.py:904] (2/8) Epoch 19, batch 7050, loss[loss=0.1917, simple_loss=0.281, pruned_loss=0.05119, over 16302.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2968, pruned_loss=0.06096, over 3118503.95 frames. ], batch size: 35, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:39:22,079 INFO [optim.py:368] (2/8) 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 22:39:22,611 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:27,395 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189768.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:50,863 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:00,708 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:19,979 INFO [train.py:904] (2/8) Epoch 19, batch 7100, loss[loss=0.2382, simple_loss=0.3065, pruned_loss=0.0849, over 11474.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2949, pruned_loss=0.06055, over 3119438.75 frames. ], batch size: 247, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:41:20,854 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8536, 4.6923, 4.8753, 5.0319, 5.2185, 4.6153, 5.2294, 5.2148], device='cuda:2'), covar=tensor([0.1759, 0.1220, 0.1491, 0.0688, 0.0490, 0.0998, 0.0513, 0.0621], device='cuda:2'), in_proj_covar=tensor([0.0606, 0.0745, 0.0877, 0.0767, 0.0575, 0.0604, 0.0621, 0.0718], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:41:23,650 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 22:41:38,323 INFO [train.py:904] (2/8) Epoch 19, batch 7150, loss[loss=0.2032, simple_loss=0.2885, pruned_loss=0.05898, over 15493.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2935, pruned_loss=0.06061, over 3102339.22 frames. ], batch size: 190, lr: 3.54e-03, grad_scale: 4.0 2023-04-30 22:41:45,519 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1505, 4.3378, 4.4772, 4.2126, 4.3839, 4.8294, 4.3367, 4.1004], device='cuda:2'), covar=tensor([0.1930, 0.1904, 0.2133, 0.2061, 0.2332, 0.1052, 0.1684, 0.2572], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0573, 0.0632, 0.0479, 0.0636, 0.0662, 0.0494, 0.0643], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:41:45,529 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:41:58,949 INFO [optim.py:368] (2/8) 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 22:42:27,216 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7932, 3.0820, 3.3686, 1.9949, 2.8360, 2.1826, 3.4170, 3.3368], device='cuda:2'), covar=tensor([0.0250, 0.0813, 0.0587, 0.1977, 0.0836, 0.0993, 0.0598, 0.0878], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0144, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:42:32,695 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9737, 2.0810, 2.1698, 3.4607, 2.1136, 2.3472, 2.1951, 2.2271], device='cuda:2'), covar=tensor([0.1312, 0.3487, 0.2852, 0.0611, 0.4004, 0.2464, 0.3456, 0.3307], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0430, 0.0354, 0.0317, 0.0429, 0.0495, 0.0403, 0.0503], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:42:51,647 INFO [train.py:904] (2/8) Epoch 19, batch 7200, loss[loss=0.1829, simple_loss=0.278, pruned_loss=0.04385, over 16606.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2914, pruned_loss=0.05897, over 3108224.90 frames. ], batch size: 35, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:42:55,719 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189904.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:44:12,493 INFO [train.py:904] (2/8) Epoch 19, batch 7250, loss[loss=0.2241, simple_loss=0.2976, pruned_loss=0.07526, over 11383.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2894, pruned_loss=0.05805, over 3107834.06 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:44:12,830 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189952.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:44:34,219 INFO [optim.py:368] (2/8) 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 22:45:19,909 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:45:32,080 INFO [train.py:904] (2/8) Epoch 19, batch 7300, loss[loss=0.2148, simple_loss=0.3089, pruned_loss=0.06036, over 16474.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2887, pruned_loss=0.05813, over 3105262.66 frames. ], batch size: 75, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:45:33,979 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190003.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:46:04,059 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5728, 5.8521, 5.5484, 5.6182, 5.2785, 5.1063, 5.3112, 5.9187], device='cuda:2'), covar=tensor([0.1061, 0.0724, 0.1109, 0.0819, 0.0721, 0.0812, 0.1120, 0.0849], device='cuda:2'), in_proj_covar=tensor([0.0634, 0.0774, 0.0639, 0.0577, 0.0485, 0.0498, 0.0645, 0.0596], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:46:48,533 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 19, batch 7350, loss[loss=0.2213, simple_loss=0.3067, pruned_loss=0.06796, over 16890.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2902, pruned_loss=0.05961, over 3077850.42 frames. ], batch size: 109, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:47:01,678 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:47:10,690 INFO [optim.py:368] (2/8) 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 22:47:24,614 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-04-30 22:47:29,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4094, 3.4617, 2.1094, 3.8901, 2.5753, 3.8965, 2.1502, 2.7131], device='cuda:2'), covar=tensor([0.0272, 0.0401, 0.1739, 0.0187, 0.0897, 0.0560, 0.1545, 0.0807], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0155, 0.0174, 0.0213, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:47:40,480 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:47:57,274 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 22:48:08,858 INFO [train.py:904] (2/8) Epoch 19, batch 7400, loss[loss=0.2519, simple_loss=0.3154, pruned_loss=0.09414, over 11347.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2912, pruned_loss=0.06059, over 3057536.76 frames. ], batch size: 247, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:09,841 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190139.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 22:49:30,223 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:49:30,831 INFO [train.py:904] (2/8) Epoch 19, batch 7450, loss[loss=0.2152, simple_loss=0.3022, pruned_loss=0.0641, over 16941.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2924, pruned_loss=0.06136, over 3075975.36 frames. ], batch size: 109, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:41,102 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 22:50:05,069 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-30 22:50:09,217 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4010, 2.9405, 2.6079, 2.2623, 2.2119, 2.2687, 2.9259, 2.7953], device='cuda:2'), covar=tensor([0.2464, 0.0825, 0.1666, 0.2526, 0.2360, 0.2063, 0.0538, 0.1381], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0264, 0.0299, 0.0305, 0.0293, 0.0250, 0.0287, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 22:50:11,768 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 22:50:51,707 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190200.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 22:50:54,217 INFO [train.py:904] (2/8) Epoch 19, batch 7500, loss[loss=0.216, simple_loss=0.2965, pruned_loss=0.0678, over 16919.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2922, pruned_loss=0.06014, over 3085433.54 frames. ], batch size: 109, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:50:58,477 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3380, 2.3566, 2.3847, 4.1968, 2.2573, 2.7926, 2.3954, 2.5719], device='cuda:2'), covar=tensor([0.1249, 0.3488, 0.2820, 0.0445, 0.3920, 0.2322, 0.3420, 0.2938], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0433, 0.0356, 0.0319, 0.0431, 0.0497, 0.0404, 0.0505], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:50:59,305 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:51:10,223 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:52:13,611 INFO [train.py:904] (2/8) Epoch 19, batch 7550, loss[loss=0.1959, simple_loss=0.2775, pruned_loss=0.05718, over 16537.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2917, pruned_loss=0.06056, over 3078224.70 frames. ], batch size: 75, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:52:34,940 INFO [optim.py:368] (2/8) 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 22:53:22,455 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190296.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:53:31,301 INFO [train.py:904] (2/8) Epoch 19, batch 7600, loss[loss=0.2094, simple_loss=0.2955, pruned_loss=0.06167, over 16721.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.291, pruned_loss=0.06106, over 3072893.86 frames. ], batch size: 89, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:53:52,915 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190316.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:36,982 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:48,951 INFO [train.py:904] (2/8) Epoch 19, batch 7650, loss[loss=0.2202, simple_loss=0.3134, pruned_loss=0.06344, over 17143.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2913, pruned_loss=0.06069, over 3099690.66 frames. ], batch size: 49, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:55:02,242 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:10,292 INFO [optim.py:368] (2/8) 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 22:55:27,529 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190377.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:38,685 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:02,489 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:02,555 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6407, 3.0343, 3.3068, 1.9144, 2.8404, 2.1530, 3.1792, 3.3089], device='cuda:2'), covar=tensor([0.0273, 0.0837, 0.0566, 0.2125, 0.0830, 0.1009, 0.0663, 0.0978], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0151, 0.0144, 0.0128, 0.0143, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:56:06,255 INFO [train.py:904] (2/8) Epoch 19, batch 7700, loss[loss=0.2049, simple_loss=0.2922, pruned_loss=0.05883, over 16585.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2913, pruned_loss=0.06128, over 3099234.80 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:56:11,114 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5818, 3.6386, 4.0306, 2.0106, 4.2038, 4.2212, 3.0892, 3.0523], device='cuda:2'), covar=tensor([0.0769, 0.0243, 0.0184, 0.1168, 0.0071, 0.0147, 0.0387, 0.0458], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0106, 0.0096, 0.0138, 0.0078, 0.0121, 0.0126, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 22:56:12,135 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7601, 3.8602, 4.1205, 4.0706, 4.1029, 3.8611, 3.8774, 3.9032], device='cuda:2'), covar=tensor([0.0383, 0.0660, 0.0427, 0.0485, 0.0502, 0.0497, 0.0935, 0.0549], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0428, 0.0417, 0.0390, 0.0464, 0.0437, 0.0530, 0.0351], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 22:56:15,128 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:26,135 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 22:56:50,414 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:01,421 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3504, 5.7163, 5.4633, 5.4905, 5.1478, 5.1290, 5.1221, 5.8449], device='cuda:2'), covar=tensor([0.1305, 0.0807, 0.1041, 0.0885, 0.0883, 0.0747, 0.1168, 0.0884], device='cuda:2'), in_proj_covar=tensor([0.0637, 0.0772, 0.0642, 0.0579, 0.0485, 0.0498, 0.0647, 0.0595], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:57:18,843 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0925, 5.1040, 4.8828, 4.2432, 4.9852, 1.9616, 4.7250, 4.6299], device='cuda:2'), covar=tensor([0.0078, 0.0057, 0.0181, 0.0353, 0.0078, 0.2518, 0.0131, 0.0200], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0143, 0.0189, 0.0172, 0.0164, 0.0198, 0.0178, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 22:57:21,407 INFO [train.py:904] (2/8) Epoch 19, batch 7750, loss[loss=0.23, simple_loss=0.2972, pruned_loss=0.08139, over 11573.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2905, pruned_loss=0.06033, over 3117991.61 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:57:35,047 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:43,808 INFO [optim.py:368] (2/8) 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 22:58:30,599 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190495.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 22:58:39,266 INFO [train.py:904] (2/8) Epoch 19, batch 7800, loss[loss=0.2174, simple_loss=0.301, pruned_loss=0.06692, over 16926.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2919, pruned_loss=0.06149, over 3094973.75 frames. ], batch size: 109, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:58:47,777 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:59:57,134 INFO [train.py:904] (2/8) Epoch 19, batch 7850, loss[loss=0.1988, simple_loss=0.2794, pruned_loss=0.05906, over 16651.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2925, pruned_loss=0.06149, over 3082609.83 frames. ], batch size: 57, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:00:17,836 INFO [optim.py:368] (2/8) 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 23:00:56,948 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190591.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:01:12,371 INFO [train.py:904] (2/8) Epoch 19, batch 7900, loss[loss=0.2569, simple_loss=0.3209, pruned_loss=0.09646, over 11425.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2918, pruned_loss=0.06088, over 3079429.56 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:01:23,317 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190609.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:02:32,482 INFO [train.py:904] (2/8) Epoch 19, batch 7950, loss[loss=0.1957, simple_loss=0.2717, pruned_loss=0.0599, over 17053.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2924, pruned_loss=0.06153, over 3071359.07 frames. ], batch size: 53, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:02:32,969 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190652.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:02:54,091 INFO [optim.py:368] (2/8) 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 23:03:00,815 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190670.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:03:04,351 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190672.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:08,598 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:49,817 INFO [train.py:904] (2/8) Epoch 19, batch 8000, loss[loss=0.2153, simple_loss=0.2964, pruned_loss=0.06711, over 16278.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2926, pruned_loss=0.06177, over 3070192.41 frames. ], batch size: 165, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:04:28,399 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 23:04:44,490 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190736.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:07,642 INFO [train.py:904] (2/8) Epoch 19, batch 8050, loss[loss=0.1933, simple_loss=0.2796, pruned_loss=0.05345, over 16781.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2925, pruned_loss=0.06183, over 3072271.62 frames. ], batch size: 76, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:05:12,146 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190755.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:28,125 INFO [optim.py:368] (2/8) 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 23:06:12,395 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190795.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:06:22,312 INFO [train.py:904] (2/8) Epoch 19, batch 8100, loss[loss=0.1849, simple_loss=0.2852, pruned_loss=0.04229, over 16845.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2916, pruned_loss=0.06048, over 3106166.97 frames. ], batch size: 96, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:06:23,044 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-30 23:06:31,293 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:07:25,606 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190843.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:07:39,350 INFO [train.py:904] (2/8) Epoch 19, batch 8150, loss[loss=0.1894, simple_loss=0.2734, pruned_loss=0.05267, over 16888.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2898, pruned_loss=0.05996, over 3097872.44 frames. ], batch size: 116, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:07:44,592 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:08:01,318 INFO [optim.py:368] (2/8) 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 23:08:40,592 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 23:08:57,387 INFO [train.py:904] (2/8) Epoch 19, batch 8200, loss[loss=0.1942, simple_loss=0.2731, pruned_loss=0.05764, over 16853.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2876, pruned_loss=0.0596, over 3090794.79 frames. ], batch size: 42, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:11,508 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190947.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:10:19,456 INFO [train.py:904] (2/8) Epoch 19, batch 8250, loss[loss=0.1924, simple_loss=0.2736, pruned_loss=0.05565, over 12097.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2862, pruned_loss=0.05683, over 3082172.55 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:30,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4124, 3.3437, 3.4766, 3.5593, 3.5943, 3.3169, 3.5359, 3.6516], device='cuda:2'), covar=tensor([0.1407, 0.1032, 0.1125, 0.0658, 0.0687, 0.2259, 0.0921, 0.0798], device='cuda:2'), in_proj_covar=tensor([0.0599, 0.0744, 0.0869, 0.0760, 0.0575, 0.0597, 0.0619, 0.0715], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:10:39,904 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190965.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:10:40,157 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9142, 2.0894, 2.3117, 3.1830, 2.1516, 2.2777, 2.2549, 2.1920], device='cuda:2'), covar=tensor([0.1230, 0.3693, 0.2706, 0.0693, 0.4550, 0.2647, 0.3551, 0.3606], device='cuda:2'), in_proj_covar=tensor([0.0385, 0.0429, 0.0352, 0.0316, 0.0427, 0.0492, 0.0400, 0.0500], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:10:41,155 INFO [optim.py:368] (2/8) 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 23:10:48,420 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 23:10:51,831 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:11:29,769 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0205, 4.1533, 4.2627, 3.1081, 3.6663, 4.1837, 3.7540, 2.8607], device='cuda:2'), covar=tensor([0.0386, 0.0048, 0.0037, 0.0308, 0.0102, 0.0090, 0.0075, 0.0355], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0079, 0.0080, 0.0133, 0.0093, 0.0107, 0.0091, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 23:11:39,003 INFO [train.py:904] (2/8) Epoch 19, batch 8300, loss[loss=0.183, simple_loss=0.2683, pruned_loss=0.04884, over 12353.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2836, pruned_loss=0.05432, over 3051874.44 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:12:10,266 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191020.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:12:27,701 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191031.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:12:40,293 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-04-30 23:13:00,823 INFO [train.py:904] (2/8) Epoch 19, batch 8350, loss[loss=0.1905, simple_loss=0.2873, pruned_loss=0.04687, over 15396.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2833, pruned_loss=0.05233, over 3060316.36 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:13:08,054 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191055.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:24,342 INFO [optim.py:368] (2/8) 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 23:13:28,639 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191068.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:51,935 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0656, 4.0736, 4.4357, 4.4285, 4.4095, 4.1787, 4.1461, 4.1924], device='cuda:2'), covar=tensor([0.0402, 0.0849, 0.0486, 0.0473, 0.0513, 0.0485, 0.1124, 0.0542], device='cuda:2'), in_proj_covar=tensor([0.0393, 0.0431, 0.0418, 0.0394, 0.0465, 0.0439, 0.0535, 0.0355], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 23:14:22,694 INFO [train.py:904] (2/8) Epoch 19, batch 8400, loss[loss=0.1768, simple_loss=0.2564, pruned_loss=0.04864, over 12197.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2805, pruned_loss=0.05002, over 3055355.52 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:14:24,465 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:06,908 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:11,012 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191132.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:42,689 INFO [train.py:904] (2/8) Epoch 19, batch 8450, loss[loss=0.1785, simple_loss=0.2628, pruned_loss=0.04712, over 12459.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2786, pruned_loss=0.04857, over 3051026.59 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:16:06,334 INFO [optim.py:368] (2/8) 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 23:16:36,998 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:16:50,100 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:17:04,704 INFO [train.py:904] (2/8) Epoch 19, batch 8500, loss[loss=0.1486, simple_loss=0.2477, pruned_loss=0.02473, over 16793.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.275, pruned_loss=0.04602, over 3061164.34 frames. ], batch size: 83, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:22,325 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191246.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:18:24,092 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191247.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:18:32,469 INFO [train.py:904] (2/8) Epoch 19, batch 8550, loss[loss=0.1572, simple_loss=0.2442, pruned_loss=0.03511, over 11825.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.272, pruned_loss=0.04465, over 3054982.50 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:57,579 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191265.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:19:00,615 INFO [optim.py:368] (2/8) 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 23:20:00,161 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191295.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:20:05,718 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7546, 2.8012, 2.6908, 4.5612, 3.0871, 4.1896, 1.6504, 3.1409], device='cuda:2'), covar=tensor([0.1303, 0.0756, 0.1028, 0.0135, 0.0127, 0.0329, 0.1522, 0.0662], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0179, 0.0201, 0.0209, 0.0193, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-04-30 23:20:13,437 INFO [train.py:904] (2/8) Epoch 19, batch 8600, loss[loss=0.1628, simple_loss=0.2651, pruned_loss=0.0302, over 16864.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2721, pruned_loss=0.04365, over 3048337.88 frames. ], batch size: 90, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:20:37,121 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:03,000 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4557, 3.3190, 2.6895, 2.0778, 2.1399, 2.2061, 3.4487, 2.9870], device='cuda:2'), covar=tensor([0.2882, 0.0747, 0.1822, 0.3140, 0.2992, 0.2320, 0.0451, 0.1547], device='cuda:2'), in_proj_covar=tensor([0.0318, 0.0259, 0.0294, 0.0299, 0.0286, 0.0246, 0.0282, 0.0321], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-04-30 23:21:11,776 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:52,926 INFO [train.py:904] (2/8) Epoch 19, batch 8650, loss[loss=0.1701, simple_loss=0.2686, pruned_loss=0.03582, over 16643.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2704, pruned_loss=0.04212, over 3060582.92 frames. ], batch size: 134, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:22:25,675 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:22:29,946 INFO [optim.py:368] (2/8) 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 23:22:55,705 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:23:39,021 INFO [train.py:904] (2/8) Epoch 19, batch 8700, loss[loss=0.1921, simple_loss=0.2925, pruned_loss=0.04586, over 16719.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2685, pruned_loss=0.04077, over 3075525.02 frames. ], batch size: 134, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:23:40,196 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0352, 3.3366, 3.3006, 2.1320, 3.0258, 3.3715, 3.2418, 1.8098], device='cuda:2'), covar=tensor([0.0608, 0.0063, 0.0065, 0.0477, 0.0123, 0.0092, 0.0088, 0.0573], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0079, 0.0080, 0.0134, 0.0094, 0.0106, 0.0091, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-04-30 23:24:20,674 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:24:23,580 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:24:28,806 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 23:24:56,917 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5640, 3.6395, 3.4271, 3.1816, 3.1919, 3.5715, 3.3206, 3.3906], device='cuda:2'), covar=tensor([0.0635, 0.0601, 0.0320, 0.0302, 0.0673, 0.0513, 0.1571, 0.0549], device='cuda:2'), in_proj_covar=tensor([0.0271, 0.0393, 0.0319, 0.0310, 0.0327, 0.0362, 0.0221, 0.0380], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:25:04,845 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-30 23:25:14,445 INFO [train.py:904] (2/8) Epoch 19, batch 8750, loss[loss=0.1812, simple_loss=0.2826, pruned_loss=0.03987, over 16267.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2681, pruned_loss=0.04029, over 3088380.69 frames. ], batch size: 165, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:25:58,340 INFO [optim.py:368] (2/8) 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 23:26:39,672 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:27:07,632 INFO [train.py:904] (2/8) Epoch 19, batch 8800, loss[loss=0.1731, simple_loss=0.2592, pruned_loss=0.0435, over 12373.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2665, pruned_loss=0.03955, over 3073775.07 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:27:30,558 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 23:28:28,986 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191541.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:28:51,736 INFO [train.py:904] (2/8) Epoch 19, batch 8850, loss[loss=0.1597, simple_loss=0.2674, pruned_loss=0.02597, over 15465.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2688, pruned_loss=0.03912, over 3054708.57 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:29:28,628 INFO [optim.py:368] (2/8) 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 23:30:26,410 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 23:30:39,776 INFO [train.py:904] (2/8) Epoch 19, batch 8900, loss[loss=0.1994, simple_loss=0.2889, pruned_loss=0.05497, over 16803.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2693, pruned_loss=0.03859, over 3065704.34 frames. ], batch size: 124, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:32:42,039 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 23:32:45,267 INFO [train.py:904] (2/8) Epoch 19, batch 8950, loss[loss=0.16, simple_loss=0.2585, pruned_loss=0.0307, over 16498.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2691, pruned_loss=0.03918, over 3067562.80 frames. ], batch size: 68, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:32:50,177 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7633, 1.3262, 1.6753, 1.6541, 1.8241, 1.8264, 1.6402, 1.7500], device='cuda:2'), covar=tensor([0.0231, 0.0401, 0.0224, 0.0290, 0.0307, 0.0178, 0.0420, 0.0134], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0185, 0.0170, 0.0173, 0.0185, 0.0143, 0.0188, 0.0138], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:33:21,158 INFO [optim.py:368] (2/8) 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 23:34:34,392 INFO [train.py:904] (2/8) Epoch 19, batch 9000, loss[loss=0.1655, simple_loss=0.252, pruned_loss=0.03951, over 16848.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2655, pruned_loss=0.03787, over 3057681.47 frames. ], batch size: 124, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:34:34,393 INFO [train.py:929] (2/8) Computing validation loss 2023-04-30 23:34:42,117 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7903, 5.7272, 5.6798, 5.3183, 5.6898, 3.5194, 5.5992, 5.6608], device='cuda:2'), covar=tensor([0.0058, 0.0062, 0.0144, 0.0253, 0.0079, 0.1519, 0.0115, 0.0112], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0140, 0.0183, 0.0165, 0.0161, 0.0195, 0.0174, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:34:44,202 INFO [train.py:938] (2/8) Epoch 19, validation: loss=0.1462, simple_loss=0.2506, pruned_loss=0.02087, over 944034.00 frames. 2023-04-30 23:34:44,203 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-04-30 23:35:25,484 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:35:31,021 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:36:06,370 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-30 23:36:27,797 INFO [train.py:904] (2/8) Epoch 19, batch 9050, loss[loss=0.1652, simple_loss=0.2654, pruned_loss=0.03256, over 16800.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2663, pruned_loss=0.03826, over 3050308.23 frames. ], batch size: 102, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:37:04,256 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:37:39,966 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 19, batch 9100, loss[loss=0.1432, simple_loss=0.2505, pruned_loss=0.01799, over 16915.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.266, pruned_loss=0.0385, over 3061277.41 frames. ], batch size: 96, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:38:13,639 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2733, 5.2363, 5.0866, 4.6247, 4.7541, 5.1501, 5.0760, 4.7648], device='cuda:2'), covar=tensor([0.0597, 0.0632, 0.0309, 0.0327, 0.0935, 0.0614, 0.0298, 0.0742], device='cuda:2'), in_proj_covar=tensor([0.0269, 0.0389, 0.0316, 0.0307, 0.0323, 0.0358, 0.0219, 0.0376], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:39:32,422 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191836.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:39:43,531 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:39:53,152 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4669, 2.4977, 2.1525, 2.3188, 2.8697, 2.5627, 2.8814, 3.0751], device='cuda:2'), covar=tensor([0.0102, 0.0433, 0.0537, 0.0471, 0.0297, 0.0388, 0.0247, 0.0247], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0224, 0.0215, 0.0216, 0.0225, 0.0222, 0.0222, 0.0216], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:40:08,190 INFO [train.py:904] (2/8) Epoch 19, batch 9150, loss[loss=0.1736, simple_loss=0.2612, pruned_loss=0.04296, over 12435.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2667, pruned_loss=0.03842, over 3051581.30 frames. ], batch size: 250, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:40:46,340 INFO [optim.py:368] (2/8) 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 23:41:01,122 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 23:41:30,261 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:41:52,732 INFO [train.py:904] (2/8) Epoch 19, batch 9200, loss[loss=0.1569, simple_loss=0.245, pruned_loss=0.03442, over 16385.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2621, pruned_loss=0.03753, over 3047331.30 frames. ], batch size: 35, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:42:12,811 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-30 23:43:29,321 INFO [train.py:904] (2/8) Epoch 19, batch 9250, loss[loss=0.147, simple_loss=0.2475, pruned_loss=0.02331, over 16838.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2617, pruned_loss=0.03736, over 3035257.71 frames. ], batch size: 102, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:44:05,909 INFO [optim.py:368] (2/8) 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 23:45:23,368 INFO [train.py:904] (2/8) Epoch 19, batch 9300, loss[loss=0.1481, simple_loss=0.2467, pruned_loss=0.02482, over 16892.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2607, pruned_loss=0.03726, over 3051968.79 frames. ], batch size: 96, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:46:09,690 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:09,784 INFO [train.py:904] (2/8) Epoch 19, batch 9350, loss[loss=0.1743, simple_loss=0.2636, pruned_loss=0.04251, over 16991.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2605, pruned_loss=0.03694, over 3067302.13 frames. ], batch size: 109, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:47:46,335 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:47,130 INFO [optim.py:368] (2/8) 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 23:48:49,308 INFO [train.py:904] (2/8) Epoch 19, batch 9400, loss[loss=0.1614, simple_loss=0.2627, pruned_loss=0.03006, over 16339.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2611, pruned_loss=0.03683, over 3074851.61 frames. ], batch size: 146, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:48:50,354 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:29,959 INFO [train.py:904] (2/8) Epoch 19, batch 9450, loss[loss=0.1528, simple_loss=0.2495, pruned_loss=0.02802, over 16615.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2625, pruned_loss=0.03699, over 3067881.97 frames. ], batch size: 68, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:50:35,964 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:37,965 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 23:50:51,015 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192163.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:51:05,238 INFO [optim.py:368] (2/8) 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 23:51:47,877 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1314, 3.9744, 4.2516, 4.3383, 4.4574, 3.9982, 4.4397, 4.4769], device='cuda:2'), covar=tensor([0.1754, 0.1268, 0.1396, 0.0726, 0.0527, 0.1402, 0.0650, 0.0754], device='cuda:2'), in_proj_covar=tensor([0.0586, 0.0724, 0.0845, 0.0743, 0.0557, 0.0582, 0.0599, 0.0697], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:52:10,464 INFO [train.py:904] (2/8) Epoch 19, batch 9500, loss[loss=0.1763, simple_loss=0.2659, pruned_loss=0.04332, over 16565.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2616, pruned_loss=0.03686, over 3065796.31 frames. ], batch size: 62, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:52:14,148 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6445, 2.1180, 1.7519, 1.8858, 2.3661, 2.0988, 2.1188, 2.5138], device='cuda:2'), covar=tensor([0.0178, 0.0410, 0.0542, 0.0489, 0.0269, 0.0364, 0.0193, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0223, 0.0214, 0.0216, 0.0224, 0.0221, 0.0220, 0.0214], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:52:39,613 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:45,044 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:49,328 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:59,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0911, 4.0829, 4.4381, 4.4150, 4.4101, 4.1726, 4.1729, 4.1785], device='cuda:2'), covar=tensor([0.0357, 0.1055, 0.0547, 0.0461, 0.0539, 0.0492, 0.0903, 0.0404], device='cuda:2'), in_proj_covar=tensor([0.0377, 0.0409, 0.0403, 0.0377, 0.0445, 0.0420, 0.0508, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-04-30 23:53:55,148 INFO [train.py:904] (2/8) Epoch 19, batch 9550, loss[loss=0.1574, simple_loss=0.2575, pruned_loss=0.02862, over 16834.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2611, pruned_loss=0.03696, over 3064141.66 frames. ], batch size: 96, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:54:34,513 INFO [optim.py:368] (2/8) 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 23:54:35,241 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4489, 3.4399, 3.6235, 1.8593, 3.7633, 3.8795, 2.9750, 3.0502], device='cuda:2'), covar=tensor([0.0725, 0.0223, 0.0194, 0.1216, 0.0082, 0.0130, 0.0435, 0.0379], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0101, 0.0089, 0.0133, 0.0074, 0.0114, 0.0120, 0.0124], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-04-30 23:54:55,496 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:00,663 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:25,589 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 23:55:31,111 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8779, 4.8468, 4.6824, 4.2447, 4.7243, 1.7064, 4.5118, 4.5804], device='cuda:2'), covar=tensor([0.0090, 0.0104, 0.0216, 0.0331, 0.0121, 0.2643, 0.0151, 0.0208], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0138, 0.0181, 0.0162, 0.0159, 0.0194, 0.0171, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-04-30 23:55:38,440 INFO [train.py:904] (2/8) Epoch 19, batch 9600, loss[loss=0.1968, simple_loss=0.2948, pruned_loss=0.0494, over 16223.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2625, pruned_loss=0.03776, over 3060871.38 frames. ], batch size: 165, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:55:43,574 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 23:57:09,999 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5854, 2.9992, 3.2823, 1.8939, 2.8638, 2.0328, 3.1383, 3.1723], device='cuda:2'), covar=tensor([0.0289, 0.0821, 0.0508, 0.2165, 0.0765, 0.1078, 0.0656, 0.0891], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0146, 0.0139, 0.0124, 0.0138, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-04-30 23:57:27,639 INFO [train.py:904] (2/8) Epoch 19, batch 9650, loss[loss=0.1754, simple_loss=0.2715, pruned_loss=0.03964, over 16914.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2643, pruned_loss=0.03801, over 3059144.59 frames. ], batch size: 116, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:58:09,763 INFO [optim.py:368] (2/8) 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 23:58:53,957 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:09,402 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:15,171 INFO [train.py:904] (2/8) Epoch 19, batch 9700, loss[loss=0.167, simple_loss=0.26, pruned_loss=0.03703, over 15440.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2633, pruned_loss=0.03798, over 3039779.21 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:19,357 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192433.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:00:57,138 INFO [train.py:904] (2/8) Epoch 19, batch 9750, loss[loss=0.1749, simple_loss=0.2699, pruned_loss=0.03999, over 16336.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.263, pruned_loss=0.03856, over 3048124.50 frames. ], batch size: 146, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:58,978 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:09,805 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:14,107 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:31,653 INFO [optim.py:368] (2/8) 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 00:02:23,611 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:02:37,067 INFO [train.py:904] (2/8) Epoch 19, batch 9800, loss[loss=0.1642, simple_loss=0.2735, pruned_loss=0.02744, over 15490.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2628, pruned_loss=0.03726, over 3061585.22 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:02:55,411 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192511.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:04:23,520 INFO [train.py:904] (2/8) Epoch 19, batch 9850, loss[loss=0.163, simple_loss=0.2575, pruned_loss=0.03428, over 16254.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2636, pruned_loss=0.03683, over 3061785.32 frames. ], batch size: 165, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:05:00,568 INFO [optim.py:368] (2/8) 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 00:05:11,431 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192575.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:05:11,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5893, 3.5923, 3.5596, 2.8684, 3.5049, 1.9177, 3.3003, 2.9844], device='cuda:2'), covar=tensor([0.0116, 0.0101, 0.0163, 0.0181, 0.0092, 0.2477, 0.0110, 0.0228], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0139, 0.0182, 0.0162, 0.0159, 0.0195, 0.0172, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:05:15,481 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:06:01,031 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7147, 2.6186, 2.4893, 4.5305, 3.0042, 4.0535, 1.6499, 2.9307], device='cuda:2'), covar=tensor([0.1377, 0.0833, 0.1247, 0.0158, 0.0133, 0.0391, 0.1517, 0.0776], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0167, 0.0188, 0.0174, 0.0194, 0.0206, 0.0192, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:06:14,630 INFO [train.py:904] (2/8) Epoch 19, batch 9900, loss[loss=0.1787, simple_loss=0.2711, pruned_loss=0.04311, over 12613.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2634, pruned_loss=0.03671, over 3038518.44 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:06:46,512 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:08:13,281 INFO [train.py:904] (2/8) Epoch 19, batch 9950, loss[loss=0.155, simple_loss=0.2555, pruned_loss=0.02723, over 16714.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2651, pruned_loss=0.03676, over 3039187.94 frames. ], batch size: 76, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:08:54,785 INFO [optim.py:368] (2/8) 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 00:09:04,359 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:14,946 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:26,376 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0347, 4.0708, 3.9362, 3.5916, 3.6275, 4.0079, 3.7273, 3.7512], device='cuda:2'), covar=tensor([0.0630, 0.0916, 0.0399, 0.0321, 0.0776, 0.0593, 0.0854, 0.0611], device='cuda:2'), in_proj_covar=tensor([0.0268, 0.0382, 0.0312, 0.0302, 0.0318, 0.0355, 0.0215, 0.0370], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-05-01 00:09:28,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7749, 5.0327, 5.1587, 5.0030, 5.0670, 5.5561, 5.0602, 4.8072], device='cuda:2'), covar=tensor([0.0988, 0.1813, 0.2048, 0.1826, 0.2439, 0.0954, 0.1571, 0.2206], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0541, 0.0599, 0.0450, 0.0601, 0.0629, 0.0471, 0.0601], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 00:09:53,486 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 00:10:14,415 INFO [train.py:904] (2/8) Epoch 19, batch 10000, loss[loss=0.1752, simple_loss=0.2617, pruned_loss=0.04432, over 12247.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2634, pruned_loss=0.03599, over 3073026.61 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:11:22,436 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192734.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:25,782 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192735.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:50,828 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:56,958 INFO [train.py:904] (2/8) Epoch 19, batch 10050, loss[loss=0.1824, simple_loss=0.2732, pruned_loss=0.04585, over 16990.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2639, pruned_loss=0.03607, over 3075688.03 frames. ], batch size: 109, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:12:01,251 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 00:12:02,953 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192755.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:09,632 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:32,901 INFO [optim.py:368] (2/8) 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 00:13:08,085 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:21,321 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192796.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:28,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0173, 3.2992, 3.5449, 2.1154, 2.9853, 2.2766, 3.4581, 3.4650], device='cuda:2'), covar=tensor([0.0287, 0.0761, 0.0518, 0.1945, 0.0765, 0.0945, 0.0720, 0.1009], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0150, 0.0159, 0.0146, 0.0139, 0.0124, 0.0138, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:13:30,579 INFO [train.py:904] (2/8) Epoch 19, batch 10100, loss[loss=0.1697, simple_loss=0.2527, pruned_loss=0.04338, over 12469.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2637, pruned_loss=0.03612, over 3076141.32 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:13:37,883 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6388, 4.6198, 4.3834, 3.9252, 4.5156, 1.6807, 4.2513, 4.2470], device='cuda:2'), covar=tensor([0.0131, 0.0129, 0.0220, 0.0330, 0.0137, 0.2715, 0.0136, 0.0228], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0141, 0.0183, 0.0162, 0.0160, 0.0197, 0.0173, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:13:38,853 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:48,330 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192811.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:14:22,529 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0006, 2.7188, 2.9230, 2.0531, 2.7276, 2.1854, 2.6067, 2.9223], device='cuda:2'), covar=tensor([0.0355, 0.0967, 0.0421, 0.1822, 0.0768, 0.0872, 0.0763, 0.0964], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0149, 0.0159, 0.0146, 0.0139, 0.0123, 0.0137, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:14:28,831 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192831.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:15:13,689 INFO [train.py:904] (2/8) Epoch 20, batch 0, loss[loss=0.1764, simple_loss=0.2666, pruned_loss=0.04308, over 17196.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2666, pruned_loss=0.04308, over 17196.00 frames. ], batch size: 44, lr: 3.43e-03, grad_scale: 8.0 2023-05-01 00:15:13,689 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 00:15:21,155 INFO [train.py:938] (2/8) Epoch 20, validation: loss=0.146, simple_loss=0.2496, pruned_loss=0.02121, over 944034.00 frames. 2023-05-01 00:15:21,156 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 00:15:31,953 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:45,160 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6503, 2.4143, 2.3493, 4.6089, 2.3266, 2.7932, 2.4220, 2.5335], device='cuda:2'), covar=tensor([0.1118, 0.3599, 0.3002, 0.0376, 0.4129, 0.2534, 0.3539, 0.3576], device='cuda:2'), in_proj_covar=tensor([0.0383, 0.0423, 0.0352, 0.0310, 0.0423, 0.0484, 0.0395, 0.0492], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:15:49,198 INFO [optim.py:368] (2/8) 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 00:15:53,702 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192875.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:54,014 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 00:15:56,027 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:16:17,628 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192892.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:16:31,046 INFO [train.py:904] (2/8) Epoch 20, batch 50, loss[loss=0.2461, simple_loss=0.3138, pruned_loss=0.08923, over 11995.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2731, pruned_loss=0.05424, over 750119.96 frames. ], batch size: 246, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:16:31,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4031, 5.7833, 5.4856, 5.5595, 5.1346, 5.1948, 5.2152, 5.9190], device='cuda:2'), covar=tensor([0.1344, 0.1117, 0.1605, 0.0926, 0.0996, 0.0755, 0.1251, 0.0889], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0766, 0.0623, 0.0567, 0.0480, 0.0489, 0.0639, 0.0584], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:16:59,572 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:02,457 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:36,897 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9553, 4.9275, 5.4136, 5.3608, 5.4216, 5.0455, 4.9902, 4.7958], device='cuda:2'), covar=tensor([0.0371, 0.0505, 0.0409, 0.0510, 0.0565, 0.0458, 0.1153, 0.0469], device='cuda:2'), in_proj_covar=tensor([0.0380, 0.0413, 0.0405, 0.0379, 0.0449, 0.0425, 0.0514, 0.0344], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:17:37,169 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 00:17:38,938 INFO [train.py:904] (2/8) Epoch 20, batch 100, loss[loss=0.1796, simple_loss=0.2726, pruned_loss=0.04325, over 17120.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2688, pruned_loss=0.04931, over 1324343.58 frames. ], batch size: 49, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:18:07,333 INFO [optim.py:368] (2/8) 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 00:18:07,670 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192972.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:18:15,940 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6422, 4.5649, 4.9606, 4.9505, 5.0280, 4.6512, 4.6326, 4.4844], device='cuda:2'), covar=tensor([0.0366, 0.0556, 0.0382, 0.0441, 0.0545, 0.0487, 0.1079, 0.0609], device='cuda:2'), in_proj_covar=tensor([0.0382, 0.0416, 0.0407, 0.0381, 0.0452, 0.0427, 0.0517, 0.0346], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:18:48,432 INFO [train.py:904] (2/8) Epoch 20, batch 150, loss[loss=0.1645, simple_loss=0.2595, pruned_loss=0.03481, over 17113.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2667, pruned_loss=0.04836, over 1765203.88 frames. ], batch size: 47, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:19:27,068 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193029.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:43,937 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:52,879 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:55,120 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 00:19:58,179 INFO [train.py:904] (2/8) Epoch 20, batch 200, loss[loss=0.2061, simple_loss=0.2782, pruned_loss=0.06702, over 16749.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2658, pruned_loss=0.04786, over 2112872.25 frames. ], batch size: 124, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:20:03,580 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193055.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:27,491 INFO [optim.py:368] (2/8) 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 00:20:39,506 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5600, 4.6993, 4.8606, 4.6734, 4.6983, 5.3124, 4.7686, 4.4205], device='cuda:2'), covar=tensor([0.1653, 0.2278, 0.2773, 0.2175, 0.2993, 0.1204, 0.1957, 0.2838], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0562, 0.0622, 0.0469, 0.0627, 0.0651, 0.0489, 0.0624], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 00:20:42,423 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-01 00:20:50,737 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:53,813 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193091.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:57,997 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4485, 5.8132, 5.5570, 5.6169, 5.1938, 5.1973, 5.2556, 5.9353], device='cuda:2'), covar=tensor([0.1392, 0.0985, 0.1249, 0.0900, 0.0991, 0.0713, 0.1107, 0.1037], device='cuda:2'), in_proj_covar=tensor([0.0638, 0.0787, 0.0638, 0.0583, 0.0492, 0.0501, 0.0655, 0.0601], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:21:00,957 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:08,806 INFO [train.py:904] (2/8) Epoch 20, batch 250, loss[loss=0.1723, simple_loss=0.2611, pruned_loss=0.04168, over 16708.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2638, pruned_loss=0.04694, over 2369401.93 frames. ], batch size: 57, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:21:09,283 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:57,542 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193137.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:22:07,341 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2528, 2.1817, 2.2944, 3.9986, 2.2909, 2.5365, 2.2076, 2.3516], device='cuda:2'), covar=tensor([0.1449, 0.3795, 0.2998, 0.0600, 0.3715, 0.2541, 0.4037, 0.3002], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0431, 0.0358, 0.0318, 0.0429, 0.0494, 0.0402, 0.0502], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:22:17,119 INFO [train.py:904] (2/8) Epoch 20, batch 300, loss[loss=0.1838, simple_loss=0.26, pruned_loss=0.05381, over 16816.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2625, pruned_loss=0.04653, over 2583159.95 frames. ], batch size: 124, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:22:46,393 INFO [optim.py:368] (2/8) 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 00:23:05,584 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193187.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:23:10,399 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0015, 4.0989, 2.6381, 4.6745, 3.3082, 4.6386, 2.6155, 3.3304], device='cuda:2'), covar=tensor([0.0278, 0.0403, 0.1494, 0.0288, 0.0814, 0.0574, 0.1531, 0.0701], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0157, 0.0176, 0.0212, 0.0202, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:23:28,161 INFO [train.py:904] (2/8) Epoch 20, batch 350, loss[loss=0.1706, simple_loss=0.2641, pruned_loss=0.03857, over 16726.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.26, pruned_loss=0.04517, over 2752961.23 frames. ], batch size: 57, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:23:39,636 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:24:37,895 INFO [train.py:904] (2/8) Epoch 20, batch 400, loss[loss=0.1529, simple_loss=0.2426, pruned_loss=0.03159, over 17231.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2588, pruned_loss=0.04546, over 2864781.56 frames. ], batch size: 45, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:25:03,042 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0118, 4.4925, 3.2249, 2.4153, 2.8054, 2.6182, 4.8210, 3.7003], device='cuda:2'), covar=tensor([0.2713, 0.0524, 0.1678, 0.2862, 0.2952, 0.2138, 0.0348, 0.1379], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0264, 0.0298, 0.0303, 0.0287, 0.0251, 0.0286, 0.0328], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 00:25:05,916 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193272.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:06,023 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193272.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:25:06,669 INFO [optim.py:368] (2/8) 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 00:25:23,435 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193284.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:46,656 INFO [train.py:904] (2/8) Epoch 20, batch 450, loss[loss=0.1793, simple_loss=0.2545, pruned_loss=0.05206, over 16747.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.258, pruned_loss=0.0446, over 2975148.30 frames. ], batch size: 124, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:26:13,643 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:26,049 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:48,822 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:57,285 INFO [train.py:904] (2/8) Epoch 20, batch 500, loss[loss=0.1796, simple_loss=0.2722, pruned_loss=0.04347, over 17038.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2558, pruned_loss=0.04364, over 3058642.13 frames. ], batch size: 53, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:27:26,048 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:47,556 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2821, 4.1435, 4.5376, 2.3907, 4.7560, 4.8594, 3.4521, 3.7233], device='cuda:2'), covar=tensor([0.0640, 0.0252, 0.0208, 0.1142, 0.0073, 0.0136, 0.0416, 0.0393], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0106, 0.0095, 0.0139, 0.0078, 0.0121, 0.0125, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:27:51,479 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193391.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:59,427 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:07,092 INFO [train.py:904] (2/8) Epoch 20, batch 550, loss[loss=0.1545, simple_loss=0.239, pruned_loss=0.03493, over 17028.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2544, pruned_loss=0.04289, over 3107451.97 frames. ], batch size: 41, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:28:32,240 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:46,260 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7932, 4.2530, 3.0026, 2.2756, 2.6965, 2.5720, 4.6311, 3.5788], device='cuda:2'), covar=tensor([0.2941, 0.0577, 0.1752, 0.2890, 0.2779, 0.2098, 0.0333, 0.1333], device='cuda:2'), in_proj_covar=tensor([0.0325, 0.0265, 0.0300, 0.0305, 0.0289, 0.0253, 0.0287, 0.0330], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 00:28:56,785 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:29:07,419 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6988, 1.8168, 2.2962, 2.5001, 2.6073, 2.6067, 1.9113, 2.7757], device='cuda:2'), covar=tensor([0.0183, 0.0466, 0.0324, 0.0284, 0.0297, 0.0310, 0.0483, 0.0170], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0190, 0.0175, 0.0177, 0.0191, 0.0148, 0.0192, 0.0141], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:29:14,799 INFO [train.py:904] (2/8) Epoch 20, batch 600, loss[loss=0.1664, simple_loss=0.2439, pruned_loss=0.0444, over 16224.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2532, pruned_loss=0.04289, over 3156404.77 frames. ], batch size: 164, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:29:18,175 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193454.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:29:21,708 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6631, 3.7013, 2.1904, 4.0145, 2.9453, 3.8653, 2.3715, 2.9973], device='cuda:2'), covar=tensor([0.0261, 0.0418, 0.1604, 0.0346, 0.0696, 0.0789, 0.1346, 0.0684], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0177, 0.0194, 0.0158, 0.0176, 0.0214, 0.0203, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:29:22,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1919, 5.1229, 4.8942, 4.3860, 4.9685, 1.9030, 4.7554, 4.7856], device='cuda:2'), covar=tensor([0.0083, 0.0086, 0.0202, 0.0389, 0.0106, 0.2784, 0.0139, 0.0229], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0147, 0.0190, 0.0170, 0.0166, 0.0203, 0.0180, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:29:43,211 INFO [optim.py:368] (2/8) 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 00:29:55,228 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193482.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:30:03,351 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193487.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:30:21,022 INFO [train.py:904] (2/8) Epoch 20, batch 650, loss[loss=0.1534, simple_loss=0.2349, pruned_loss=0.03591, over 12211.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2512, pruned_loss=0.04166, over 3183824.81 frames. ], batch size: 246, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:30:39,889 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193515.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:31:07,551 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:31:29,637 INFO [train.py:904] (2/8) Epoch 20, batch 700, loss[loss=0.1897, simple_loss=0.2594, pruned_loss=0.06, over 12148.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2509, pruned_loss=0.04111, over 3202743.82 frames. ], batch size: 246, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:31:48,973 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193567.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:31:57,449 INFO [optim.py:368] (2/8) 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 00:32:25,086 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5378, 3.5695, 2.2284, 3.8469, 2.8607, 3.7536, 2.3711, 2.9003], device='cuda:2'), covar=tensor([0.0260, 0.0413, 0.1439, 0.0330, 0.0709, 0.0746, 0.1300, 0.0686], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0196, 0.0161, 0.0178, 0.0216, 0.0204, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:32:35,677 INFO [train.py:904] (2/8) Epoch 20, batch 750, loss[loss=0.1642, simple_loss=0.2451, pruned_loss=0.04162, over 16801.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2518, pruned_loss=0.04145, over 3231412.31 frames. ], batch size: 102, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:32:54,641 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:28,522 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:42,945 INFO [train.py:904] (2/8) Epoch 20, batch 800, loss[loss=0.1439, simple_loss=0.2209, pruned_loss=0.03342, over 16813.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2515, pruned_loss=0.04125, over 3253683.76 frames. ], batch size: 39, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:34:10,661 INFO [optim.py:368] (2/8) 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 00:34:16,665 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:45,105 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:52,173 INFO [train.py:904] (2/8) Epoch 20, batch 850, loss[loss=0.1662, simple_loss=0.2455, pruned_loss=0.04341, over 15584.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2513, pruned_loss=0.04104, over 3268998.10 frames. ], batch size: 191, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:35:02,457 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3290, 4.3790, 4.6868, 4.7193, 4.7168, 4.4439, 4.4529, 4.3216], device='cuda:2'), covar=tensor([0.0390, 0.0626, 0.0420, 0.0360, 0.0492, 0.0428, 0.0827, 0.0631], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0440, 0.0427, 0.0399, 0.0476, 0.0448, 0.0541, 0.0362], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:35:50,808 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:35:59,338 INFO [train.py:904] (2/8) Epoch 20, batch 900, loss[loss=0.1414, simple_loss=0.2276, pruned_loss=0.02762, over 16797.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2508, pruned_loss=0.04035, over 3287771.45 frames. ], batch size: 39, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:36:14,713 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 00:36:28,231 INFO [optim.py:368] (2/8) 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 00:36:33,803 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193777.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:37:09,296 INFO [train.py:904] (2/8) Epoch 20, batch 950, loss[loss=0.1596, simple_loss=0.2602, pruned_loss=0.02948, over 17137.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2505, pruned_loss=0.03984, over 3300972.64 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:37:20,552 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193810.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:37:21,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5890, 3.6321, 3.3694, 3.0002, 3.2203, 3.4917, 3.3421, 3.2858], device='cuda:2'), covar=tensor([0.0624, 0.0663, 0.0319, 0.0339, 0.0617, 0.0455, 0.1279, 0.0546], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0420, 0.0340, 0.0334, 0.0351, 0.0391, 0.0236, 0.0408], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:38:03,615 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.70 vs. limit=5.0 2023-05-01 00:38:09,531 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3660, 3.5390, 3.6642, 3.6403, 3.6485, 3.5015, 3.5153, 3.5438], device='cuda:2'), covar=tensor([0.0404, 0.0688, 0.0462, 0.0422, 0.0563, 0.0494, 0.0732, 0.0480], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0439, 0.0425, 0.0398, 0.0475, 0.0447, 0.0539, 0.0362], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:38:17,860 INFO [train.py:904] (2/8) Epoch 20, batch 1000, loss[loss=0.1698, simple_loss=0.2621, pruned_loss=0.03872, over 16751.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2494, pruned_loss=0.0396, over 3293582.17 frames. ], batch size: 57, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:38:39,219 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193867.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:38:45,939 INFO [optim.py:368] (2/8) 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 00:39:00,602 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193884.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:39:24,779 INFO [train.py:904] (2/8) Epoch 20, batch 1050, loss[loss=0.1612, simple_loss=0.2512, pruned_loss=0.03555, over 16729.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2493, pruned_loss=0.03952, over 3297520.80 frames. ], batch size: 62, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:39:43,670 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:18,121 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:25,162 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:35,457 INFO [train.py:904] (2/8) Epoch 20, batch 1100, loss[loss=0.1423, simple_loss=0.2348, pruned_loss=0.02489, over 16867.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2489, pruned_loss=0.03951, over 3307973.20 frames. ], batch size: 42, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:41:01,676 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:04,217 INFO [optim.py:368] (2/8) 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 00:41:24,329 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:46,949 INFO [train.py:904] (2/8) Epoch 20, batch 1150, loss[loss=0.1638, simple_loss=0.2641, pruned_loss=0.03177, over 17035.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2485, pruned_loss=0.03922, over 3300134.50 frames. ], batch size: 50, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:42:56,046 INFO [train.py:904] (2/8) Epoch 20, batch 1200, loss[loss=0.1609, simple_loss=0.2503, pruned_loss=0.03574, over 17009.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2485, pruned_loss=0.03885, over 3307697.12 frames. ], batch size: 55, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:43:25,016 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:43:25,793 INFO [optim.py:368] (2/8) 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 00:43:33,063 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194077.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:43:45,890 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7415, 2.9943, 2.8720, 1.9380, 2.5644, 1.9251, 3.2447, 3.2478], device='cuda:2'), covar=tensor([0.0262, 0.0933, 0.0765, 0.2251, 0.1135, 0.1244, 0.0618, 0.1046], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0161, 0.0167, 0.0153, 0.0145, 0.0129, 0.0145, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:44:06,933 INFO [train.py:904] (2/8) Epoch 20, batch 1250, loss[loss=0.1655, simple_loss=0.2659, pruned_loss=0.0326, over 17253.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.249, pruned_loss=0.03901, over 3311545.49 frames. ], batch size: 52, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:44:17,554 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194110.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:44:38,137 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:44:38,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9543, 2.8870, 2.6477, 4.3220, 3.6006, 4.2169, 1.6449, 3.0987], device='cuda:2'), covar=tensor([0.1242, 0.0669, 0.1126, 0.0202, 0.0188, 0.0405, 0.1514, 0.0763], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0173, 0.0192, 0.0184, 0.0200, 0.0213, 0.0198, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:44:49,867 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:45:02,584 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 00:45:15,923 INFO [train.py:904] (2/8) Epoch 20, batch 1300, loss[loss=0.1652, simple_loss=0.2416, pruned_loss=0.04444, over 16543.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2486, pruned_loss=0.03924, over 3299296.91 frames. ], batch size: 75, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:45:26,572 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194158.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:45:43,535 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9262, 4.9749, 5.3823, 5.4115, 5.3910, 5.0870, 4.9933, 4.8104], device='cuda:2'), covar=tensor([0.0375, 0.0616, 0.0452, 0.0417, 0.0437, 0.0404, 0.0962, 0.0491], device='cuda:2'), in_proj_covar=tensor([0.0406, 0.0446, 0.0431, 0.0403, 0.0481, 0.0452, 0.0546, 0.0365], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:45:46,645 INFO [optim.py:368] (2/8) 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 00:45:51,136 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1041, 5.1177, 4.8761, 3.6150, 4.9169, 1.6607, 4.5213, 4.7336], device='cuda:2'), covar=tensor([0.0139, 0.0118, 0.0270, 0.0803, 0.0171, 0.3537, 0.0227, 0.0370], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0151, 0.0196, 0.0175, 0.0172, 0.0207, 0.0185, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:46:27,312 INFO [train.py:904] (2/8) Epoch 20, batch 1350, loss[loss=0.144, simple_loss=0.2251, pruned_loss=0.03144, over 16985.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2484, pruned_loss=0.03914, over 3304967.21 frames. ], batch size: 41, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:47:20,532 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194240.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:47:36,939 INFO [train.py:904] (2/8) Epoch 20, batch 1400, loss[loss=0.1396, simple_loss=0.2232, pruned_loss=0.02801, over 16253.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2488, pruned_loss=0.03944, over 3311279.94 frames. ], batch size: 36, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:47:59,694 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 00:48:03,377 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194271.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:48:06,429 INFO [optim.py:368] (2/8) 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 00:48:44,297 INFO [train.py:904] (2/8) Epoch 20, batch 1450, loss[loss=0.1424, simple_loss=0.2325, pruned_loss=0.02613, over 17259.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2478, pruned_loss=0.03935, over 3314396.78 frames. ], batch size: 52, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:06,335 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194318.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:08,664 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:54,003 INFO [train.py:904] (2/8) Epoch 20, batch 1500, loss[loss=0.1374, simple_loss=0.2211, pruned_loss=0.02688, over 16802.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2484, pruned_loss=0.03983, over 3322492.13 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:55,457 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:50:07,918 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194362.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:50:24,392 INFO [optim.py:368] (2/8) 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 00:50:31,282 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:50:40,816 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5479, 3.6172, 3.3971, 3.0155, 3.2117, 3.4897, 3.3272, 3.3596], device='cuda:2'), covar=tensor([0.0592, 0.0645, 0.0314, 0.0289, 0.0609, 0.0426, 0.1386, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0426, 0.0345, 0.0339, 0.0357, 0.0397, 0.0239, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:51:03,653 INFO [train.py:904] (2/8) Epoch 20, batch 1550, loss[loss=0.2026, simple_loss=0.2846, pruned_loss=0.0603, over 15417.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2493, pruned_loss=0.04122, over 3307274.32 frames. ], batch size: 190, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:51:20,874 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:34,180 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194423.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:51:40,589 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:52:13,192 INFO [train.py:904] (2/8) Epoch 20, batch 1600, loss[loss=0.1701, simple_loss=0.246, pruned_loss=0.04708, over 16723.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2512, pruned_loss=0.04167, over 3319733.74 frames. ], batch size: 134, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:52:20,886 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4837, 2.3287, 2.3660, 4.1805, 2.1826, 2.6755, 2.3978, 2.4379], device='cuda:2'), covar=tensor([0.1182, 0.3461, 0.2984, 0.0529, 0.4284, 0.2562, 0.3521, 0.3454], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0440, 0.0364, 0.0326, 0.0435, 0.0506, 0.0409, 0.0515], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:52:43,858 INFO [optim.py:368] (2/8) 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 00:53:22,714 INFO [train.py:904] (2/8) Epoch 20, batch 1650, loss[loss=0.1707, simple_loss=0.2726, pruned_loss=0.03437, over 16747.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2535, pruned_loss=0.04255, over 3324335.98 frames. ], batch size: 57, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:54:16,710 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194540.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:54:33,460 INFO [train.py:904] (2/8) Epoch 20, batch 1700, loss[loss=0.1777, simple_loss=0.2625, pruned_loss=0.04646, over 15526.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2546, pruned_loss=0.04291, over 3315658.54 frames. ], batch size: 191, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:55:05,696 INFO [optim.py:368] (2/8) 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 00:55:25,647 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 20, batch 1750, loss[loss=0.1752, simple_loss=0.2542, pruned_loss=0.04807, over 16491.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.256, pruned_loss=0.04369, over 3299637.43 frames. ], batch size: 146, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:55:53,767 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 00:56:27,183 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194634.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:56:51,746 INFO [train.py:904] (2/8) Epoch 20, batch 1800, loss[loss=0.1755, simple_loss=0.2705, pruned_loss=0.04029, over 16744.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2571, pruned_loss=0.04354, over 3300458.34 frames. ], batch size: 62, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:57:06,067 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7992, 3.8450, 1.9677, 4.4677, 2.9417, 4.3848, 2.1583, 3.0694], device='cuda:2'), covar=tensor([0.0323, 0.0413, 0.2130, 0.0292, 0.0863, 0.0455, 0.1994, 0.0769], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0180, 0.0198, 0.0165, 0.0179, 0.0220, 0.0205, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 00:57:10,149 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9073, 4.0004, 4.2866, 4.2590, 4.2859, 4.0203, 4.0832, 4.0404], device='cuda:2'), covar=tensor([0.0406, 0.0609, 0.0417, 0.0454, 0.0527, 0.0502, 0.0784, 0.0553], device='cuda:2'), in_proj_covar=tensor([0.0411, 0.0452, 0.0435, 0.0410, 0.0488, 0.0462, 0.0553, 0.0370], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 00:57:22,476 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:57:23,424 INFO [optim.py:368] (2/8) 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 00:57:37,257 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8673, 2.6874, 2.6179, 1.8982, 2.5578, 2.7483, 2.5894, 1.8499], device='cuda:2'), covar=tensor([0.0453, 0.0098, 0.0081, 0.0388, 0.0145, 0.0123, 0.0121, 0.0392], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0082, 0.0082, 0.0135, 0.0097, 0.0108, 0.0094, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:2') 2023-05-01 00:57:50,527 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194695.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:57:58,019 INFO [train.py:904] (2/8) Epoch 20, batch 1850, loss[loss=0.1408, simple_loss=0.2251, pruned_loss=0.02823, over 16754.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.258, pruned_loss=0.04356, over 3306842.45 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:57:58,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2360, 5.8310, 6.0151, 5.7029, 5.7894, 6.3546, 5.8298, 5.5514], device='cuda:2'), covar=tensor([0.0869, 0.1864, 0.1941, 0.2111, 0.2808, 0.1002, 0.1607, 0.2358], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0592, 0.0651, 0.0492, 0.0657, 0.0683, 0.0507, 0.0656], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 00:58:08,652 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:10,821 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 00:58:21,440 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194718.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:58:33,760 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:40,794 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5732, 3.6213, 3.4068, 3.0032, 3.2552, 3.4726, 3.3475, 3.2989], device='cuda:2'), covar=tensor([0.0632, 0.0746, 0.0296, 0.0279, 0.0525, 0.0472, 0.1434, 0.0479], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0428, 0.0346, 0.0340, 0.0358, 0.0399, 0.0239, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 00:59:06,818 INFO [train.py:904] (2/8) Epoch 20, batch 1900, loss[loss=0.1846, simple_loss=0.2679, pruned_loss=0.05063, over 16540.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2567, pruned_loss=0.04253, over 3308506.40 frames. ], batch size: 75, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:59:14,295 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 00:59:38,382 INFO [optim.py:368] (2/8) 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 00:59:41,138 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:00:16,324 INFO [train.py:904] (2/8) Epoch 20, batch 1950, loss[loss=0.1364, simple_loss=0.2243, pruned_loss=0.02429, over 16989.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2558, pruned_loss=0.04177, over 3318256.87 frames. ], batch size: 41, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 01:00:19,394 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 01:00:44,358 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 01:00:49,228 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3367, 4.2241, 4.2412, 3.9784, 3.9910, 4.2895, 4.0173, 4.0465], device='cuda:2'), covar=tensor([0.0738, 0.0804, 0.0348, 0.0310, 0.0738, 0.0610, 0.0796, 0.0668], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0428, 0.0346, 0.0340, 0.0357, 0.0399, 0.0239, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:01:17,138 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:01:23,588 INFO [train.py:904] (2/8) Epoch 20, batch 2000, loss[loss=0.1963, simple_loss=0.2705, pruned_loss=0.06107, over 16814.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.256, pruned_loss=0.04155, over 3314362.99 frames. ], batch size: 124, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:01:54,983 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 20, batch 2050, loss[loss=0.1487, simple_loss=0.238, pruned_loss=0.02975, over 16859.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2558, pruned_loss=0.0412, over 3319132.24 frames. ], batch size: 42, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:02:41,450 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194908.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:02:55,027 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7485, 3.9246, 2.9761, 2.2633, 2.6230, 2.4224, 4.0512, 3.4479], device='cuda:2'), covar=tensor([0.2700, 0.0609, 0.1736, 0.3149, 0.2665, 0.2081, 0.0554, 0.1497], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0267, 0.0300, 0.0304, 0.0293, 0.0253, 0.0289, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 01:03:11,906 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 01:03:18,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6324, 2.5512, 1.6443, 2.7433, 2.0890, 2.7906, 1.9435, 2.2838], device='cuda:2'), covar=tensor([0.0291, 0.0377, 0.1542, 0.0264, 0.0676, 0.0490, 0.1488, 0.0664], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0180, 0.0196, 0.0164, 0.0178, 0.0220, 0.0203, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:03:26,644 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0132, 2.1366, 2.6930, 2.9792, 2.7631, 3.4759, 2.4874, 3.4106], device='cuda:2'), covar=tensor([0.0260, 0.0484, 0.0301, 0.0319, 0.0356, 0.0202, 0.0405, 0.0162], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0193, 0.0178, 0.0181, 0.0195, 0.0152, 0.0195, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:03:41,679 INFO [train.py:904] (2/8) Epoch 20, batch 2100, loss[loss=0.1701, simple_loss=0.2697, pruned_loss=0.03523, over 16648.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2578, pruned_loss=0.04228, over 3319894.70 frames. ], batch size: 62, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:04:12,821 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 01:04:34,696 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:45,221 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 01:04:50,814 INFO [train.py:904] (2/8) Epoch 20, batch 2150, loss[loss=0.1568, simple_loss=0.2414, pruned_loss=0.0361, over 17219.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2583, pruned_loss=0.04266, over 3315656.71 frames. ], batch size: 43, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:05:01,399 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:03,757 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:13,355 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195018.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 01:05:18,549 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:49,743 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195044.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:05:55,271 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5782, 2.3289, 2.2212, 4.2922, 2.2329, 2.7285, 2.3462, 2.4998], device='cuda:2'), covar=tensor([0.1203, 0.3609, 0.3212, 0.0524, 0.4322, 0.2687, 0.3459, 0.3895], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0440, 0.0363, 0.0326, 0.0433, 0.0506, 0.0409, 0.0515], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:06:01,906 INFO [train.py:904] (2/8) Epoch 20, batch 2200, loss[loss=0.1763, simple_loss=0.2763, pruned_loss=0.03818, over 17042.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2593, pruned_loss=0.04334, over 3322252.62 frames. ], batch size: 50, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:06:09,186 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:20,444 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195066.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:06:28,229 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:30,141 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195072.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 01:06:33,763 INFO [optim.py:368] (2/8) 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 01:07:10,887 INFO [train.py:904] (2/8) Epoch 20, batch 2250, loss[loss=0.1707, simple_loss=0.2543, pruned_loss=0.04358, over 16484.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2602, pruned_loss=0.04377, over 3328700.41 frames. ], batch size: 146, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:07:11,361 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8399, 3.1296, 2.5509, 5.0294, 3.9935, 4.4291, 1.5844, 3.1824], device='cuda:2'), covar=tensor([0.1378, 0.0710, 0.1312, 0.0196, 0.0212, 0.0394, 0.1653, 0.0774], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0174, 0.0192, 0.0187, 0.0203, 0.0215, 0.0198, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:07:15,439 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195105.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 01:07:54,046 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:08:21,143 INFO [train.py:904] (2/8) Epoch 20, batch 2300, loss[loss=0.1758, simple_loss=0.2578, pruned_loss=0.04686, over 16677.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2604, pruned_loss=0.04356, over 3338352.44 frames. ], batch size: 83, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:08:24,574 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5339, 3.8400, 3.9579, 2.9278, 3.6327, 4.0208, 3.6974, 2.2837], device='cuda:2'), covar=tensor([0.0499, 0.0208, 0.0052, 0.0339, 0.0105, 0.0090, 0.0092, 0.0469], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0132, 0.0095, 0.0107, 0.0093, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 01:08:51,383 INFO [optim.py:368] (2/8) 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 01:09:03,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9081, 3.7086, 4.2441, 1.9681, 4.5009, 4.5163, 3.2145, 3.3727], device='cuda:2'), covar=tensor([0.0734, 0.0279, 0.0233, 0.1248, 0.0077, 0.0214, 0.0430, 0.0437], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0098, 0.0140, 0.0079, 0.0125, 0.0127, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:09:04,504 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4121, 4.4293, 4.7377, 4.7330, 4.7759, 4.4696, 4.4968, 4.3144], device='cuda:2'), covar=tensor([0.0396, 0.0672, 0.0434, 0.0476, 0.0477, 0.0437, 0.0838, 0.0707], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0447, 0.0432, 0.0406, 0.0482, 0.0458, 0.0546, 0.0368], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 01:09:29,600 INFO [train.py:904] (2/8) Epoch 20, batch 2350, loss[loss=0.1759, simple_loss=0.2514, pruned_loss=0.05017, over 16927.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2609, pruned_loss=0.04428, over 3330208.68 frames. ], batch size: 109, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:09:31,097 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:10:36,693 INFO [train.py:904] (2/8) Epoch 20, batch 2400, loss[loss=0.1842, simple_loss=0.2623, pruned_loss=0.05301, over 16411.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2613, pruned_loss=0.04426, over 3329316.40 frames. ], batch size: 146, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:11:07,015 INFO [optim.py:368] (2/8) 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 01:11:29,181 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:11:37,002 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0508, 4.2218, 2.9582, 4.9751, 3.3632, 4.7997, 2.9586, 3.4279], device='cuda:2'), covar=tensor([0.0251, 0.0319, 0.1347, 0.0165, 0.0739, 0.0368, 0.1307, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0180, 0.0196, 0.0165, 0.0178, 0.0220, 0.0204, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:11:45,805 INFO [train.py:904] (2/8) Epoch 20, batch 2450, loss[loss=0.1917, simple_loss=0.2687, pruned_loss=0.05733, over 16478.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2609, pruned_loss=0.04388, over 3328487.91 frames. ], batch size: 146, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:12:08,515 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7040, 4.2703, 4.2784, 3.2500, 3.7138, 4.2706, 3.8600, 2.4714], device='cuda:2'), covar=tensor([0.0457, 0.0064, 0.0045, 0.0281, 0.0112, 0.0083, 0.0087, 0.0423], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0132, 0.0095, 0.0107, 0.0093, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 01:12:14,230 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1539, 5.6531, 5.8577, 5.5135, 5.5972, 6.1966, 5.7558, 5.5145], device='cuda:2'), covar=tensor([0.0873, 0.1905, 0.2160, 0.2013, 0.2542, 0.0914, 0.1265, 0.2001], device='cuda:2'), in_proj_covar=tensor([0.0415, 0.0607, 0.0664, 0.0505, 0.0669, 0.0696, 0.0513, 0.0673], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 01:12:35,686 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:12:54,075 INFO [train.py:904] (2/8) Epoch 20, batch 2500, loss[loss=0.1848, simple_loss=0.2721, pruned_loss=0.0488, over 17062.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2608, pruned_loss=0.04372, over 3314460.42 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:12:54,822 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 01:13:15,893 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195367.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 01:13:26,926 INFO [optim.py:368] (2/8) 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 01:13:50,872 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:14:01,553 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195400.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 01:14:04,077 INFO [train.py:904] (2/8) Epoch 20, batch 2550, loss[loss=0.1634, simple_loss=0.2494, pruned_loss=0.03875, over 16462.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2609, pruned_loss=0.04356, over 3323747.74 frames. ], batch size: 75, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:14:38,105 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:11,790 INFO [train.py:904] (2/8) Epoch 20, batch 2600, loss[loss=0.1468, simple_loss=0.2389, pruned_loss=0.02736, over 16851.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2608, pruned_loss=0.04351, over 3318541.11 frames. ], batch size: 42, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:15:13,377 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:40,233 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4524, 5.8180, 5.6246, 5.7051, 5.2885, 5.2594, 5.2876, 5.9888], device='cuda:2'), covar=tensor([0.1472, 0.1079, 0.1051, 0.0880, 0.0886, 0.0726, 0.1161, 0.0887], device='cuda:2'), in_proj_covar=tensor([0.0661, 0.0817, 0.0665, 0.0610, 0.0511, 0.0515, 0.0681, 0.0624], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:15:42,986 INFO [optim.py:368] (2/8) 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 01:16:00,902 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4272, 5.4055, 5.2753, 4.8389, 4.9384, 5.3784, 5.2640, 4.9580], device='cuda:2'), covar=tensor([0.0592, 0.0433, 0.0260, 0.0310, 0.0998, 0.0354, 0.0250, 0.0722], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0432, 0.0349, 0.0344, 0.0361, 0.0400, 0.0241, 0.0417], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:16:20,750 INFO [train.py:904] (2/8) Epoch 20, batch 2650, loss[loss=0.1616, simple_loss=0.2577, pruned_loss=0.03278, over 17133.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2609, pruned_loss=0.04273, over 3324544.58 frames. ], batch size: 48, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:16:22,192 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:16:24,520 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6920, 1.7753, 2.2819, 2.4770, 2.5810, 2.6486, 1.9214, 2.7677], device='cuda:2'), covar=tensor([0.0157, 0.0485, 0.0312, 0.0287, 0.0287, 0.0265, 0.0522, 0.0157], device='cuda:2'), in_proj_covar=tensor([0.0187, 0.0194, 0.0179, 0.0183, 0.0195, 0.0153, 0.0196, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:16:48,650 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8620, 4.8303, 5.3261, 5.3028, 5.3114, 4.9656, 4.9082, 4.6882], device='cuda:2'), covar=tensor([0.0325, 0.0540, 0.0358, 0.0374, 0.0507, 0.0386, 0.0983, 0.0500], device='cuda:2'), in_proj_covar=tensor([0.0406, 0.0448, 0.0432, 0.0406, 0.0480, 0.0457, 0.0548, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 01:17:03,658 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 01:17:20,775 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-01 01:17:28,727 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:29,628 INFO [train.py:904] (2/8) Epoch 20, batch 2700, loss[loss=0.181, simple_loss=0.2753, pruned_loss=0.04337, over 17112.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2611, pruned_loss=0.04223, over 3320546.59 frames. ], batch size: 48, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:18:00,661 INFO [optim.py:368] (2/8) 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 01:18:31,052 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 01:18:39,437 INFO [train.py:904] (2/8) Epoch 20, batch 2750, loss[loss=0.1804, simple_loss=0.2587, pruned_loss=0.05102, over 16751.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2608, pruned_loss=0.04189, over 3320723.37 frames. ], batch size: 83, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:19:40,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5678, 6.0052, 5.7419, 5.8141, 5.3984, 5.4615, 5.3542, 6.1464], device='cuda:2'), covar=tensor([0.1447, 0.0997, 0.1065, 0.0904, 0.0894, 0.0649, 0.1224, 0.0857], device='cuda:2'), in_proj_covar=tensor([0.0667, 0.0822, 0.0670, 0.0615, 0.0515, 0.0519, 0.0686, 0.0628], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:19:47,489 INFO [train.py:904] (2/8) Epoch 20, batch 2800, loss[loss=0.1654, simple_loss=0.2472, pruned_loss=0.04177, over 16834.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2594, pruned_loss=0.04185, over 3324727.88 frames. ], batch size: 102, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:20:07,216 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195667.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 01:20:18,614 INFO [optim.py:368] (2/8) 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 01:20:39,430 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7875, 4.0275, 2.5695, 4.6476, 3.0125, 4.5591, 2.5839, 3.2197], device='cuda:2'), covar=tensor([0.0301, 0.0333, 0.1484, 0.0207, 0.0787, 0.0414, 0.1469, 0.0715], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0195, 0.0165, 0.0178, 0.0219, 0.0203, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:20:52,430 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195700.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:20:54,302 INFO [train.py:904] (2/8) Epoch 20, batch 2850, loss[loss=0.1749, simple_loss=0.2475, pruned_loss=0.05117, over 16877.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2592, pruned_loss=0.04192, over 3327596.73 frames. ], batch size: 109, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:21:13,264 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195715.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:27,871 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 01:21:29,684 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:56,478 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195748.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:21:56,485 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:22:02,475 INFO [train.py:904] (2/8) Epoch 20, batch 2900, loss[loss=0.1835, simple_loss=0.2593, pruned_loss=0.05386, over 16916.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2581, pruned_loss=0.04177, over 3337291.15 frames. ], batch size: 116, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:22:33,088 INFO [optim.py:368] (2/8) 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 01:22:33,332 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:22:44,761 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1184, 4.4571, 4.4862, 3.1680, 3.7861, 4.4817, 4.0179, 2.7854], device='cuda:2'), covar=tensor([0.0447, 0.0073, 0.0044, 0.0380, 0.0138, 0.0108, 0.0089, 0.0451], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0083, 0.0082, 0.0134, 0.0097, 0.0109, 0.0094, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:2') 2023-05-01 01:23:10,990 INFO [train.py:904] (2/8) Epoch 20, batch 2950, loss[loss=0.1872, simple_loss=0.263, pruned_loss=0.05573, over 16481.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2567, pruned_loss=0.04165, over 3343564.63 frames. ], batch size: 75, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,947 INFO [train.py:904] (2/8) Epoch 20, batch 3000, loss[loss=0.1702, simple_loss=0.2613, pruned_loss=0.03961, over 17116.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2576, pruned_loss=0.04253, over 3341278.88 frames. ], batch size: 47, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,948 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 01:24:25,661 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2122, 3.2970, 3.1465, 4.9709, 3.9135, 4.5513, 2.0252, 3.5012], device='cuda:2'), covar=tensor([0.1125, 0.0629, 0.0921, 0.0162, 0.0151, 0.0335, 0.1377, 0.0624], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0173, 0.0192, 0.0187, 0.0204, 0.0214, 0.0197, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:24:27,126 INFO [train.py:938] (2/8) Epoch 20, validation: loss=0.1354, simple_loss=0.2409, pruned_loss=0.01492, over 944034.00 frames. 2023-05-01 01:24:27,127 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 01:24:58,716 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 20, batch 3050, loss[loss=0.1698, simple_loss=0.2633, pruned_loss=0.03817, over 17078.00 frames. ], tot_loss[loss=0.171, simple_loss=0.257, pruned_loss=0.04249, over 3340148.10 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:26:46,791 INFO [train.py:904] (2/8) Epoch 20, batch 3100, loss[loss=0.1584, simple_loss=0.2487, pruned_loss=0.03403, over 16016.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2576, pruned_loss=0.04307, over 3341360.68 frames. ], batch size: 35, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:27:16,978 INFO [optim.py:368] (2/8) 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 01:27:55,507 INFO [train.py:904] (2/8) Epoch 20, batch 3150, loss[loss=0.1835, simple_loss=0.2614, pruned_loss=0.05286, over 16419.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2563, pruned_loss=0.04298, over 3341760.12 frames. ], batch size: 146, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:28:30,459 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 01:28:57,195 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:29:03,218 INFO [train.py:904] (2/8) Epoch 20, batch 3200, loss[loss=0.1822, simple_loss=0.2731, pruned_loss=0.04569, over 16703.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2549, pruned_loss=0.04192, over 3346724.41 frames. ], batch size: 62, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:29:35,518 INFO [optim.py:368] (2/8) 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 01:29:47,655 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5523, 3.4717, 2.8326, 2.2002, 2.3198, 2.2525, 3.6110, 3.1013], device='cuda:2'), covar=tensor([0.2597, 0.0689, 0.1549, 0.2731, 0.2629, 0.2188, 0.0527, 0.1599], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0270, 0.0303, 0.0307, 0.0297, 0.0255, 0.0292, 0.0336], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 01:30:04,479 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=196096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:30:11,566 INFO [train.py:904] (2/8) Epoch 20, batch 3250, loss[loss=0.225, simple_loss=0.2984, pruned_loss=0.07585, over 15339.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.256, pruned_loss=0.04263, over 3329094.08 frames. ], batch size: 190, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:16,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9909, 4.4364, 4.4348, 3.2893, 3.6020, 4.3861, 3.9626, 2.5913], device='cuda:2'), covar=tensor([0.0444, 0.0053, 0.0045, 0.0317, 0.0129, 0.0095, 0.0077, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0083, 0.0083, 0.0134, 0.0097, 0.0109, 0.0094, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:2') 2023-05-01 01:31:19,979 INFO [train.py:904] (2/8) Epoch 20, batch 3300, loss[loss=0.1672, simple_loss=0.2573, pruned_loss=0.03859, over 17010.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2574, pruned_loss=0.04318, over 3324272.75 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:52,355 INFO [optim.py:368] (2/8) 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 01:31:59,240 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 01:32:20,544 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9254, 4.3708, 4.3840, 3.2476, 3.6335, 4.3387, 3.9689, 2.5119], device='cuda:2'), covar=tensor([0.0458, 0.0062, 0.0050, 0.0332, 0.0130, 0.0104, 0.0088, 0.0464], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0083, 0.0082, 0.0134, 0.0097, 0.0109, 0.0094, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:2') 2023-05-01 01:32:28,250 INFO [train.py:904] (2/8) Epoch 20, batch 3350, loss[loss=0.1324, simple_loss=0.2181, pruned_loss=0.02335, over 16785.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2588, pruned_loss=0.04352, over 3318709.88 frames. ], batch size: 39, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:33:04,051 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 01:33:24,694 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2488, 5.8738, 6.0073, 5.6884, 5.8557, 6.3711, 5.7691, 5.4846], device='cuda:2'), covar=tensor([0.0833, 0.1769, 0.2204, 0.2131, 0.2290, 0.0858, 0.1685, 0.2472], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0607, 0.0667, 0.0508, 0.0673, 0.0703, 0.0518, 0.0678], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 01:33:33,482 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3456, 2.3173, 2.3186, 4.2371, 2.2186, 2.7409, 2.3663, 2.4978], device='cuda:2'), covar=tensor([0.1331, 0.3546, 0.3081, 0.0514, 0.4267, 0.2584, 0.3702, 0.3485], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0442, 0.0366, 0.0329, 0.0437, 0.0511, 0.0413, 0.0519], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:33:35,776 INFO [train.py:904] (2/8) Epoch 20, batch 3400, loss[loss=0.1502, simple_loss=0.2429, pruned_loss=0.02875, over 17223.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2581, pruned_loss=0.04279, over 3320474.74 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:34:06,838 INFO [optim.py:368] (2/8) 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 01:34:26,072 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5591, 3.6425, 4.0396, 2.1488, 3.2200, 2.5312, 3.9840, 3.8102], device='cuda:2'), covar=tensor([0.0271, 0.0968, 0.0513, 0.2076, 0.0809, 0.1009, 0.0620, 0.1153], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:34:44,339 INFO [train.py:904] (2/8) Epoch 20, batch 3450, loss[loss=0.1602, simple_loss=0.2572, pruned_loss=0.03163, over 17120.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2568, pruned_loss=0.04226, over 3316192.33 frames. ], batch size: 49, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:35:10,269 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:35:50,044 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-05-01 01:35:50,467 INFO [train.py:904] (2/8) Epoch 20, batch 3500, loss[loss=0.1559, simple_loss=0.2438, pruned_loss=0.03402, over 16839.00 frames. ], tot_loss[loss=0.17, simple_loss=0.256, pruned_loss=0.04201, over 3307014.03 frames. ], batch size: 96, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:36:03,253 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196360.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:36:23,663 INFO [optim.py:368] (2/8) 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 01:36:35,813 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:37:01,667 INFO [train.py:904] (2/8) Epoch 20, batch 3550, loss[loss=0.1479, simple_loss=0.2329, pruned_loss=0.03148, over 16856.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2554, pruned_loss=0.04116, over 3318571.01 frames. ], batch size: 42, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:37:28,437 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:37:38,002 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-01 01:38:06,990 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4649, 5.8517, 5.4351, 5.7997, 5.3458, 5.0946, 5.4890, 5.9871], device='cuda:2'), covar=tensor([0.2429, 0.1588, 0.2342, 0.1491, 0.1619, 0.1423, 0.2275, 0.2074], device='cuda:2'), in_proj_covar=tensor([0.0679, 0.0840, 0.0683, 0.0626, 0.0526, 0.0529, 0.0700, 0.0642], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:38:10,214 INFO [train.py:904] (2/8) Epoch 20, batch 3600, loss[loss=0.1608, simple_loss=0.256, pruned_loss=0.03282, over 17119.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2539, pruned_loss=0.04046, over 3313254.98 frames. ], batch size: 49, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:38:41,979 INFO [optim.py:368] (2/8) 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 01:39:20,691 INFO [train.py:904] (2/8) Epoch 20, batch 3650, loss[loss=0.1584, simple_loss=0.2353, pruned_loss=0.04077, over 16341.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2526, pruned_loss=0.04103, over 3316882.90 frames. ], batch size: 165, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:39:40,450 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 01:40:32,649 INFO [train.py:904] (2/8) Epoch 20, batch 3700, loss[loss=0.1559, simple_loss=0.2339, pruned_loss=0.0389, over 16330.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2512, pruned_loss=0.04275, over 3290433.44 frames. ], batch size: 165, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:40:48,460 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 01:41:07,135 INFO [optim.py:368] (2/8) 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 01:41:11,787 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:41:47,079 INFO [train.py:904] (2/8) Epoch 20, batch 3750, loss[loss=0.1627, simple_loss=0.2365, pruned_loss=0.04439, over 16884.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2524, pruned_loss=0.04407, over 3277627.27 frames. ], batch size: 96, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:42:00,559 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2507, 2.2191, 2.3536, 4.0614, 2.2700, 2.5527, 2.2967, 2.4313], device='cuda:2'), covar=tensor([0.1427, 0.3537, 0.2659, 0.0524, 0.3624, 0.2495, 0.3753, 0.2819], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0444, 0.0366, 0.0329, 0.0436, 0.0513, 0.0413, 0.0520], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:42:21,420 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2154, 4.0982, 4.2683, 4.3925, 4.5019, 4.1066, 4.2795, 4.4999], device='cuda:2'), covar=tensor([0.1602, 0.1107, 0.1303, 0.0684, 0.0569, 0.1239, 0.1966, 0.0659], device='cuda:2'), in_proj_covar=tensor([0.0654, 0.0813, 0.0948, 0.0831, 0.0621, 0.0647, 0.0665, 0.0774], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:42:33,727 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 01:42:39,028 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:42:58,322 INFO [train.py:904] (2/8) Epoch 20, batch 3800, loss[loss=0.1704, simple_loss=0.2575, pruned_loss=0.04161, over 16713.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2533, pruned_loss=0.04534, over 3284294.82 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:43:31,148 INFO [optim.py:368] (2/8) 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 01:43:35,321 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:44:10,774 INFO [train.py:904] (2/8) Epoch 20, batch 3850, loss[loss=0.1907, simple_loss=0.2769, pruned_loss=0.05219, over 16699.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2541, pruned_loss=0.04585, over 3285774.83 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:44:32,369 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:45:05,444 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 01:45:24,169 INFO [train.py:904] (2/8) Epoch 20, batch 3900, loss[loss=0.1893, simple_loss=0.2607, pruned_loss=0.05894, over 16829.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2536, pruned_loss=0.04643, over 3277629.86 frames. ], batch size: 116, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:45:57,453 INFO [optim.py:368] (2/8) 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 01:46:04,943 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196779.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:46:19,609 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8741, 4.8014, 4.7544, 4.4485, 4.4457, 4.8523, 4.6796, 4.5833], device='cuda:2'), covar=tensor([0.0669, 0.0853, 0.0321, 0.0305, 0.0897, 0.0536, 0.0439, 0.0637], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0432, 0.0348, 0.0344, 0.0361, 0.0400, 0.0240, 0.0417], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:46:36,464 INFO [train.py:904] (2/8) Epoch 20, batch 3950, loss[loss=0.1917, simple_loss=0.2586, pruned_loss=0.06237, over 16885.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.253, pruned_loss=0.04682, over 3282446.83 frames. ], batch size: 116, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:16,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3485, 4.3054, 4.2548, 4.0106, 4.0228, 4.3421, 4.0411, 4.1321], device='cuda:2'), covar=tensor([0.0676, 0.0720, 0.0299, 0.0252, 0.0689, 0.0468, 0.0675, 0.0568], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0432, 0.0349, 0.0344, 0.0361, 0.0401, 0.0240, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:47:31,623 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196840.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:44,625 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7311, 3.7902, 2.3421, 4.1058, 2.9844, 4.1657, 2.4065, 3.0041], device='cuda:2'), covar=tensor([0.0239, 0.0327, 0.1550, 0.0251, 0.0637, 0.0542, 0.1522, 0.0691], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0178, 0.0195, 0.0164, 0.0177, 0.0219, 0.0203, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:47:44,726 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3667, 2.3281, 2.3133, 4.1468, 2.2570, 2.6979, 2.3403, 2.5067], device='cuda:2'), covar=tensor([0.1277, 0.3594, 0.2943, 0.0525, 0.3780, 0.2355, 0.3716, 0.3014], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0446, 0.0368, 0.0330, 0.0437, 0.0515, 0.0415, 0.0523], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:47:46,067 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2286, 2.1862, 2.2634, 3.9620, 2.2009, 2.5424, 2.2514, 2.3768], device='cuda:2'), covar=tensor([0.1406, 0.3779, 0.2968, 0.0572, 0.3865, 0.2621, 0.3815, 0.3056], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0445, 0.0368, 0.0330, 0.0437, 0.0515, 0.0415, 0.0523], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:47:48,960 INFO [train.py:904] (2/8) Epoch 20, batch 4000, loss[loss=0.1815, simple_loss=0.2613, pruned_loss=0.05088, over 16844.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2533, pruned_loss=0.04732, over 3286340.64 frames. ], batch size: 96, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:52,320 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:48:15,333 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5771, 4.8336, 4.6716, 4.6431, 4.3788, 4.3211, 4.3059, 4.9023], device='cuda:2'), covar=tensor([0.1159, 0.0848, 0.0918, 0.0803, 0.0797, 0.1189, 0.1101, 0.0832], device='cuda:2'), in_proj_covar=tensor([0.0670, 0.0824, 0.0675, 0.0618, 0.0520, 0.0524, 0.0693, 0.0636], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:48:21,944 INFO [optim.py:368] (2/8) 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 01:48:32,441 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:01,021 INFO [train.py:904] (2/8) Epoch 20, batch 4050, loss[loss=0.1653, simple_loss=0.2544, pruned_loss=0.0381, over 16294.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2541, pruned_loss=0.04681, over 3272325.44 frames. ], batch size: 165, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:49:14,858 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0988, 2.0681, 1.6845, 1.7467, 2.2916, 1.9424, 1.9801, 2.3469], device='cuda:2'), covar=tensor([0.0169, 0.0404, 0.0560, 0.0471, 0.0254, 0.0351, 0.0191, 0.0245], device='cuda:2'), in_proj_covar=tensor([0.0204, 0.0234, 0.0223, 0.0225, 0.0235, 0.0233, 0.0236, 0.0230], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:49:19,937 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:47,769 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:52,796 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9838, 3.8095, 3.7238, 2.2547, 3.4171, 3.7789, 3.4021, 2.0145], device='cuda:2'), covar=tensor([0.0615, 0.0060, 0.0056, 0.0473, 0.0105, 0.0113, 0.0106, 0.0488], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0083, 0.0082, 0.0133, 0.0096, 0.0107, 0.0093, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 01:49:57,981 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196942.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:50:12,375 INFO [train.py:904] (2/8) Epoch 20, batch 4100, loss[loss=0.1819, simple_loss=0.2683, pruned_loss=0.04776, over 16454.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2556, pruned_loss=0.0463, over 3263477.59 frames. ], batch size: 146, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:50:44,593 INFO [optim.py:368] (2/8) 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 01:50:47,883 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:51:17,396 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6890, 1.7937, 1.5719, 1.4074, 1.8916, 1.5273, 1.6065, 1.8669], device='cuda:2'), covar=tensor([0.0168, 0.0282, 0.0385, 0.0357, 0.0189, 0.0270, 0.0144, 0.0192], device='cuda:2'), in_proj_covar=tensor([0.0206, 0.0235, 0.0224, 0.0226, 0.0236, 0.0234, 0.0237, 0.0231], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:51:23,808 INFO [train.py:904] (2/8) Epoch 20, batch 4150, loss[loss=0.2529, simple_loss=0.3184, pruned_loss=0.09364, over 11600.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2626, pruned_loss=0.04902, over 3225352.30 frames. ], batch size: 246, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:51:45,576 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197016.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:00,053 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:40,716 INFO [train.py:904] (2/8) Epoch 20, batch 4200, loss[loss=0.2284, simple_loss=0.3171, pruned_loss=0.06983, over 16224.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2696, pruned_loss=0.05057, over 3209516.47 frames. ], batch size: 35, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:52:58,762 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:53:06,448 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 01:53:12,969 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 20, batch 4250, loss[loss=0.1741, simple_loss=0.2644, pruned_loss=0.04191, over 17112.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2731, pruned_loss=0.05026, over 3186307.15 frames. ], batch size: 47, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:54:39,853 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:55:04,298 INFO [train.py:904] (2/8) Epoch 20, batch 4300, loss[loss=0.2201, simple_loss=0.3001, pruned_loss=0.07008, over 11996.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2741, pruned_loss=0.04932, over 3174327.30 frames. ], batch size: 247, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:55:28,683 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3821, 2.8252, 3.0670, 1.8791, 2.6983, 2.0827, 2.9709, 3.0248], device='cuda:2'), covar=tensor([0.0291, 0.0856, 0.0553, 0.2046, 0.0883, 0.0959, 0.0657, 0.0954], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:55:37,988 INFO [optim.py:368] (2/8) 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 01:56:12,063 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7591, 3.7761, 2.5159, 4.6273, 2.9241, 4.4879, 2.6108, 3.1451], device='cuda:2'), covar=tensor([0.0267, 0.0355, 0.1385, 0.0109, 0.0744, 0.0402, 0.1277, 0.0666], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0160, 0.0175, 0.0215, 0.0200, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:56:18,918 INFO [train.py:904] (2/8) Epoch 20, batch 4350, loss[loss=0.1981, simple_loss=0.2903, pruned_loss=0.05293, over 16453.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2774, pruned_loss=0.05055, over 3189518.09 frames. ], batch size: 146, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:56:29,923 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:05,460 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:09,110 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:31,880 INFO [train.py:904] (2/8) Epoch 20, batch 4400, loss[loss=0.2045, simple_loss=0.2933, pruned_loss=0.05789, over 16928.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2792, pruned_loss=0.05139, over 3183208.88 frames. ], batch size: 109, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:57:32,416 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5722, 4.6381, 4.4460, 4.1604, 4.1703, 4.5685, 4.2723, 4.2316], device='cuda:2'), covar=tensor([0.0493, 0.0308, 0.0248, 0.0245, 0.0773, 0.0336, 0.0459, 0.0523], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0420, 0.0340, 0.0335, 0.0351, 0.0388, 0.0233, 0.0406], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:57:34,786 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6859, 2.9188, 2.9563, 4.8637, 3.9303, 4.2632, 1.5984, 3.0107], device='cuda:2'), covar=tensor([0.1332, 0.0745, 0.1053, 0.0117, 0.0322, 0.0340, 0.1613, 0.0840], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0185, 0.0205, 0.0212, 0.0198, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 01:58:04,301 INFO [optim.py:368] (2/8) 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 01:58:14,449 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:58:42,795 INFO [train.py:904] (2/8) Epoch 20, batch 4450, loss[loss=0.2102, simple_loss=0.3019, pruned_loss=0.0593, over 16417.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2825, pruned_loss=0.0525, over 3202361.53 frames. ], batch size: 146, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:59:17,315 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3580, 4.2008, 4.3118, 4.5025, 4.6209, 4.1932, 4.5939, 4.6587], device='cuda:2'), covar=tensor([0.1468, 0.1097, 0.1556, 0.0778, 0.0543, 0.1301, 0.0668, 0.0606], device='cuda:2'), in_proj_covar=tensor([0.0633, 0.0786, 0.0910, 0.0800, 0.0598, 0.0627, 0.0641, 0.0743], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 01:59:55,983 INFO [train.py:904] (2/8) Epoch 20, batch 4500, loss[loss=0.1811, simple_loss=0.2688, pruned_loss=0.04666, over 16274.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2838, pruned_loss=0.05344, over 3224603.95 frames. ], batch size: 35, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:00:30,608 INFO [optim.py:368] (2/8) 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 02:01:08,124 INFO [train.py:904] (2/8) Epoch 20, batch 4550, loss[loss=0.2364, simple_loss=0.3318, pruned_loss=0.07047, over 16632.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2846, pruned_loss=0.0544, over 3237292.77 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:01:27,217 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197415.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:01:39,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9791, 2.3361, 2.4162, 2.6321, 2.0753, 3.2316, 1.7991, 2.7679], device='cuda:2'), covar=tensor([0.1123, 0.0637, 0.0962, 0.0185, 0.0156, 0.0313, 0.1424, 0.0673], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0173, 0.0192, 0.0186, 0.0206, 0.0213, 0.0199, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:01:56,098 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197435.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:02:03,343 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3829, 4.2139, 4.0243, 2.7436, 3.7291, 4.1536, 3.6713, 2.3987], device='cuda:2'), covar=tensor([0.0552, 0.0028, 0.0042, 0.0386, 0.0076, 0.0076, 0.0080, 0.0430], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0082, 0.0081, 0.0133, 0.0096, 0.0107, 0.0094, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:02:18,308 INFO [train.py:904] (2/8) Epoch 20, batch 4600, loss[loss=0.1988, simple_loss=0.2844, pruned_loss=0.05658, over 16648.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2857, pruned_loss=0.05482, over 3217980.04 frames. ], batch size: 62, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:02:52,218 INFO [optim.py:368] (2/8) 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 02:02:52,779 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197476.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:03:02,412 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:30,139 INFO [train.py:904] (2/8) Epoch 20, batch 4650, loss[loss=0.1942, simple_loss=0.279, pruned_loss=0.05468, over 16707.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2849, pruned_loss=0.05485, over 3207677.01 frames. ], batch size: 124, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:03:40,039 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1609, 3.1324, 1.9583, 3.4765, 2.3501, 3.4931, 2.0449, 2.6085], device='cuda:2'), covar=tensor([0.0340, 0.0456, 0.1703, 0.0202, 0.0929, 0.0461, 0.1650, 0.0814], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0175, 0.0192, 0.0158, 0.0174, 0.0213, 0.0199, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:03:41,303 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:49,537 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197516.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:20,822 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:32,449 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197545.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:39,475 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4036, 4.5104, 4.6380, 4.4281, 4.4964, 5.0274, 4.5478, 4.2501], device='cuda:2'), covar=tensor([0.1396, 0.1690, 0.1874, 0.1972, 0.2474, 0.1075, 0.1539, 0.2553], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0577, 0.0631, 0.0479, 0.0639, 0.0671, 0.0493, 0.0644], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:04:42,155 INFO [train.py:904] (2/8) Epoch 20, batch 4700, loss[loss=0.1708, simple_loss=0.2586, pruned_loss=0.04157, over 16761.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2824, pruned_loss=0.05363, over 3215703.27 frames. ], batch size: 89, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:04:52,429 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:18,233 INFO [optim.py:368] (2/8) 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 02:05:19,875 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197577.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:05:30,679 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:55,395 INFO [train.py:904] (2/8) Epoch 20, batch 4750, loss[loss=0.1676, simple_loss=0.2507, pruned_loss=0.04224, over 17064.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2783, pruned_loss=0.05177, over 3209577.58 frames. ], batch size: 53, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:06:01,864 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:06:59,939 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7826, 2.7248, 2.5960, 1.8673, 2.5873, 2.6983, 2.5551, 1.8005], device='cuda:2'), covar=tensor([0.0480, 0.0087, 0.0075, 0.0371, 0.0118, 0.0132, 0.0125, 0.0421], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0082, 0.0081, 0.0132, 0.0095, 0.0106, 0.0093, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:07:08,982 INFO [train.py:904] (2/8) Epoch 20, batch 4800, loss[loss=0.1747, simple_loss=0.2614, pruned_loss=0.04397, over 17038.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2746, pruned_loss=0.04986, over 3195344.40 frames. ], batch size: 41, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:07:27,678 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 02:07:45,093 INFO [optim.py:368] (2/8) 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 02:08:24,127 INFO [train.py:904] (2/8) Epoch 20, batch 4850, loss[loss=0.1715, simple_loss=0.2674, pruned_loss=0.03777, over 16389.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2748, pruned_loss=0.04936, over 3183385.08 frames. ], batch size: 146, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:08:46,264 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-05-01 02:08:57,111 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0576, 3.9382, 4.1339, 4.2570, 4.3869, 3.9114, 4.2911, 4.3974], device='cuda:2'), covar=tensor([0.1467, 0.1152, 0.1245, 0.0664, 0.0472, 0.1422, 0.0724, 0.0655], device='cuda:2'), in_proj_covar=tensor([0.0620, 0.0767, 0.0891, 0.0782, 0.0585, 0.0613, 0.0628, 0.0727], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:09:40,078 INFO [train.py:904] (2/8) Epoch 20, batch 4900, loss[loss=0.1737, simple_loss=0.2706, pruned_loss=0.03839, over 15427.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.274, pruned_loss=0.04816, over 3156327.13 frames. ], batch size: 191, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:09:55,921 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0359, 2.0698, 1.7653, 1.7921, 2.2929, 2.0028, 1.9124, 2.3746], device='cuda:2'), covar=tensor([0.0183, 0.0404, 0.0528, 0.0462, 0.0229, 0.0353, 0.0192, 0.0274], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0229, 0.0222, 0.0222, 0.0232, 0.0231, 0.0232, 0.0228], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:10:08,659 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197771.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:10:15,934 INFO [optim.py:368] (2/8) 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 02:10:48,289 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8038, 4.0853, 3.7608, 3.5487, 3.2309, 3.9760, 3.6073, 3.6208], device='cuda:2'), covar=tensor([0.0838, 0.0548, 0.0444, 0.0386, 0.1424, 0.0493, 0.1473, 0.0687], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0409, 0.0331, 0.0327, 0.0342, 0.0378, 0.0228, 0.0395], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:10:52,823 INFO [train.py:904] (2/8) Epoch 20, batch 4950, loss[loss=0.1962, simple_loss=0.2863, pruned_loss=0.05309, over 16461.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2734, pruned_loss=0.04728, over 3181182.78 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:04,524 INFO [train.py:904] (2/8) Epoch 20, batch 5000, loss[loss=0.1859, simple_loss=0.2828, pruned_loss=0.04448, over 16748.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2743, pruned_loss=0.04702, over 3192313.70 frames. ], batch size: 89, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:32,783 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197872.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:12:38,856 INFO [optim.py:368] (2/8) 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 02:13:01,287 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197892.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:13,653 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197901.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:14,485 INFO [train.py:904] (2/8) Epoch 20, batch 5050, loss[loss=0.1844, simple_loss=0.2785, pruned_loss=0.04516, over 16873.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2746, pruned_loss=0.04674, over 3196710.99 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:13:35,054 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3358, 4.2154, 4.3829, 4.5507, 4.7371, 4.3079, 4.6932, 4.7346], device='cuda:2'), covar=tensor([0.1827, 0.1176, 0.1517, 0.0740, 0.0448, 0.1090, 0.0544, 0.0562], device='cuda:2'), in_proj_covar=tensor([0.0625, 0.0773, 0.0900, 0.0790, 0.0588, 0.0618, 0.0633, 0.0732], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:13:38,920 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 02:14:25,041 INFO [train.py:904] (2/8) Epoch 20, batch 5100, loss[loss=0.1712, simple_loss=0.2605, pruned_loss=0.04095, over 16454.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2734, pruned_loss=0.04635, over 3187827.27 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:14:27,271 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197953.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:15:00,607 INFO [optim.py:368] (2/8) 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 02:15:41,431 INFO [train.py:904] (2/8) Epoch 20, batch 5150, loss[loss=0.1834, simple_loss=0.2824, pruned_loss=0.04221, over 16485.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2734, pruned_loss=0.0464, over 3153356.76 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:16:52,361 INFO [train.py:904] (2/8) Epoch 20, batch 5200, loss[loss=0.1667, simple_loss=0.2578, pruned_loss=0.03781, over 16536.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2721, pruned_loss=0.04575, over 3158554.51 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:17:06,544 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-05-01 02:17:18,569 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8253, 4.9145, 5.2547, 5.1907, 5.2110, 4.9075, 4.8370, 4.6643], device='cuda:2'), covar=tensor([0.0296, 0.0492, 0.0303, 0.0373, 0.0499, 0.0332, 0.0958, 0.0421], device='cuda:2'), in_proj_covar=tensor([0.0389, 0.0431, 0.0415, 0.0388, 0.0461, 0.0439, 0.0528, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 02:17:19,631 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198071.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:17:27,749 INFO [optim.py:368] (2/8) 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 02:17:34,778 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198081.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:18:03,837 INFO [train.py:904] (2/8) Epoch 20, batch 5250, loss[loss=0.1681, simple_loss=0.2598, pruned_loss=0.03816, over 16538.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2697, pruned_loss=0.04525, over 3171335.96 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:18:28,863 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198119.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:18:48,782 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 02:19:01,611 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:19:15,056 INFO [train.py:904] (2/8) Epoch 20, batch 5300, loss[loss=0.1668, simple_loss=0.2555, pruned_loss=0.03898, over 15281.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2654, pruned_loss=0.04359, over 3191740.63 frames. ], batch size: 190, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:19:32,122 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-05-01 02:19:45,039 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198172.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:19:49,806 INFO [optim.py:368] (2/8) 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 02:20:26,914 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198201.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:28,367 INFO [train.py:904] (2/8) Epoch 20, batch 5350, loss[loss=0.1639, simple_loss=0.2622, pruned_loss=0.03275, over 16883.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2641, pruned_loss=0.04291, over 3212930.59 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:20:54,787 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198220.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:59,146 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3463, 4.5997, 4.8050, 4.7725, 4.7637, 4.4939, 4.1606, 4.2684], device='cuda:2'), covar=tensor([0.0545, 0.0714, 0.0543, 0.0637, 0.0858, 0.0580, 0.1604, 0.0605], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0434, 0.0418, 0.0390, 0.0465, 0.0442, 0.0531, 0.0354], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 02:21:35,323 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198249.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:40,981 INFO [train.py:904] (2/8) Epoch 20, batch 5400, loss[loss=0.1965, simple_loss=0.2915, pruned_loss=0.05079, over 16896.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2667, pruned_loss=0.04358, over 3209218.52 frames. ], batch size: 109, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:22:15,985 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 20, batch 5450, loss[loss=0.2081, simple_loss=0.2912, pruned_loss=0.06254, over 16872.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2698, pruned_loss=0.04512, over 3191205.85 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:23:04,397 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6341, 4.0831, 3.1385, 2.3653, 2.7800, 2.6171, 4.4041, 3.6614], device='cuda:2'), covar=tensor([0.2875, 0.0602, 0.1601, 0.2531, 0.2625, 0.1835, 0.0396, 0.1031], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0268, 0.0302, 0.0306, 0.0294, 0.0252, 0.0292, 0.0331], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:23:12,771 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 02:23:18,428 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2669, 3.4261, 3.5730, 3.5533, 3.5644, 3.3960, 3.4056, 3.4745], device='cuda:2'), covar=tensor([0.0430, 0.0839, 0.0537, 0.0456, 0.0548, 0.0771, 0.0962, 0.0610], device='cuda:2'), in_proj_covar=tensor([0.0390, 0.0432, 0.0416, 0.0388, 0.0463, 0.0441, 0.0529, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 02:23:37,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6366, 3.6055, 4.2764, 1.9621, 4.4996, 4.5161, 3.2531, 3.1664], device='cuda:2'), covar=tensor([0.0833, 0.0301, 0.0155, 0.1236, 0.0051, 0.0105, 0.0389, 0.0459], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0110, 0.0098, 0.0140, 0.0080, 0.0124, 0.0129, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:24:01,047 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8898, 3.7546, 3.8209, 4.0805, 4.1340, 3.8230, 4.1675, 4.1853], device='cuda:2'), covar=tensor([0.1654, 0.1386, 0.1856, 0.0859, 0.0845, 0.1730, 0.0905, 0.0963], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0767, 0.0893, 0.0779, 0.0586, 0.0614, 0.0626, 0.0726], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:24:14,759 INFO [train.py:904] (2/8) Epoch 20, batch 5500, loss[loss=0.2403, simple_loss=0.3152, pruned_loss=0.08274, over 11910.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2765, pruned_loss=0.04903, over 3173730.89 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:24:38,081 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9886, 4.9943, 4.8568, 4.5142, 4.5110, 4.9169, 4.7748, 4.6168], device='cuda:2'), covar=tensor([0.0613, 0.0550, 0.0284, 0.0297, 0.0933, 0.0518, 0.0391, 0.0598], device='cuda:2'), in_proj_covar=tensor([0.0290, 0.0421, 0.0339, 0.0334, 0.0350, 0.0390, 0.0233, 0.0405], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:24:51,695 INFO [optim.py:368] (2/8) 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 02:25:34,175 INFO [train.py:904] (2/8) Epoch 20, batch 5550, loss[loss=0.1947, simple_loss=0.2849, pruned_loss=0.05227, over 17023.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.283, pruned_loss=0.05365, over 3152385.69 frames. ], batch size: 41, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:25:51,829 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 02:26:30,782 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198437.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:26:30,908 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4004, 4.4109, 4.2665, 3.5292, 4.3400, 1.7147, 4.0561, 3.9047], device='cuda:2'), covar=tensor([0.0100, 0.0089, 0.0183, 0.0331, 0.0093, 0.2752, 0.0138, 0.0261], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0148, 0.0193, 0.0175, 0.0169, 0.0203, 0.0183, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:26:37,786 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7435, 3.7714, 3.8748, 3.6470, 3.8541, 4.2372, 3.8701, 3.6071], device='cuda:2'), covar=tensor([0.2281, 0.2354, 0.2442, 0.2384, 0.2361, 0.1777, 0.1762, 0.2569], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0572, 0.0624, 0.0475, 0.0633, 0.0663, 0.0491, 0.0640], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:26:45,761 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 02:26:53,777 INFO [train.py:904] (2/8) Epoch 20, batch 5600, loss[loss=0.2122, simple_loss=0.3036, pruned_loss=0.06037, over 16718.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2882, pruned_loss=0.05842, over 3110014.52 frames. ], batch size: 89, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:27:11,843 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4284, 1.6877, 2.0876, 2.3298, 2.4499, 2.7228, 1.7961, 2.6096], device='cuda:2'), covar=tensor([0.0195, 0.0501, 0.0292, 0.0315, 0.0303, 0.0181, 0.0498, 0.0141], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0190, 0.0175, 0.0180, 0.0191, 0.0149, 0.0191, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:27:26,815 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-01 02:27:34,779 INFO [optim.py:368] (2/8) 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 02:28:14,129 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198499.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:28:18,605 INFO [train.py:904] (2/8) Epoch 20, batch 5650, loss[loss=0.2717, simple_loss=0.3319, pruned_loss=0.1058, over 11521.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2934, pruned_loss=0.06248, over 3068885.08 frames. ], batch size: 247, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:13,816 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 02:29:32,484 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198548.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:29:33,841 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6000, 2.1890, 1.8770, 2.0206, 2.5137, 2.1696, 2.3573, 2.6770], device='cuda:2'), covar=tensor([0.0188, 0.0402, 0.0518, 0.0460, 0.0246, 0.0381, 0.0217, 0.0258], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0229, 0.0222, 0.0221, 0.0232, 0.0230, 0.0231, 0.0227], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:29:36,982 INFO [train.py:904] (2/8) Epoch 20, batch 5700, loss[loss=0.199, simple_loss=0.3017, pruned_loss=0.04815, over 16710.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2947, pruned_loss=0.06316, over 3083856.83 frames. ], batch size: 89, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:46,738 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6091, 3.6655, 2.1910, 4.2941, 2.7720, 4.1802, 2.3424, 2.8649], device='cuda:2'), covar=tensor([0.0320, 0.0415, 0.1768, 0.0155, 0.0871, 0.0495, 0.1596, 0.0840], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0157, 0.0175, 0.0213, 0.0200, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:29:50,279 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:14,597 INFO [optim.py:368] (2/8) 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 02:30:21,723 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3240, 3.1009, 3.3738, 1.7609, 3.5834, 3.6040, 2.8084, 2.6899], device='cuda:2'), covar=tensor([0.0845, 0.0291, 0.0247, 0.1253, 0.0085, 0.0180, 0.0509, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0140, 0.0080, 0.0124, 0.0128, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:30:47,286 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:55,555 INFO [train.py:904] (2/8) Epoch 20, batch 5750, loss[loss=0.225, simple_loss=0.3073, pruned_loss=0.07134, over 16849.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2984, pruned_loss=0.06529, over 3053362.38 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:31:02,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7166, 3.6829, 2.2383, 4.3711, 2.8901, 4.2519, 2.4603, 3.0047], device='cuda:2'), covar=tensor([0.0295, 0.0449, 0.1737, 0.0181, 0.0852, 0.0500, 0.1550, 0.0800], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0157, 0.0175, 0.0213, 0.0200, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:31:08,899 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 02:31:17,261 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:32:08,120 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 02:32:16,951 INFO [train.py:904] (2/8) Epoch 20, batch 5800, loss[loss=0.1793, simple_loss=0.2696, pruned_loss=0.04453, over 16440.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2982, pruned_loss=0.06456, over 3035149.96 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:32:53,794 INFO [optim.py:368] (2/8) 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 02:32:54,384 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:33:34,961 INFO [train.py:904] (2/8) Epoch 20, batch 5850, loss[loss=0.2091, simple_loss=0.2794, pruned_loss=0.06941, over 11600.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2958, pruned_loss=0.0627, over 3038696.57 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:34:05,321 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9472, 5.4406, 5.6078, 5.3385, 5.3800, 5.9912, 5.4364, 5.2279], device='cuda:2'), covar=tensor([0.1053, 0.1769, 0.1895, 0.1894, 0.2446, 0.0920, 0.1604, 0.2366], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0581, 0.0635, 0.0481, 0.0641, 0.0670, 0.0498, 0.0649], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:34:27,436 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5667, 4.5872, 4.9556, 4.9292, 4.9486, 4.6352, 4.6197, 4.4641], device='cuda:2'), covar=tensor([0.0355, 0.0610, 0.0406, 0.0423, 0.0465, 0.0420, 0.0918, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0442, 0.0426, 0.0396, 0.0474, 0.0449, 0.0538, 0.0360], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 02:34:30,885 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198737.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:34:53,519 INFO [train.py:904] (2/8) Epoch 20, batch 5900, loss[loss=0.1872, simple_loss=0.2729, pruned_loss=0.0507, over 15463.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.296, pruned_loss=0.06312, over 3042344.51 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:35:25,020 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9440, 4.7227, 4.9646, 5.1537, 5.3876, 4.7457, 5.3471, 5.3703], device='cuda:2'), covar=tensor([0.2102, 0.1451, 0.1857, 0.0857, 0.0661, 0.1008, 0.0776, 0.0669], device='cuda:2'), in_proj_covar=tensor([0.0619, 0.0763, 0.0886, 0.0775, 0.0584, 0.0611, 0.0626, 0.0721], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:35:26,417 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4479, 2.9075, 2.7105, 2.3540, 2.3375, 2.3093, 2.8318, 2.9352], device='cuda:2'), covar=tensor([0.2144, 0.0615, 0.1429, 0.2211, 0.2065, 0.1957, 0.0458, 0.1108], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0267, 0.0302, 0.0305, 0.0294, 0.0252, 0.0292, 0.0331], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:35:27,507 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3973, 3.3707, 1.8610, 3.7359, 2.5183, 3.7083, 2.1725, 2.7727], device='cuda:2'), covar=tensor([0.0293, 0.0411, 0.1883, 0.0331, 0.0918, 0.0638, 0.1597, 0.0849], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0158, 0.0175, 0.0213, 0.0201, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:35:34,248 INFO [optim.py:368] (2/8) 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 02:35:48,536 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:36:14,567 INFO [train.py:904] (2/8) Epoch 20, batch 5950, loss[loss=0.2121, simple_loss=0.2979, pruned_loss=0.06315, over 16735.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2963, pruned_loss=0.0614, over 3062156.79 frames. ], batch size: 124, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:02,256 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6824, 4.8781, 4.9867, 4.8188, 4.8716, 5.4143, 4.8763, 4.6361], device='cuda:2'), covar=tensor([0.1200, 0.1940, 0.2223, 0.2049, 0.2428, 0.0979, 0.1753, 0.2535], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0584, 0.0639, 0.0484, 0.0644, 0.0673, 0.0502, 0.0653], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:37:31,029 INFO [train.py:904] (2/8) Epoch 20, batch 6000, loss[loss=0.1958, simple_loss=0.2771, pruned_loss=0.05727, over 16662.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2948, pruned_loss=0.06055, over 3081684.31 frames. ], batch size: 57, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:31,029 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 02:37:41,824 INFO [train.py:938] (2/8) Epoch 20, validation: loss=0.1516, simple_loss=0.2644, pruned_loss=0.01942, over 944034.00 frames. 2023-05-01 02:37:41,824 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 02:37:42,250 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:46,504 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198855.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:38:09,261 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0482, 4.0337, 3.9725, 3.2569, 4.0132, 1.7085, 3.7765, 3.5141], device='cuda:2'), covar=tensor([0.0137, 0.0109, 0.0193, 0.0339, 0.0102, 0.2851, 0.0155, 0.0288], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0174, 0.0168, 0.0202, 0.0182, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:38:17,161 INFO [optim.py:368] (2/8) 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 02:38:28,024 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:38:30,826 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 02:38:56,337 INFO [train.py:904] (2/8) Epoch 20, batch 6050, loss[loss=0.2002, simple_loss=0.2883, pruned_loss=0.05608, over 15356.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2939, pruned_loss=0.06045, over 3078973.86 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:39:14,622 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198913.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:36,221 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198927.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:05,086 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198944.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:16,850 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9554, 2.4060, 1.9606, 2.1923, 2.7564, 2.4097, 2.7038, 2.9075], device='cuda:2'), covar=tensor([0.0157, 0.0379, 0.0516, 0.0417, 0.0240, 0.0345, 0.0201, 0.0218], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0228, 0.0220, 0.0221, 0.0231, 0.0228, 0.0229, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:40:17,596 INFO [train.py:904] (2/8) Epoch 20, batch 6100, loss[loss=0.1899, simple_loss=0.2835, pruned_loss=0.0481, over 16828.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2932, pruned_loss=0.05973, over 3082725.92 frames. ], batch size: 102, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:40:46,239 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:53,272 INFO [optim.py:368] (2/8) 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 02:41:12,463 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:41:32,341 INFO [train.py:904] (2/8) Epoch 20, batch 6150, loss[loss=0.19, simple_loss=0.2774, pruned_loss=0.05131, over 16891.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2917, pruned_loss=0.05955, over 3085686.53 frames. ], batch size: 116, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:42:49,581 INFO [train.py:904] (2/8) Epoch 20, batch 6200, loss[loss=0.1766, simple_loss=0.2722, pruned_loss=0.04047, over 16699.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2897, pruned_loss=0.05857, over 3095558.44 frames. ], batch size: 89, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:43:19,270 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8066, 2.7302, 2.5883, 1.9266, 2.6113, 2.6794, 2.5723, 1.9073], device='cuda:2'), covar=tensor([0.0443, 0.0079, 0.0081, 0.0351, 0.0118, 0.0134, 0.0116, 0.0394], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0132, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:43:19,542 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 02:43:22,218 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9275, 5.0116, 5.4298, 5.3646, 5.3933, 5.0441, 4.9885, 4.7670], device='cuda:2'), covar=tensor([0.0350, 0.0572, 0.0312, 0.0426, 0.0456, 0.0350, 0.0973, 0.0459], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0442, 0.0426, 0.0398, 0.0476, 0.0451, 0.0541, 0.0362], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 02:43:28,024 INFO [optim.py:368] (2/8) 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 02:44:06,526 INFO [train.py:904] (2/8) Epoch 20, batch 6250, loss[loss=0.1963, simple_loss=0.286, pruned_loss=0.05332, over 15454.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2888, pruned_loss=0.05797, over 3094308.64 frames. ], batch size: 191, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:44:56,041 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2965, 3.7159, 3.8250, 2.2553, 3.2591, 2.4244, 3.8924, 3.9339], device='cuda:2'), covar=tensor([0.0229, 0.0707, 0.0509, 0.1963, 0.0739, 0.0969, 0.0510, 0.0894], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0161, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 02:45:16,569 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 02:45:21,382 INFO [train.py:904] (2/8) Epoch 20, batch 6300, loss[loss=0.1908, simple_loss=0.2783, pruned_loss=0.05164, over 15371.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2889, pruned_loss=0.05758, over 3096447.16 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:26,461 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199155.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:45:31,560 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-05-01 02:45:39,550 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4093, 1.6254, 2.0847, 2.3191, 2.3623, 2.5838, 1.7884, 2.5901], device='cuda:2'), covar=tensor([0.0220, 0.0529, 0.0306, 0.0350, 0.0329, 0.0229, 0.0548, 0.0132], device='cuda:2'), in_proj_covar=tensor([0.0183, 0.0191, 0.0176, 0.0180, 0.0192, 0.0150, 0.0193, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:45:43,187 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8905, 5.2896, 5.4338, 5.1900, 5.2576, 5.8056, 5.2715, 5.0646], device='cuda:2'), covar=tensor([0.0951, 0.1699, 0.2318, 0.1910, 0.2499, 0.0957, 0.1571, 0.2341], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0584, 0.0642, 0.0485, 0.0644, 0.0673, 0.0501, 0.0651], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:45:59,376 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:00,084 INFO [optim.py:368] (2/8) 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 02:46:03,032 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7810, 6.1061, 5.7994, 5.8817, 5.4113, 5.3371, 5.5487, 6.1907], device='cuda:2'), covar=tensor([0.1145, 0.0738, 0.0949, 0.0766, 0.0837, 0.0654, 0.1139, 0.0853], device='cuda:2'), in_proj_covar=tensor([0.0641, 0.0780, 0.0646, 0.0590, 0.0493, 0.0504, 0.0657, 0.0605], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:46:03,483 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 02:46:09,895 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 02:46:38,965 INFO [train.py:904] (2/8) Epoch 20, batch 6350, loss[loss=0.2829, simple_loss=0.3345, pruned_loss=0.1156, over 11747.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2892, pruned_loss=0.0585, over 3086381.63 frames. ], batch size: 247, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:46:40,592 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:46,155 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 02:46:48,061 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199208.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:30,807 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199237.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:47:34,237 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199239.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:41,388 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-01 02:47:52,850 INFO [train.py:904] (2/8) Epoch 20, batch 6400, loss[loss=0.2041, simple_loss=0.2807, pruned_loss=0.06376, over 16471.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2892, pruned_loss=0.05949, over 3100418.99 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:48:09,794 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 02:48:21,692 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199271.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:48:29,164 INFO [optim.py:368] (2/8) 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 02:48:39,449 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199283.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:49:07,103 INFO [train.py:904] (2/8) Epoch 20, batch 6450, loss[loss=0.2057, simple_loss=0.2934, pruned_loss=0.05895, over 16374.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2889, pruned_loss=0.05862, over 3109137.35 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:49:33,336 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:49:59,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7379, 1.7605, 1.6338, 1.5171, 1.9173, 1.5666, 1.6674, 1.8803], device='cuda:2'), covar=tensor([0.0239, 0.0288, 0.0372, 0.0325, 0.0224, 0.0257, 0.0247, 0.0229], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0228, 0.0221, 0.0221, 0.0231, 0.0228, 0.0229, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 02:50:24,669 INFO [train.py:904] (2/8) Epoch 20, batch 6500, loss[loss=0.2004, simple_loss=0.2904, pruned_loss=0.05517, over 16773.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2873, pruned_loss=0.05812, over 3104778.66 frames. ], batch size: 124, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:51:02,007 INFO [optim.py:368] (2/8) 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 02:51:39,117 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:51:41,758 INFO [train.py:904] (2/8) Epoch 20, batch 6550, loss[loss=0.2082, simple_loss=0.3061, pruned_loss=0.05519, over 16684.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2905, pruned_loss=0.05921, over 3113451.89 frames. ], batch size: 57, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:52:15,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3635, 2.9202, 2.6485, 2.2586, 2.2139, 2.2375, 2.8710, 2.8209], device='cuda:2'), covar=tensor([0.2433, 0.0739, 0.1660, 0.2645, 0.2400, 0.2106, 0.0544, 0.1397], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0267, 0.0302, 0.0307, 0.0294, 0.0253, 0.0292, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 02:52:56,160 INFO [train.py:904] (2/8) Epoch 20, batch 6600, loss[loss=0.2331, simple_loss=0.2998, pruned_loss=0.08321, over 11471.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2928, pruned_loss=0.05999, over 3093729.42 frames. ], batch size: 249, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:53:05,863 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:53:10,077 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199461.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:53:33,375 INFO [optim.py:368] (2/8) 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 02:53:59,240 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199494.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:54:11,565 INFO [train.py:904] (2/8) Epoch 20, batch 6650, loss[loss=0.2075, simple_loss=0.2907, pruned_loss=0.06217, over 16716.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2931, pruned_loss=0.06097, over 3069528.79 frames. ], batch size: 39, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:54:20,056 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199508.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:54:25,411 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 02:54:37,226 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199519.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:54:56,458 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199532.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:55:00,513 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199535.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:55:06,218 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199539.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:55:25,178 INFO [train.py:904] (2/8) Epoch 20, batch 6700, loss[loss=0.224, simple_loss=0.2966, pruned_loss=0.07571, over 11264.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2919, pruned_loss=0.06099, over 3074709.01 frames. ], batch size: 246, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:55:30,690 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199555.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:55:32,560 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199556.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:55:37,595 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-05-01 02:56:02,760 INFO [optim.py:368] (2/8) 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 02:56:13,403 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:18,211 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:31,488 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199596.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:56:38,605 INFO [train.py:904] (2/8) Epoch 20, batch 6750, loss[loss=0.2024, simple_loss=0.2914, pruned_loss=0.05666, over 16456.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2914, pruned_loss=0.06116, over 3090216.92 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:57:20,693 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:57:53,090 INFO [train.py:904] (2/8) Epoch 20, batch 6800, loss[loss=0.2399, simple_loss=0.3124, pruned_loss=0.08369, over 11632.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2917, pruned_loss=0.06085, over 3105857.87 frames. ], batch size: 248, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:58:29,264 INFO [optim.py:368] (2/8) 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 02:59:05,436 INFO [train.py:904] (2/8) Epoch 20, batch 6850, loss[loss=0.1934, simple_loss=0.2903, pruned_loss=0.04819, over 16764.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2925, pruned_loss=0.0611, over 3098575.07 frames. ], batch size: 124, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:59:56,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0094, 2.0874, 2.1068, 3.5655, 2.0872, 2.4750, 2.1811, 2.2455], device='cuda:2'), covar=tensor([0.1388, 0.3656, 0.3032, 0.0586, 0.3984, 0.2398, 0.3605, 0.3178], device='cuda:2'), in_proj_covar=tensor([0.0393, 0.0435, 0.0359, 0.0320, 0.0430, 0.0502, 0.0406, 0.0511], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:00:20,622 INFO [train.py:904] (2/8) Epoch 20, batch 6900, loss[loss=0.1827, simple_loss=0.2837, pruned_loss=0.04083, over 17118.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2951, pruned_loss=0.06093, over 3094712.40 frames. ], batch size: 49, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:00:26,875 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199756.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:00:59,795 INFO [optim.py:368] (2/8) 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 03:01:07,635 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 03:01:17,002 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-01 03:01:36,157 INFO [train.py:904] (2/8) Epoch 20, batch 6950, loss[loss=0.2366, simple_loss=0.3211, pruned_loss=0.07608, over 15191.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2965, pruned_loss=0.06235, over 3093565.90 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:01:54,385 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199814.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:02:22,490 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:02:47,651 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199850.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:02:49,629 INFO [train.py:904] (2/8) Epoch 20, batch 7000, loss[loss=0.1843, simple_loss=0.2867, pruned_loss=0.04096, over 16485.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2965, pruned_loss=0.0621, over 3062520.65 frames. ], batch size: 75, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:03:10,545 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 03:03:29,995 INFO [optim.py:368] (2/8) 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 03:03:33,438 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:03:49,865 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199891.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:03:55,641 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7693, 1.3249, 1.6988, 1.6328, 1.7664, 1.8988, 1.5897, 1.8099], device='cuda:2'), covar=tensor([0.0239, 0.0378, 0.0202, 0.0276, 0.0251, 0.0168, 0.0384, 0.0138], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0192, 0.0175, 0.0180, 0.0192, 0.0150, 0.0192, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:04:07,200 INFO [train.py:904] (2/8) Epoch 20, batch 7050, loss[loss=0.2192, simple_loss=0.3007, pruned_loss=0.06879, over 15354.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2976, pruned_loss=0.06202, over 3071061.94 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:04:17,856 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:05:18,815 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0709, 3.6196, 3.5756, 2.3229, 3.3172, 3.5949, 3.2985, 2.0661], device='cuda:2'), covar=tensor([0.0571, 0.0055, 0.0052, 0.0413, 0.0102, 0.0109, 0.0105, 0.0437], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0133, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 03:05:25,773 INFO [train.py:904] (2/8) Epoch 20, batch 7100, loss[loss=0.2037, simple_loss=0.2833, pruned_loss=0.06206, over 16800.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2962, pruned_loss=0.06194, over 3058291.25 frames. ], batch size: 124, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:05:54,787 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 03:06:45,331 INFO [train.py:904] (2/8) Epoch 20, batch 7150, loss[loss=0.2074, simple_loss=0.2939, pruned_loss=0.06051, over 15193.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2943, pruned_loss=0.06202, over 3051616.83 frames. ], batch size: 190, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:07:59,654 INFO [train.py:904] (2/8) Epoch 20, batch 7200, loss[loss=0.1885, simple_loss=0.2954, pruned_loss=0.04075, over 16558.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2916, pruned_loss=0.05997, over 3050834.12 frames. ], batch size: 75, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:08:06,669 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200056.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:08:25,535 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 03:08:37,902 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6738, 3.1756, 3.3188, 2.1440, 2.8770, 2.0549, 3.3353, 3.3999], device='cuda:2'), covar=tensor([0.0236, 0.0734, 0.0555, 0.1848, 0.0811, 0.1053, 0.0581, 0.0745], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0161, 0.0165, 0.0151, 0.0143, 0.0129, 0.0143, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 03:08:40,489 INFO [optim.py:368] (2/8) 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 03:08:43,225 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 03:09:19,963 INFO [train.py:904] (2/8) Epoch 20, batch 7250, loss[loss=0.1903, simple_loss=0.2744, pruned_loss=0.05313, over 16710.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2887, pruned_loss=0.05799, over 3078275.84 frames. ], batch size: 134, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:09:23,330 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200104.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:09:37,484 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:10:31,948 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 20, batch 7300, loss[loss=0.1931, simple_loss=0.2911, pruned_loss=0.0475, over 16920.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2879, pruned_loss=0.05766, over 3077218.48 frames. ], batch size: 96, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:10:50,981 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200162.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:11:14,597 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200177.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:11:15,880 INFO [optim.py:368] (2/8) 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 03:11:36,995 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200191.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:11:40,888 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.14 vs. limit=5.0 2023-05-01 03:11:47,182 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200198.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:11:52,586 INFO [train.py:904] (2/8) Epoch 20, batch 7350, loss[loss=0.2259, simple_loss=0.3176, pruned_loss=0.06707, over 16511.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.29, pruned_loss=0.05972, over 3034943.31 frames. ], batch size: 146, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:12:51,305 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200238.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:12:52,207 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200239.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:13:12,410 INFO [train.py:904] (2/8) Epoch 20, batch 7400, loss[loss=0.2008, simple_loss=0.3008, pruned_loss=0.05043, over 16741.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2908, pruned_loss=0.06004, over 3035712.78 frames. ], batch size: 102, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:13:33,123 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:13:52,749 INFO [optim.py:368] (2/8) 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 03:13:55,472 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:14:32,167 INFO [train.py:904] (2/8) Epoch 20, batch 7450, loss[loss=0.2014, simple_loss=0.2963, pruned_loss=0.05324, over 16282.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2922, pruned_loss=0.06121, over 3040885.93 frames. ], batch size: 146, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:15:19,988 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8531, 2.5435, 2.5947, 1.8501, 2.5457, 2.6471, 2.6011, 1.9311], device='cuda:2'), covar=tensor([0.0454, 0.0133, 0.0086, 0.0352, 0.0139, 0.0146, 0.0116, 0.0420], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0133, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 03:15:36,238 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200340.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:15:54,169 INFO [train.py:904] (2/8) Epoch 20, batch 7500, loss[loss=0.2217, simple_loss=0.2977, pruned_loss=0.07282, over 11181.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2922, pruned_loss=0.06047, over 3034769.06 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:15:55,924 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0263, 5.3038, 5.0342, 5.0528, 4.8129, 4.7707, 4.7400, 5.3962], device='cuda:2'), covar=tensor([0.1186, 0.0879, 0.1062, 0.0934, 0.0798, 0.0954, 0.1200, 0.0864], device='cuda:2'), in_proj_covar=tensor([0.0643, 0.0789, 0.0651, 0.0593, 0.0497, 0.0509, 0.0661, 0.0610], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:16:02,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8970, 4.1334, 3.9241, 3.9996, 3.6830, 3.7848, 3.8084, 4.1182], device='cuda:2'), covar=tensor([0.1134, 0.0899, 0.1075, 0.0884, 0.0821, 0.1645, 0.0969, 0.1025], device='cuda:2'), in_proj_covar=tensor([0.0643, 0.0789, 0.0651, 0.0593, 0.0496, 0.0509, 0.0661, 0.0609], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:16:19,191 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1699, 4.2150, 4.0475, 3.7320, 3.7626, 4.1796, 3.8815, 3.8885], device='cuda:2'), covar=tensor([0.0638, 0.0521, 0.0294, 0.0333, 0.0728, 0.0434, 0.0765, 0.0644], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0402, 0.0323, 0.0321, 0.0333, 0.0373, 0.0224, 0.0388], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:16:34,101 INFO [optim.py:368] (2/8) 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 03:17:03,869 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3219, 3.1975, 3.1594, 3.4263, 3.4252, 3.2674, 3.3834, 3.4513], device='cuda:2'), covar=tensor([0.1510, 0.1499, 0.1778, 0.0983, 0.1047, 0.2851, 0.1557, 0.1419], device='cuda:2'), in_proj_covar=tensor([0.0612, 0.0754, 0.0881, 0.0770, 0.0580, 0.0603, 0.0622, 0.0718], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:17:11,462 INFO [train.py:904] (2/8) Epoch 20, batch 7550, loss[loss=0.2222, simple_loss=0.2901, pruned_loss=0.07712, over 11408.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2917, pruned_loss=0.06109, over 3018975.31 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:17:23,796 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7418, 4.9927, 5.1565, 4.9639, 5.0552, 5.5450, 5.0254, 4.7941], device='cuda:2'), covar=tensor([0.1079, 0.1668, 0.2167, 0.1717, 0.2250, 0.0921, 0.1602, 0.2382], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0581, 0.0639, 0.0479, 0.0637, 0.0669, 0.0501, 0.0648], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 03:17:56,621 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3323, 2.9241, 2.6923, 2.2771, 2.2677, 2.2891, 2.8382, 2.8338], device='cuda:2'), covar=tensor([0.2452, 0.0712, 0.1579, 0.2303, 0.2439, 0.2065, 0.0459, 0.1221], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0270, 0.0305, 0.0310, 0.0298, 0.0257, 0.0295, 0.0336], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 03:18:26,295 INFO [train.py:904] (2/8) Epoch 20, batch 7600, loss[loss=0.2264, simple_loss=0.2904, pruned_loss=0.08124, over 11166.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2903, pruned_loss=0.06041, over 3034130.87 frames. ], batch size: 246, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:18:37,025 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7679, 3.7724, 3.8773, 3.6777, 3.8801, 4.2082, 3.8639, 3.6515], device='cuda:2'), covar=tensor([0.2011, 0.2275, 0.2641, 0.2525, 0.2323, 0.1850, 0.1733, 0.2527], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0582, 0.0640, 0.0480, 0.0639, 0.0671, 0.0502, 0.0649], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 03:19:06,493 INFO [optim.py:368] (2/8) 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 03:19:44,067 INFO [train.py:904] (2/8) Epoch 20, batch 7650, loss[loss=0.1964, simple_loss=0.2902, pruned_loss=0.05129, over 17165.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.291, pruned_loss=0.06101, over 3042228.92 frames. ], batch size: 46, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:20:33,303 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200533.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:21:02,175 INFO [train.py:904] (2/8) Epoch 20, batch 7700, loss[loss=0.1949, simple_loss=0.2797, pruned_loss=0.05509, over 16560.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2912, pruned_loss=0.06146, over 3047504.67 frames. ], batch size: 57, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:21:22,795 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200565.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:21:44,265 INFO [optim.py:368] (2/8) 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 03:22:21,773 INFO [train.py:904] (2/8) Epoch 20, batch 7750, loss[loss=0.2294, simple_loss=0.3009, pruned_loss=0.079, over 11649.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.291, pruned_loss=0.06121, over 3045555.20 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:22:32,122 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 03:22:38,649 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200613.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:06,513 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3525, 4.4173, 4.7887, 4.7705, 4.7482, 4.4605, 4.4223, 4.3026], device='cuda:2'), covar=tensor([0.0378, 0.0596, 0.0347, 0.0345, 0.0541, 0.0410, 0.0984, 0.0512], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0439, 0.0425, 0.0396, 0.0474, 0.0447, 0.0538, 0.0360], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 03:23:11,917 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200635.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:23:28,191 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:36,962 INFO [train.py:904] (2/8) Epoch 20, batch 7800, loss[loss=0.2044, simple_loss=0.2927, pruned_loss=0.05812, over 16763.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2919, pruned_loss=0.06203, over 3042133.93 frames. ], batch size: 83, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:24:18,682 INFO [optim.py:368] (2/8) 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 03:24:23,020 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3750, 3.2703, 3.7672, 1.8461, 3.9127, 3.9401, 2.9584, 2.8143], device='cuda:2'), covar=tensor([0.0808, 0.0279, 0.0173, 0.1161, 0.0071, 0.0173, 0.0413, 0.0475], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0107, 0.0096, 0.0138, 0.0078, 0.0122, 0.0126, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 03:24:53,417 INFO [train.py:904] (2/8) Epoch 20, batch 7850, loss[loss=0.2393, simple_loss=0.3055, pruned_loss=0.08661, over 11473.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2929, pruned_loss=0.06233, over 3024609.25 frames. ], batch size: 247, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:24:59,617 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:25:39,106 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 03:26:08,718 INFO [train.py:904] (2/8) Epoch 20, batch 7900, loss[loss=0.2111, simple_loss=0.2953, pruned_loss=0.06338, over 16682.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2914, pruned_loss=0.06112, over 3051303.46 frames. ], batch size: 62, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:26:49,222 INFO [optim.py:368] (2/8) 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 03:27:27,127 INFO [train.py:904] (2/8) Epoch 20, batch 7950, loss[loss=0.2016, simple_loss=0.2814, pruned_loss=0.06084, over 16595.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2921, pruned_loss=0.06157, over 3060712.73 frames. ], batch size: 57, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:28:09,350 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 03:28:16,259 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200833.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:28:44,605 INFO [train.py:904] (2/8) Epoch 20, batch 8000, loss[loss=0.1968, simple_loss=0.2836, pruned_loss=0.05504, over 17065.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.293, pruned_loss=0.06225, over 3055376.02 frames. ], batch size: 53, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:29:24,818 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200881.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:29:59,852 INFO [train.py:904] (2/8) Epoch 20, batch 8050, loss[loss=0.2011, simple_loss=0.2931, pruned_loss=0.0546, over 16827.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2927, pruned_loss=0.06164, over 3066984.23 frames. ], batch size: 102, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:30:08,109 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6032, 4.6753, 4.4560, 4.2035, 4.1464, 4.6034, 4.3442, 4.2692], device='cuda:2'), covar=tensor([0.0668, 0.0605, 0.0332, 0.0300, 0.0953, 0.0508, 0.0458, 0.0633], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0408, 0.0329, 0.0324, 0.0338, 0.0376, 0.0227, 0.0394], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:30:12,842 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 03:30:50,042 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:31:15,303 INFO [train.py:904] (2/8) Epoch 20, batch 8100, loss[loss=0.1947, simple_loss=0.2862, pruned_loss=0.05155, over 16933.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2923, pruned_loss=0.06137, over 3049111.72 frames. ], batch size: 116, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:31:57,146 INFO [optim.py:368] (2/8) 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 03:32:04,948 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200983.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:31,846 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:33,316 INFO [train.py:904] (2/8) Epoch 20, batch 8150, loss[loss=0.1849, simple_loss=0.2614, pruned_loss=0.05423, over 17047.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2896, pruned_loss=0.05987, over 3080278.75 frames. ], batch size: 53, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:33:06,095 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 03:33:36,236 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0523, 4.8522, 5.1125, 5.2803, 5.4930, 4.8994, 5.4675, 5.4661], device='cuda:2'), covar=tensor([0.1954, 0.1303, 0.1609, 0.0725, 0.0577, 0.0828, 0.0572, 0.0636], device='cuda:2'), in_proj_covar=tensor([0.0613, 0.0758, 0.0884, 0.0768, 0.0584, 0.0604, 0.0622, 0.0720], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:33:52,111 INFO [train.py:904] (2/8) Epoch 20, batch 8200, loss[loss=0.1831, simple_loss=0.2768, pruned_loss=0.04467, over 16931.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2874, pruned_loss=0.05912, over 3087529.87 frames. ], batch size: 90, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:34:16,077 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5554, 3.5814, 3.3752, 3.0668, 3.1969, 3.5047, 3.3192, 3.3392], device='cuda:2'), covar=tensor([0.0597, 0.0672, 0.0292, 0.0291, 0.0573, 0.0476, 0.1336, 0.0509], device='cuda:2'), in_proj_covar=tensor([0.0282, 0.0407, 0.0326, 0.0323, 0.0337, 0.0374, 0.0226, 0.0392], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:34:24,246 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:34:36,686 INFO [optim.py:368] (2/8) 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 03:34:53,835 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-01 03:35:15,217 INFO [train.py:904] (2/8) Epoch 20, batch 8250, loss[loss=0.1729, simple_loss=0.2661, pruned_loss=0.03988, over 15388.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2857, pruned_loss=0.0573, over 3031182.61 frames. ], batch size: 190, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:35:29,700 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0705, 4.0177, 4.0032, 2.7879, 3.9966, 1.4388, 3.6537, 3.5824], device='cuda:2'), covar=tensor([0.0202, 0.0176, 0.0259, 0.0750, 0.0184, 0.3912, 0.0256, 0.0483], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0147, 0.0189, 0.0173, 0.0166, 0.0200, 0.0179, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:35:35,809 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 03:35:57,996 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 03:36:06,047 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:36:38,968 INFO [train.py:904] (2/8) Epoch 20, batch 8300, loss[loss=0.1893, simple_loss=0.2843, pruned_loss=0.04713, over 16936.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2827, pruned_loss=0.05404, over 3033684.23 frames. ], batch size: 109, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:37:21,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0743, 1.5283, 1.9065, 2.0442, 2.1945, 2.3010, 1.8183, 2.2691], device='cuda:2'), covar=tensor([0.0220, 0.0499, 0.0301, 0.0324, 0.0293, 0.0187, 0.0452, 0.0160], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0189, 0.0173, 0.0178, 0.0190, 0.0147, 0.0191, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:37:22,447 INFO [optim.py:368] (2/8) 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 03:38:00,751 INFO [train.py:904] (2/8) Epoch 20, batch 8350, loss[loss=0.1906, simple_loss=0.2894, pruned_loss=0.04594, over 15460.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2819, pruned_loss=0.05186, over 3030099.32 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:38:11,459 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9376, 2.1450, 2.3026, 3.2353, 2.1646, 2.3759, 2.2661, 2.2009], device='cuda:2'), covar=tensor([0.1197, 0.3658, 0.2706, 0.0677, 0.4769, 0.2612, 0.3974, 0.3856], device='cuda:2'), in_proj_covar=tensor([0.0387, 0.0434, 0.0357, 0.0317, 0.0426, 0.0497, 0.0403, 0.0506], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:39:23,231 INFO [train.py:904] (2/8) Epoch 20, batch 8400, loss[loss=0.1596, simple_loss=0.2568, pruned_loss=0.03119, over 16707.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2797, pruned_loss=0.05014, over 3025450.21 frames. ], batch size: 89, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:25,951 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:39:37,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6754, 4.7104, 4.5493, 4.2056, 4.2133, 4.6349, 4.4431, 4.3372], device='cuda:2'), covar=tensor([0.0556, 0.0551, 0.0303, 0.0314, 0.0910, 0.0473, 0.0401, 0.0704], device='cuda:2'), in_proj_covar=tensor([0.0281, 0.0406, 0.0325, 0.0321, 0.0336, 0.0374, 0.0226, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:40:00,755 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:08,502 INFO [optim.py:368] (2/8) 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 03:40:11,078 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6840, 2.6447, 1.8106, 2.8187, 2.1021, 2.8110, 2.1224, 2.4776], device='cuda:2'), covar=tensor([0.0267, 0.0330, 0.1200, 0.0281, 0.0619, 0.0417, 0.1136, 0.0552], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0170, 0.0188, 0.0152, 0.0171, 0.0207, 0.0195, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 03:40:44,263 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201301.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:45,023 INFO [train.py:904] (2/8) Epoch 20, batch 8450, loss[loss=0.1695, simple_loss=0.2632, pruned_loss=0.03786, over 17050.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.278, pruned_loss=0.04859, over 3029670.18 frames. ], batch size: 53, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:41:05,340 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:41:39,385 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:02,466 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:07,348 INFO [train.py:904] (2/8) Epoch 20, batch 8500, loss[loss=0.1928, simple_loss=0.2786, pruned_loss=0.05355, over 16671.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2744, pruned_loss=0.04607, over 3042550.42 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:42:09,926 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-01 03:42:50,125 INFO [optim.py:368] (2/8) 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 03:43:31,052 INFO [train.py:904] (2/8) Epoch 20, batch 8550, loss[loss=0.1824, simple_loss=0.2833, pruned_loss=0.04073, over 16155.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2725, pruned_loss=0.04489, over 3044550.29 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:44:18,611 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:45:09,657 INFO [train.py:904] (2/8) Epoch 20, batch 8600, loss[loss=0.1871, simple_loss=0.2799, pruned_loss=0.04717, over 16682.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2723, pruned_loss=0.04371, over 3047473.72 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:45:32,502 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 03:46:02,940 INFO [optim.py:368] (2/8) 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 03:46:48,552 INFO [train.py:904] (2/8) Epoch 20, batch 8650, loss[loss=0.1758, simple_loss=0.2653, pruned_loss=0.04311, over 12315.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2703, pruned_loss=0.04231, over 3034856.62 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:48:02,811 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0949, 4.1373, 4.4414, 4.4155, 4.4299, 4.2291, 4.1694, 4.1597], device='cuda:2'), covar=tensor([0.0330, 0.0606, 0.0427, 0.0422, 0.0422, 0.0358, 0.0818, 0.0466], device='cuda:2'), in_proj_covar=tensor([0.0394, 0.0435, 0.0422, 0.0391, 0.0468, 0.0439, 0.0529, 0.0355], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 03:48:36,194 INFO [train.py:904] (2/8) Epoch 20, batch 8700, loss[loss=0.182, simple_loss=0.2727, pruned_loss=0.04567, over 15331.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2674, pruned_loss=0.04083, over 3057310.36 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:49:28,969 INFO [optim.py:368] (2/8) 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 03:50:13,600 INFO [train.py:904] (2/8) Epoch 20, batch 8750, loss[loss=0.1998, simple_loss=0.292, pruned_loss=0.05386, over 16764.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2669, pruned_loss=0.04022, over 3054414.87 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:50:31,548 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201609.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:22,015 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201630.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:51,032 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201644.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:52:07,672 INFO [train.py:904] (2/8) Epoch 20, batch 8800, loss[loss=0.1776, simple_loss=0.2691, pruned_loss=0.04309, over 16925.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2656, pruned_loss=0.03919, over 3062955.88 frames. ], batch size: 109, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:52:24,874 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201661.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:53:06,515 INFO [optim.py:368] (2/8) 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 03:53:39,411 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8765, 2.0758, 2.3262, 3.1567, 2.0742, 2.2302, 2.2385, 2.1259], device='cuda:2'), covar=tensor([0.1291, 0.3836, 0.2775, 0.0693, 0.4512, 0.2678, 0.3413, 0.3994], device='cuda:2'), in_proj_covar=tensor([0.0387, 0.0433, 0.0357, 0.0316, 0.0427, 0.0497, 0.0403, 0.0506], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:53:51,629 INFO [train.py:904] (2/8) Epoch 20, batch 8850, loss[loss=0.1782, simple_loss=0.282, pruned_loss=0.03724, over 16571.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2683, pruned_loss=0.03869, over 3058201.82 frames. ], batch size: 62, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:53:57,898 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:53:59,714 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:34,733 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:45,908 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:55:33,461 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0619, 2.1113, 2.2158, 3.5657, 2.0560, 2.3960, 2.2210, 2.2332], device='cuda:2'), covar=tensor([0.1202, 0.3591, 0.2920, 0.0538, 0.4407, 0.2496, 0.3857, 0.3554], device='cuda:2'), in_proj_covar=tensor([0.0387, 0.0432, 0.0356, 0.0314, 0.0425, 0.0495, 0.0403, 0.0504], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:55:38,107 INFO [train.py:904] (2/8) Epoch 20, batch 8900, loss[loss=0.1802, simple_loss=0.2722, pruned_loss=0.04405, over 16918.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2683, pruned_loss=0.03812, over 3054600.76 frames. ], batch size: 116, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:56:08,823 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:27,846 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:46,607 INFO [optim.py:368] (2/8) 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 03:57:11,413 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3852, 2.0387, 1.7545, 1.7614, 2.3035, 1.9656, 1.8859, 2.3732], device='cuda:2'), covar=tensor([0.0156, 0.0388, 0.0534, 0.0463, 0.0241, 0.0362, 0.0207, 0.0269], device='cuda:2'), in_proj_covar=tensor([0.0190, 0.0224, 0.0216, 0.0216, 0.0224, 0.0224, 0.0221, 0.0218], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 03:57:41,707 INFO [train.py:904] (2/8) Epoch 20, batch 8950, loss[loss=0.1658, simple_loss=0.262, pruned_loss=0.03475, over 16301.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2679, pruned_loss=0.03813, over 3080472.50 frames. ], batch size: 166, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:59:12,781 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 03:59:30,698 INFO [train.py:904] (2/8) Epoch 20, batch 9000, loss[loss=0.1831, simple_loss=0.2708, pruned_loss=0.04775, over 16995.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2652, pruned_loss=0.0371, over 3089345.17 frames. ], batch size: 116, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:59:30,699 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 03:59:41,107 INFO [train.py:938] (2/8) Epoch 20, validation: loss=0.1464, simple_loss=0.2502, pruned_loss=0.02125, over 944034.00 frames. 2023-05-01 03:59:41,108 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 04:00:41,794 INFO [optim.py:368] (2/8) 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 04:00:42,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4741, 2.3206, 2.2775, 4.3043, 2.2325, 2.6800, 2.3366, 2.4478], device='cuda:2'), covar=tensor([0.1038, 0.3550, 0.3019, 0.0373, 0.4094, 0.2417, 0.3566, 0.3244], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0431, 0.0356, 0.0314, 0.0424, 0.0494, 0.0402, 0.0503], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:01:24,589 INFO [train.py:904] (2/8) Epoch 20, batch 9050, loss[loss=0.1536, simple_loss=0.2484, pruned_loss=0.02943, over 16543.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2659, pruned_loss=0.03733, over 3099384.23 frames. ], batch size: 75, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:01:40,823 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:02:21,313 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:02:31,897 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4740, 3.4049, 3.5224, 3.5780, 3.6444, 3.3380, 3.6125, 3.6892], device='cuda:2'), covar=tensor([0.1289, 0.0988, 0.1062, 0.0697, 0.0665, 0.2019, 0.0787, 0.0827], device='cuda:2'), in_proj_covar=tensor([0.0597, 0.0736, 0.0859, 0.0754, 0.0569, 0.0590, 0.0607, 0.0702], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:03:12,549 INFO [train.py:904] (2/8) Epoch 20, batch 9100, loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.04059, over 12268.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2654, pruned_loss=0.03765, over 3094972.83 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:03:22,656 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:10,544 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 04:04:14,333 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:23,055 INFO [optim.py:368] (2/8) 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 04:04:42,902 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:05:11,348 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202000.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:05:14,319 INFO [train.py:904] (2/8) Epoch 20, batch 9150, loss[loss=0.1498, simple_loss=0.2473, pruned_loss=0.02615, over 15280.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.266, pruned_loss=0.03764, over 3080838.41 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:05:48,716 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202017.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:58,591 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202051.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:59,884 INFO [train.py:904] (2/8) Epoch 20, batch 9200, loss[loss=0.1527, simple_loss=0.248, pruned_loss=0.02872, over 16538.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2614, pruned_loss=0.03657, over 3081127.23 frames. ], batch size: 75, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:07:19,865 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:07:55,481 INFO [optim.py:368] (2/8) 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 04:08:37,603 INFO [train.py:904] (2/8) Epoch 20, batch 9250, loss[loss=0.1727, simple_loss=0.269, pruned_loss=0.03822, over 15434.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2612, pruned_loss=0.03653, over 3066094.05 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:10:09,038 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0206, 1.7980, 1.6982, 1.5042, 1.9858, 1.6201, 1.5926, 1.9415], device='cuda:2'), covar=tensor([0.0179, 0.0312, 0.0422, 0.0365, 0.0222, 0.0325, 0.0160, 0.0226], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0225, 0.0216, 0.0216, 0.0224, 0.0225, 0.0221, 0.0218], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:10:29,962 INFO [train.py:904] (2/8) Epoch 20, batch 9300, loss[loss=0.1569, simple_loss=0.2514, pruned_loss=0.03117, over 16216.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2594, pruned_loss=0.03597, over 3054560.34 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:11:23,006 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 04:11:37,344 INFO [optim.py:368] (2/8) 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 04:12:16,145 INFO [train.py:904] (2/8) Epoch 20, batch 9350, loss[loss=0.1739, simple_loss=0.2673, pruned_loss=0.0402, over 16277.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2597, pruned_loss=0.03602, over 3083438.96 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:12:22,193 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6159, 3.5914, 3.5740, 2.9116, 3.4755, 1.9675, 3.2361, 2.8964], device='cuda:2'), covar=tensor([0.0121, 0.0105, 0.0140, 0.0197, 0.0096, 0.2308, 0.0113, 0.0230], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0144, 0.0183, 0.0166, 0.0163, 0.0198, 0.0175, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:13:59,184 INFO [train.py:904] (2/8) Epoch 20, batch 9400, loss[loss=0.1823, simple_loss=0.2835, pruned_loss=0.04053, over 16939.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2598, pruned_loss=0.03603, over 3082474.01 frames. ], batch size: 116, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:14:19,754 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8823, 3.7931, 3.9238, 4.0055, 4.1076, 3.7123, 4.0825, 4.1205], device='cuda:2'), covar=tensor([0.1490, 0.1072, 0.1268, 0.0727, 0.0608, 0.1790, 0.0753, 0.0654], device='cuda:2'), in_proj_covar=tensor([0.0593, 0.0729, 0.0851, 0.0749, 0.0565, 0.0586, 0.0602, 0.0696], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:15:00,972 INFO [optim.py:368] (2/8) 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 04:15:39,427 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:15:41,530 INFO [train.py:904] (2/8) Epoch 20, batch 9450, loss[loss=0.1535, simple_loss=0.2508, pruned_loss=0.02816, over 16669.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2622, pruned_loss=0.03664, over 3081646.18 frames. ], batch size: 83, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:16:13,663 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202317.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:16:29,263 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 04:17:14,532 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202346.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:17,846 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202348.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:25,744 INFO [train.py:904] (2/8) Epoch 20, batch 9500, loss[loss=0.1709, simple_loss=0.2724, pruned_loss=0.03468, over 16142.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2606, pruned_loss=0.03601, over 3072997.51 frames. ], batch size: 165, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:17:49,552 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202362.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:55,103 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:57,190 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6088, 2.1074, 1.8325, 1.8240, 2.4368, 2.0673, 2.0189, 2.5079], device='cuda:2'), covar=tensor([0.0178, 0.0390, 0.0513, 0.0515, 0.0252, 0.0431, 0.0180, 0.0279], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0226, 0.0218, 0.0217, 0.0225, 0.0226, 0.0221, 0.0219], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:18:26,629 INFO [optim.py:368] (2/8) 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 04:19:13,221 INFO [train.py:904] (2/8) Epoch 20, batch 9550, loss[loss=0.1859, simple_loss=0.2823, pruned_loss=0.04476, over 16343.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2601, pruned_loss=0.03588, over 3087811.94 frames. ], batch size: 146, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:19:32,642 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202410.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:19:54,360 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0488, 2.0702, 2.4401, 2.9080, 2.7424, 3.2460, 2.3575, 3.2624], device='cuda:2'), covar=tensor([0.0194, 0.0504, 0.0352, 0.0288, 0.0311, 0.0187, 0.0469, 0.0152], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0186, 0.0171, 0.0175, 0.0187, 0.0144, 0.0188, 0.0140], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:19:58,748 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2310, 4.2854, 4.6209, 4.6227, 4.6064, 4.3454, 4.3328, 4.2792], device='cuda:2'), covar=tensor([0.0362, 0.0652, 0.0481, 0.0423, 0.0525, 0.0450, 0.0886, 0.0468], device='cuda:2'), in_proj_covar=tensor([0.0382, 0.0420, 0.0409, 0.0379, 0.0454, 0.0429, 0.0510, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 04:20:36,215 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5738, 3.6164, 3.4119, 3.1247, 3.2579, 3.5375, 3.3571, 3.3634], device='cuda:2'), covar=tensor([0.0599, 0.0716, 0.0288, 0.0273, 0.0523, 0.0539, 0.1093, 0.0464], device='cuda:2'), in_proj_covar=tensor([0.0271, 0.0387, 0.0314, 0.0310, 0.0321, 0.0358, 0.0217, 0.0373], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-05-01 04:20:55,235 INFO [train.py:904] (2/8) Epoch 20, batch 9600, loss[loss=0.1803, simple_loss=0.2801, pruned_loss=0.04021, over 16158.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2618, pruned_loss=0.03695, over 3067952.29 frames. ], batch size: 165, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:21:53,894 INFO [optim.py:368] (2/8) 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 04:22:41,347 INFO [train.py:904] (2/8) Epoch 20, batch 9650, loss[loss=0.1759, simple_loss=0.2587, pruned_loss=0.04656, over 12435.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2641, pruned_loss=0.03752, over 3058582.39 frames. ], batch size: 247, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:23:44,631 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 04:24:31,592 INFO [train.py:904] (2/8) Epoch 20, batch 9700, loss[loss=0.1576, simple_loss=0.26, pruned_loss=0.02757, over 17044.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2626, pruned_loss=0.03716, over 3040662.67 frames. ], batch size: 55, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:25:37,514 INFO [optim.py:368] (2/8) 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 04:26:15,967 INFO [train.py:904] (2/8) Epoch 20, batch 9750, loss[loss=0.1617, simple_loss=0.2594, pruned_loss=0.03207, over 16733.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2618, pruned_loss=0.03744, over 3049089.30 frames. ], batch size: 83, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:27:27,943 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-01 04:27:47,505 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202646.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:27:57,087 INFO [train.py:904] (2/8) Epoch 20, batch 9800, loss[loss=0.1604, simple_loss=0.2478, pruned_loss=0.03648, over 12234.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.262, pruned_loss=0.03659, over 3059548.05 frames. ], batch size: 247, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:28:57,880 INFO [optim.py:368] (2/8) 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 04:29:26,526 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:29:42,360 INFO [train.py:904] (2/8) Epoch 20, batch 9850, loss[loss=0.1593, simple_loss=0.2589, pruned_loss=0.02987, over 15372.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2632, pruned_loss=0.03637, over 3066831.32 frames. ], batch size: 192, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:30:11,387 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 04:31:34,887 INFO [train.py:904] (2/8) Epoch 20, batch 9900, loss[loss=0.1805, simple_loss=0.2815, pruned_loss=0.03976, over 16278.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.264, pruned_loss=0.03632, over 3064923.07 frames. ], batch size: 146, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:32:49,189 INFO [optim.py:368] (2/8) 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 04:33:30,682 INFO [train.py:904] (2/8) Epoch 20, batch 9950, loss[loss=0.1478, simple_loss=0.2491, pruned_loss=0.02324, over 16730.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2657, pruned_loss=0.03649, over 3054614.95 frames. ], batch size: 83, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:34:41,303 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202830.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:35:32,301 INFO [train.py:904] (2/8) Epoch 20, batch 10000, loss[loss=0.1962, simple_loss=0.2974, pruned_loss=0.04749, over 16469.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2649, pruned_loss=0.03639, over 3066101.93 frames. ], batch size: 147, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:36:36,180 INFO [optim.py:368] (2/8) 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 04:36:51,589 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202891.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:37:11,275 INFO [train.py:904] (2/8) Epoch 20, batch 10050, loss[loss=0.1705, simple_loss=0.2667, pruned_loss=0.0371, over 16780.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2651, pruned_loss=0.03619, over 3081910.92 frames. ], batch size: 76, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:38:05,712 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-01 04:38:17,146 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0377, 1.7970, 1.6177, 1.4576, 2.0051, 1.6604, 1.6263, 1.9705], device='cuda:2'), covar=tensor([0.0182, 0.0342, 0.0454, 0.0413, 0.0239, 0.0321, 0.0220, 0.0264], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0227, 0.0218, 0.0218, 0.0226, 0.0227, 0.0221, 0.0219], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:38:41,706 INFO [train.py:904] (2/8) Epoch 20, batch 10100, loss[loss=0.1651, simple_loss=0.2568, pruned_loss=0.03667, over 16607.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2655, pruned_loss=0.03661, over 3081082.35 frames. ], batch size: 57, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:39:40,380 INFO [optim.py:368] (2/8) 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 04:40:23,118 INFO [train.py:904] (2/8) Epoch 21, batch 0, loss[loss=0.1763, simple_loss=0.2688, pruned_loss=0.04192, over 17127.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2688, pruned_loss=0.04192, over 17127.00 frames. ], batch size: 48, lr: 3.26e-03, grad_scale: 8.0 2023-05-01 04:40:23,119 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 04:40:30,883 INFO [train.py:938] (2/8) Epoch 21, validation: loss=0.1451, simple_loss=0.2491, pruned_loss=0.02058, over 944034.00 frames. 2023-05-01 04:40:30,884 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 04:40:49,588 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2424, 5.8736, 5.9743, 5.6762, 5.7159, 6.3186, 5.8525, 5.5367], device='cuda:2'), covar=tensor([0.0941, 0.1812, 0.2081, 0.1960, 0.2466, 0.0920, 0.1387, 0.2112], device='cuda:2'), in_proj_covar=tensor([0.0373, 0.0549, 0.0604, 0.0456, 0.0604, 0.0637, 0.0474, 0.0610], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 04:41:11,753 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1881, 4.2925, 4.4302, 4.2460, 4.3458, 4.8440, 4.3853, 4.0681], device='cuda:2'), covar=tensor([0.1924, 0.2158, 0.2538, 0.2245, 0.2649, 0.1206, 0.1711, 0.2628], device='cuda:2'), in_proj_covar=tensor([0.0375, 0.0551, 0.0608, 0.0458, 0.0608, 0.0640, 0.0476, 0.0613], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 04:41:19,634 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:41:21,221 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 04:41:36,908 INFO [train.py:904] (2/8) Epoch 21, batch 50, loss[loss=0.1741, simple_loss=0.2557, pruned_loss=0.04622, over 16826.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2716, pruned_loss=0.05234, over 731835.93 frames. ], batch size: 83, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:42:24,827 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 04:42:43,730 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203099.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:42:48,525 INFO [train.py:904] (2/8) Epoch 21, batch 100, loss[loss=0.1823, simple_loss=0.2641, pruned_loss=0.05032, over 16871.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2674, pruned_loss=0.04913, over 1304919.48 frames. ], batch size: 96, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:43:50,264 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:43:56,573 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6556, 6.0865, 5.7528, 5.8222, 5.3547, 5.4316, 5.4354, 6.1897], device='cuda:2'), covar=tensor([0.1366, 0.1002, 0.1114, 0.0942, 0.0952, 0.0688, 0.1068, 0.0913], device='cuda:2'), in_proj_covar=tensor([0.0637, 0.0777, 0.0638, 0.0584, 0.0495, 0.0503, 0.0654, 0.0603], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:43:57,338 INFO [train.py:904] (2/8) Epoch 21, batch 150, loss[loss=0.1425, simple_loss=0.2308, pruned_loss=0.02709, over 16978.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.265, pruned_loss=0.04765, over 1758077.44 frames. ], batch size: 41, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:44:19,173 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5927, 4.3743, 4.6468, 4.7706, 4.9585, 4.4310, 4.9112, 4.9134], device='cuda:2'), covar=tensor([0.1867, 0.1370, 0.1740, 0.0894, 0.0638, 0.1134, 0.0929, 0.0818], device='cuda:2'), in_proj_covar=tensor([0.0605, 0.0740, 0.0864, 0.0764, 0.0573, 0.0593, 0.0614, 0.0710], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:44:44,683 INFO [optim.py:368] (2/8) 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 04:44:44,940 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203186.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:45:04,864 INFO [train.py:904] (2/8) Epoch 21, batch 200, loss[loss=0.1943, simple_loss=0.2934, pruned_loss=0.04763, over 17159.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2664, pruned_loss=0.04753, over 2093900.92 frames. ], batch size: 49, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:46:15,294 INFO [train.py:904] (2/8) Epoch 21, batch 250, loss[loss=0.1792, simple_loss=0.2603, pruned_loss=0.04907, over 12328.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.265, pruned_loss=0.04774, over 2360854.45 frames. ], batch size: 248, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:46:32,300 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 04:47:00,828 INFO [optim.py:368] (2/8) 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 04:47:23,329 INFO [train.py:904] (2/8) Epoch 21, batch 300, loss[loss=0.168, simple_loss=0.2549, pruned_loss=0.04057, over 16589.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2621, pruned_loss=0.04595, over 2577133.70 frames. ], batch size: 62, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:47:31,909 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 04:48:32,603 INFO [train.py:904] (2/8) Epoch 21, batch 350, loss[loss=0.1885, simple_loss=0.2675, pruned_loss=0.05478, over 16894.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2593, pruned_loss=0.04486, over 2737467.81 frames. ], batch size: 90, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:49:17,938 INFO [optim.py:368] (2/8) 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 04:49:30,089 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203394.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:49:41,962 INFO [train.py:904] (2/8) Epoch 21, batch 400, loss[loss=0.1766, simple_loss=0.262, pruned_loss=0.04559, over 16869.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.259, pruned_loss=0.04503, over 2871049.37 frames. ], batch size: 96, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:49:48,139 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 04:50:23,796 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-05-01 04:50:36,379 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203441.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:50:50,736 INFO [train.py:904] (2/8) Epoch 21, batch 450, loss[loss=0.1649, simple_loss=0.2523, pruned_loss=0.03874, over 16675.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2572, pruned_loss=0.04375, over 2969423.26 frames. ], batch size: 62, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:51:11,801 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:51:26,463 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6737, 3.5672, 3.9617, 2.1549, 4.1365, 4.1595, 3.1456, 3.1290], device='cuda:2'), covar=tensor([0.0781, 0.0264, 0.0209, 0.1168, 0.0094, 0.0180, 0.0422, 0.0464], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0107, 0.0095, 0.0139, 0.0078, 0.0121, 0.0127, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 04:51:37,695 INFO [optim.py:368] (2/8) 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 04:51:38,014 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 21, batch 500, loss[loss=0.1763, simple_loss=0.266, pruned_loss=0.04334, over 16356.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2552, pruned_loss=0.04216, over 3054107.93 frames. ], batch size: 68, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:52:17,324 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 04:52:36,215 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203528.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:52:43,062 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203534.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:53:08,036 INFO [train.py:904] (2/8) Epoch 21, batch 550, loss[loss=0.172, simple_loss=0.2684, pruned_loss=0.03781, over 17277.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2536, pruned_loss=0.04171, over 3104018.37 frames. ], batch size: 52, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:53:56,913 INFO [optim.py:368] (2/8) 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 04:54:17,583 INFO [train.py:904] (2/8) Epoch 21, batch 600, loss[loss=0.1779, simple_loss=0.2531, pruned_loss=0.05129, over 15523.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2523, pruned_loss=0.04134, over 3144833.68 frames. ], batch size: 190, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:54:23,070 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9249, 4.2143, 4.0487, 4.1221, 3.7743, 3.8147, 3.8698, 4.2113], device='cuda:2'), covar=tensor([0.1170, 0.0979, 0.0942, 0.0816, 0.0778, 0.1741, 0.0890, 0.1042], device='cuda:2'), in_proj_covar=tensor([0.0660, 0.0805, 0.0662, 0.0605, 0.0512, 0.0519, 0.0679, 0.0624], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:55:26,336 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1062, 5.6416, 5.7118, 5.3861, 5.5114, 6.1130, 5.5161, 5.2320], device='cuda:2'), covar=tensor([0.0968, 0.1956, 0.2588, 0.2258, 0.2876, 0.1068, 0.1665, 0.2236], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0583, 0.0639, 0.0481, 0.0642, 0.0672, 0.0502, 0.0644], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 04:55:27,189 INFO [train.py:904] (2/8) Epoch 21, batch 650, loss[loss=0.1601, simple_loss=0.2538, pruned_loss=0.03322, over 17079.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2514, pruned_loss=0.04053, over 3185650.28 frames. ], batch size: 53, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:55:28,632 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5433, 3.6012, 3.3304, 3.0086, 3.2025, 3.4954, 3.3140, 3.3623], device='cuda:2'), covar=tensor([0.0617, 0.0623, 0.0330, 0.0302, 0.0508, 0.0426, 0.1284, 0.0511], device='cuda:2'), in_proj_covar=tensor([0.0287, 0.0411, 0.0332, 0.0328, 0.0339, 0.0380, 0.0228, 0.0398], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 04:56:14,236 INFO [optim.py:368] (2/8) 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 04:56:24,850 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:56:34,644 INFO [train.py:904] (2/8) Epoch 21, batch 700, loss[loss=0.1471, simple_loss=0.2326, pruned_loss=0.03081, over 16628.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2518, pruned_loss=0.04084, over 3216854.49 frames. ], batch size: 76, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:57:30,288 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203741.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:31,301 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:37,674 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203747.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:44,451 INFO [train.py:904] (2/8) Epoch 21, batch 750, loss[loss=0.1742, simple_loss=0.2683, pruned_loss=0.04004, over 17296.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2523, pruned_loss=0.04136, over 3240902.20 frames. ], batch size: 52, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:58:03,875 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:33,072 INFO [optim.py:368] (2/8) 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 04:58:36,275 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:44,984 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-01 04:58:54,751 INFO [train.py:904] (2/8) Epoch 21, batch 800, loss[loss=0.1532, simple_loss=0.2408, pruned_loss=0.03278, over 16991.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2524, pruned_loss=0.04172, over 3256022.62 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:59:03,670 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203808.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:24,698 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203823.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:30,026 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:58,166 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:00:04,939 INFO [train.py:904] (2/8) Epoch 21, batch 850, loss[loss=0.1651, simple_loss=0.2611, pruned_loss=0.03452, over 17254.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2515, pruned_loss=0.04099, over 3279387.18 frames. ], batch size: 52, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:00:14,184 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6316, 2.4047, 2.4862, 4.4961, 2.3534, 2.7209, 2.5444, 2.5337], device='cuda:2'), covar=tensor([0.1280, 0.3815, 0.2977, 0.0502, 0.4361, 0.2955, 0.3426, 0.3882], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0443, 0.0366, 0.0325, 0.0434, 0.0507, 0.0414, 0.0518], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:00:53,307 INFO [optim.py:368] (2/8) 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 05:01:11,258 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203900.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:01:12,969 INFO [train.py:904] (2/8) Epoch 21, batch 900, loss[loss=0.1486, simple_loss=0.2437, pruned_loss=0.02673, over 17237.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2499, pruned_loss=0.04008, over 3296343.78 frames. ], batch size: 45, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:01:21,641 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203908.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:02:11,049 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 05:02:12,134 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-01 05:02:13,134 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-05-01 05:02:21,433 INFO [train.py:904] (2/8) Epoch 21, batch 950, loss[loss=0.1874, simple_loss=0.2602, pruned_loss=0.05729, over 16890.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2505, pruned_loss=0.04024, over 3296136.65 frames. ], batch size: 116, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:02:35,118 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203961.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:03:10,185 INFO [optim.py:368] (2/8) 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 05:03:20,727 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9209, 2.0540, 2.4118, 2.7035, 2.7849, 2.7114, 2.0055, 2.9436], device='cuda:2'), covar=tensor([0.0179, 0.0467, 0.0326, 0.0273, 0.0289, 0.0320, 0.0541, 0.0189], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0192, 0.0178, 0.0182, 0.0194, 0.0152, 0.0195, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:03:34,202 INFO [train.py:904] (2/8) Epoch 21, batch 1000, loss[loss=0.1741, simple_loss=0.2416, pruned_loss=0.05326, over 16495.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2489, pruned_loss=0.04024, over 3300887.22 frames. ], batch size: 68, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:03:40,729 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8270, 2.5520, 2.1171, 2.3601, 2.8989, 2.6808, 2.9183, 2.9747], device='cuda:2'), covar=tensor([0.0238, 0.0430, 0.0523, 0.0459, 0.0234, 0.0373, 0.0232, 0.0292], device='cuda:2'), in_proj_covar=tensor([0.0207, 0.0238, 0.0228, 0.0227, 0.0237, 0.0236, 0.0238, 0.0233], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:04:43,738 INFO [train.py:904] (2/8) Epoch 21, batch 1050, loss[loss=0.173, simple_loss=0.2483, pruned_loss=0.04884, over 16882.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2488, pruned_loss=0.04044, over 3306015.10 frames. ], batch size: 109, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:19,275 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204077.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:05:21,122 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9590, 2.1490, 2.3358, 3.4507, 2.1462, 2.3835, 2.2683, 2.2719], device='cuda:2'), covar=tensor([0.1611, 0.3773, 0.3027, 0.0847, 0.4341, 0.2701, 0.3753, 0.3667], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0446, 0.0367, 0.0328, 0.0436, 0.0511, 0.0416, 0.0522], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:05:34,277 INFO [optim.py:368] (2/8) 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 05:05:54,736 INFO [train.py:904] (2/8) Epoch 21, batch 1100, loss[loss=0.1629, simple_loss=0.2427, pruned_loss=0.04151, over 16453.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2484, pruned_loss=0.04047, over 3302277.65 frames. ], batch size: 75, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:56,333 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:18,226 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0548, 2.1430, 2.3500, 3.7069, 2.1385, 2.4272, 2.2484, 2.2984], device='cuda:2'), covar=tensor([0.1575, 0.4017, 0.2997, 0.0729, 0.4047, 0.2803, 0.3952, 0.3367], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0446, 0.0368, 0.0328, 0.0436, 0.0511, 0.0416, 0.0523], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:06:18,341 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 05:06:24,092 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:25,311 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204123.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:45,642 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204138.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:07:05,257 INFO [train.py:904] (2/8) Epoch 21, batch 1150, loss[loss=0.1539, simple_loss=0.2335, pruned_loss=0.03715, over 12273.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2484, pruned_loss=0.03952, over 3301828.85 frames. ], batch size: 248, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:07:16,684 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3939, 4.4850, 4.5973, 4.4442, 4.4843, 5.0265, 4.5598, 4.2353], device='cuda:2'), covar=tensor([0.1788, 0.2263, 0.2488, 0.2202, 0.2923, 0.1183, 0.1733, 0.2558], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0594, 0.0654, 0.0490, 0.0655, 0.0683, 0.0515, 0.0658], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:07:20,315 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:07:31,657 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204171.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:07:53,757 INFO [optim.py:368] (2/8) 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 05:08:13,781 INFO [train.py:904] (2/8) Epoch 21, batch 1200, loss[loss=0.1767, simple_loss=0.2437, pruned_loss=0.05487, over 16902.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.247, pruned_loss=0.0391, over 3303316.81 frames. ], batch size: 109, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:08:15,242 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204203.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:08:44,968 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:09:11,481 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2732, 5.9821, 6.0288, 5.6975, 5.8705, 6.3909, 5.9061, 5.6156], device='cuda:2'), covar=tensor([0.1022, 0.1978, 0.2047, 0.2227, 0.2790, 0.1098, 0.1599, 0.2154], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0595, 0.0656, 0.0491, 0.0658, 0.0685, 0.0516, 0.0661], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:09:24,354 INFO [train.py:904] (2/8) Epoch 21, batch 1250, loss[loss=0.1453, simple_loss=0.2283, pruned_loss=0.03112, over 15870.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2469, pruned_loss=0.03948, over 3296654.49 frames. ], batch size: 35, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:09:25,916 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0574, 4.7645, 5.0916, 5.2677, 5.4569, 4.7874, 5.4026, 5.4463], device='cuda:2'), covar=tensor([0.1779, 0.1308, 0.1758, 0.0736, 0.0602, 0.0950, 0.0571, 0.0627], device='cuda:2'), in_proj_covar=tensor([0.0647, 0.0793, 0.0925, 0.0814, 0.0609, 0.0633, 0.0654, 0.0755], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:09:30,688 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204256.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:10:12,294 INFO [optim.py:368] (2/8) 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 05:10:33,941 INFO [train.py:904] (2/8) Epoch 21, batch 1300, loss[loss=0.151, simple_loss=0.2453, pruned_loss=0.0284, over 17130.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2469, pruned_loss=0.04004, over 3303718.52 frames. ], batch size: 47, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:11:38,610 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1768, 5.7303, 5.7799, 5.4609, 5.6166, 6.1698, 5.6354, 5.2755], device='cuda:2'), covar=tensor([0.0900, 0.2070, 0.2347, 0.2084, 0.2589, 0.0886, 0.1706, 0.2478], device='cuda:2'), in_proj_covar=tensor([0.0405, 0.0596, 0.0656, 0.0491, 0.0658, 0.0686, 0.0517, 0.0661], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:11:42,899 INFO [train.py:904] (2/8) Epoch 21, batch 1350, loss[loss=0.1551, simple_loss=0.2483, pruned_loss=0.03097, over 16789.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2477, pruned_loss=0.04005, over 3314356.87 frames. ], batch size: 57, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:12:26,116 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 05:12:31,606 INFO [optim.py:368] (2/8) 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 05:12:52,528 INFO [train.py:904] (2/8) Epoch 21, batch 1400, loss[loss=0.1558, simple_loss=0.2481, pruned_loss=0.03169, over 17188.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2471, pruned_loss=0.03959, over 3314891.64 frames. ], batch size: 46, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:12:54,586 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204403.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:20,731 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204422.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:35,348 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204433.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:14:00,494 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204451.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:01,406 INFO [train.py:904] (2/8) Epoch 21, batch 1450, loss[loss=0.1911, simple_loss=0.2721, pruned_loss=0.05508, over 16775.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2464, pruned_loss=0.03941, over 3318071.04 frames. ], batch size: 124, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:14:26,857 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204470.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:50,905 INFO [optim.py:368] (2/8) 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 05:15:10,559 INFO [train.py:904] (2/8) Epoch 21, batch 1500, loss[loss=0.1681, simple_loss=0.2591, pruned_loss=0.03857, over 16753.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2465, pruned_loss=0.03955, over 3314876.27 frames. ], batch size: 57, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:15:12,769 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:15:33,803 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204519.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:15:51,606 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8238, 4.0100, 2.6631, 4.5472, 3.0997, 4.5401, 2.4526, 3.1740], device='cuda:2'), covar=tensor([0.0342, 0.0385, 0.1470, 0.0298, 0.0852, 0.0453, 0.1620, 0.0799], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0178, 0.0195, 0.0164, 0.0178, 0.0217, 0.0204, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:16:17,194 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:18,114 INFO [train.py:904] (2/8) Epoch 21, batch 1550, loss[loss=0.1538, simple_loss=0.249, pruned_loss=0.02933, over 17200.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.248, pruned_loss=0.0404, over 3310979.65 frames. ], batch size: 44, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:16:23,610 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204556.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:16:23,777 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4235, 2.4012, 2.4592, 4.2033, 2.3303, 2.7764, 2.3836, 2.5464], device='cuda:2'), covar=tensor([0.1340, 0.3611, 0.2986, 0.0582, 0.4172, 0.2479, 0.3641, 0.3581], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0448, 0.0370, 0.0331, 0.0437, 0.0515, 0.0419, 0.0525], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:16:47,589 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3134, 3.3809, 3.5740, 2.5017, 3.2169, 3.6486, 3.3755, 2.2105], device='cuda:2'), covar=tensor([0.0487, 0.0095, 0.0051, 0.0376, 0.0115, 0.0091, 0.0081, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0084, 0.0083, 0.0135, 0.0099, 0.0110, 0.0094, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:2') 2023-05-01 05:17:07,000 INFO [optim.py:368] (2/8) 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 05:17:26,280 INFO [train.py:904] (2/8) Epoch 21, batch 1600, loss[loss=0.1828, simple_loss=0.2639, pruned_loss=0.0508, over 16765.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.25, pruned_loss=0.04134, over 3317293.83 frames. ], batch size: 83, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:17:28,784 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204604.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:18:13,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4459, 2.8652, 3.1045, 2.0650, 2.7085, 2.1367, 3.0402, 3.1512], device='cuda:2'), covar=tensor([0.0282, 0.0847, 0.0553, 0.1862, 0.0855, 0.0966, 0.0644, 0.0885], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0153, 0.0144, 0.0130, 0.0145, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:18:13,436 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 05:18:35,818 INFO [train.py:904] (2/8) Epoch 21, batch 1650, loss[loss=0.1466, simple_loss=0.2327, pruned_loss=0.03028, over 16829.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2519, pruned_loss=0.04198, over 3320863.82 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:19:25,764 INFO [optim.py:368] (2/8) 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 05:19:27,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3219, 3.4321, 3.6520, 2.5180, 3.3040, 3.7562, 3.4756, 2.0653], device='cuda:2'), covar=tensor([0.0550, 0.0132, 0.0051, 0.0384, 0.0117, 0.0083, 0.0082, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0084, 0.0083, 0.0135, 0.0099, 0.0110, 0.0094, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:2') 2023-05-01 05:19:45,543 INFO [train.py:904] (2/8) Epoch 21, batch 1700, loss[loss=0.1601, simple_loss=0.2481, pruned_loss=0.036, over 17217.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2544, pruned_loss=0.04265, over 3316015.03 frames. ], batch size: 45, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:20:00,529 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5640, 3.3645, 3.7807, 1.8832, 3.8970, 3.9286, 3.1003, 2.8212], device='cuda:2'), covar=tensor([0.0769, 0.0252, 0.0175, 0.1262, 0.0105, 0.0179, 0.0393, 0.0494], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0108, 0.0096, 0.0138, 0.0079, 0.0122, 0.0127, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 05:20:29,910 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204733.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:20:55,678 INFO [train.py:904] (2/8) Epoch 21, batch 1750, loss[loss=0.1961, simple_loss=0.2683, pruned_loss=0.06193, over 16448.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.255, pruned_loss=0.04298, over 3296097.70 frames. ], batch size: 146, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:21:06,515 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 05:21:18,108 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7781, 2.6730, 2.6947, 4.7035, 2.5759, 3.0317, 2.6838, 2.8193], device='cuda:2'), covar=tensor([0.1173, 0.3412, 0.2810, 0.0455, 0.3922, 0.2592, 0.3471, 0.3254], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0448, 0.0369, 0.0331, 0.0436, 0.0515, 0.0418, 0.0525], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:21:37,074 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204781.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:21:46,294 INFO [optim.py:368] (2/8) 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 05:22:06,989 INFO [train.py:904] (2/8) Epoch 21, batch 1800, loss[loss=0.1593, simple_loss=0.2583, pruned_loss=0.03012, over 17141.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2555, pruned_loss=0.04247, over 3307750.91 frames. ], batch size: 49, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:22:07,590 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7157, 4.1056, 3.0260, 2.3506, 2.7073, 2.5880, 4.3107, 3.4886], device='cuda:2'), covar=tensor([0.2873, 0.0661, 0.1817, 0.2908, 0.2776, 0.2013, 0.0476, 0.1434], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0270, 0.0304, 0.0309, 0.0295, 0.0257, 0.0294, 0.0336], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:22:30,013 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:30,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8058, 2.6285, 2.5781, 1.8721, 2.5882, 2.7449, 2.5967, 1.9043], device='cuda:2'), covar=tensor([0.0479, 0.0137, 0.0088, 0.0404, 0.0148, 0.0130, 0.0124, 0.0414], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0085, 0.0083, 0.0136, 0.0100, 0.0110, 0.0094, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:2') 2023-05-01 05:22:32,985 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204821.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:39,188 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 05:23:15,229 INFO [train.py:904] (2/8) Epoch 21, batch 1850, loss[loss=0.1703, simple_loss=0.2538, pruned_loss=0.04341, over 16652.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2567, pruned_loss=0.04256, over 3303284.09 frames. ], batch size: 134, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:23:37,453 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204867.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:23:51,392 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6183, 3.6976, 2.7837, 2.2075, 2.3108, 2.2343, 3.8039, 3.1753], device='cuda:2'), covar=tensor([0.2724, 0.0612, 0.1793, 0.3098, 0.2860, 0.2239, 0.0521, 0.1560], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0271, 0.0305, 0.0310, 0.0296, 0.0257, 0.0295, 0.0336], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:23:57,968 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:24:06,221 INFO [optim.py:368] (2/8) 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 05:24:26,176 INFO [train.py:904] (2/8) Epoch 21, batch 1900, loss[loss=0.1576, simple_loss=0.236, pruned_loss=0.03957, over 16733.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2553, pruned_loss=0.0417, over 3311855.41 frames. ], batch size: 89, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:25:10,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5392, 5.9519, 5.6465, 5.7590, 5.3723, 5.3961, 5.2729, 6.0364], device='cuda:2'), covar=tensor([0.1489, 0.0994, 0.1194, 0.0883, 0.0951, 0.0685, 0.1291, 0.1050], device='cuda:2'), in_proj_covar=tensor([0.0682, 0.0832, 0.0684, 0.0627, 0.0530, 0.0534, 0.0701, 0.0643], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:25:34,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4650, 2.3059, 2.2796, 4.3670, 2.2717, 2.7521, 2.3434, 2.4407], device='cuda:2'), covar=tensor([0.1269, 0.3864, 0.3237, 0.0467, 0.4321, 0.2597, 0.3625, 0.3898], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0450, 0.0370, 0.0332, 0.0437, 0.0516, 0.0419, 0.0527], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:25:35,962 INFO [train.py:904] (2/8) Epoch 21, batch 1950, loss[loss=0.1842, simple_loss=0.2801, pruned_loss=0.04414, over 17022.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2561, pruned_loss=0.04152, over 3315695.67 frames. ], batch size: 55, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:26:05,671 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3140, 3.5484, 3.6713, 2.2510, 3.1497, 2.5551, 3.6969, 3.8555], device='cuda:2'), covar=tensor([0.0255, 0.0910, 0.0615, 0.1940, 0.0838, 0.0974, 0.0600, 0.0905], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0163, 0.0167, 0.0153, 0.0144, 0.0129, 0.0144, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:26:26,411 INFO [optim.py:368] (2/8) 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 05:26:44,684 INFO [train.py:904] (2/8) Epoch 21, batch 2000, loss[loss=0.1798, simple_loss=0.2715, pruned_loss=0.04401, over 17122.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2558, pruned_loss=0.04118, over 3314709.26 frames. ], batch size: 49, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:27:55,520 INFO [train.py:904] (2/8) Epoch 21, batch 2050, loss[loss=0.1822, simple_loss=0.249, pruned_loss=0.05768, over 16760.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2556, pruned_loss=0.04171, over 3324446.00 frames. ], batch size: 124, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:28:44,309 INFO [optim.py:368] (2/8) 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 05:29:04,145 INFO [train.py:904] (2/8) Epoch 21, batch 2100, loss[loss=0.1854, simple_loss=0.2757, pruned_loss=0.0475, over 17091.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2555, pruned_loss=0.04144, over 3331459.11 frames. ], batch size: 53, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:29:14,576 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205109.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:29:30,246 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 05:29:46,472 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3906, 5.3428, 5.2172, 4.6800, 4.8992, 5.3076, 5.2690, 4.8952], device='cuda:2'), covar=tensor([0.0633, 0.0554, 0.0290, 0.0366, 0.1057, 0.0460, 0.0265, 0.0773], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0436, 0.0352, 0.0349, 0.0361, 0.0404, 0.0240, 0.0424], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:30:03,643 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7031, 3.7556, 2.4184, 4.2452, 2.9344, 4.1930, 2.6635, 3.1852], device='cuda:2'), covar=tensor([0.0319, 0.0396, 0.1503, 0.0353, 0.0778, 0.0674, 0.1314, 0.0700], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0166, 0.0179, 0.0220, 0.0205, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:30:06,294 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-01 05:30:14,923 INFO [train.py:904] (2/8) Epoch 21, batch 2150, loss[loss=0.1798, simple_loss=0.2621, pruned_loss=0.04876, over 16776.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2573, pruned_loss=0.04236, over 3331159.94 frames. ], batch size: 124, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:30:20,388 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-01 05:30:39,995 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205170.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:30:49,798 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 05:31:19,770 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205198.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:24,886 INFO [train.py:904] (2/8) Epoch 21, batch 2200, loss[loss=0.2025, simple_loss=0.2825, pruned_loss=0.06128, over 16128.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2572, pruned_loss=0.0423, over 3328083.34 frames. ], batch size: 165, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:31:36,645 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8409, 3.9361, 2.5833, 4.6631, 3.1182, 4.5294, 2.6645, 3.3611], device='cuda:2'), covar=tensor([0.0309, 0.0367, 0.1455, 0.0228, 0.0796, 0.0489, 0.1429, 0.0651], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0181, 0.0198, 0.0167, 0.0180, 0.0221, 0.0206, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:31:53,440 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-01 05:32:34,214 INFO [train.py:904] (2/8) Epoch 21, batch 2250, loss[loss=0.1437, simple_loss=0.2272, pruned_loss=0.03013, over 16773.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2581, pruned_loss=0.04271, over 3326374.03 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:32:45,102 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:33:23,477 INFO [optim.py:368] (2/8) 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 05:33:33,364 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8492, 3.9712, 3.1350, 2.3474, 2.5339, 2.4658, 4.1033, 3.4443], device='cuda:2'), covar=tensor([0.2590, 0.0537, 0.1606, 0.3001, 0.2957, 0.2048, 0.0492, 0.1426], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0270, 0.0303, 0.0308, 0.0296, 0.0257, 0.0293, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:33:44,278 INFO [train.py:904] (2/8) Epoch 21, batch 2300, loss[loss=0.1775, simple_loss=0.2626, pruned_loss=0.04622, over 16503.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.257, pruned_loss=0.04211, over 3333814.21 frames. ], batch size: 68, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:34:53,179 INFO [train.py:904] (2/8) Epoch 21, batch 2350, loss[loss=0.1506, simple_loss=0.2407, pruned_loss=0.03027, over 16997.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2574, pruned_loss=0.04203, over 3329390.88 frames. ], batch size: 41, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:35:42,800 INFO [optim.py:368] (2/8) 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 05:35:47,907 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 05:36:02,959 INFO [train.py:904] (2/8) Epoch 21, batch 2400, loss[loss=0.1648, simple_loss=0.2574, pruned_loss=0.0361, over 16490.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2579, pruned_loss=0.04206, over 3334399.06 frames. ], batch size: 68, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:37:10,843 INFO [train.py:904] (2/8) Epoch 21, batch 2450, loss[loss=0.1646, simple_loss=0.2463, pruned_loss=0.04145, over 16249.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2583, pruned_loss=0.0416, over 3339338.16 frames. ], batch size: 165, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:37:29,569 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205465.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:37:33,920 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5354, 3.7621, 3.9743, 2.7734, 3.6245, 4.0714, 3.6643, 2.2331], device='cuda:2'), covar=tensor([0.0481, 0.0235, 0.0056, 0.0376, 0.0112, 0.0086, 0.0096, 0.0482], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:37:45,902 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205477.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:38:00,842 INFO [optim.py:368] (2/8) 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 05:38:21,260 INFO [train.py:904] (2/8) Epoch 21, batch 2500, loss[loss=0.1862, simple_loss=0.2803, pruned_loss=0.04611, over 16662.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2584, pruned_loss=0.04188, over 3332264.99 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:38:24,299 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 05:38:34,920 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6240, 2.5553, 1.8505, 2.7143, 2.0639, 2.7968, 2.0817, 2.3315], device='cuda:2'), covar=tensor([0.0310, 0.0345, 0.1301, 0.0263, 0.0669, 0.0410, 0.1173, 0.0594], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0180, 0.0197, 0.0166, 0.0179, 0.0220, 0.0205, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:38:35,197 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-05-01 05:38:53,421 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:39:30,351 INFO [train.py:904] (2/8) Epoch 21, batch 2550, loss[loss=0.1704, simple_loss=0.2513, pruned_loss=0.04475, over 16839.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2586, pruned_loss=0.04205, over 3328372.43 frames. ], batch size: 102, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:39:33,574 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:39:41,873 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-01 05:40:08,430 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0825, 5.4003, 5.1450, 5.2309, 4.8937, 4.8112, 4.8777, 5.5300], device='cuda:2'), covar=tensor([0.1318, 0.0928, 0.1043, 0.0872, 0.0945, 0.1013, 0.1222, 0.0871], device='cuda:2'), in_proj_covar=tensor([0.0676, 0.0832, 0.0682, 0.0627, 0.0528, 0.0532, 0.0697, 0.0640], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:40:19,332 INFO [optim.py:368] (2/8) 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 05:40:38,679 INFO [train.py:904] (2/8) Epoch 21, batch 2600, loss[loss=0.1716, simple_loss=0.2688, pruned_loss=0.03721, over 16144.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2589, pruned_loss=0.04227, over 3325783.00 frames. ], batch size: 35, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:41:18,067 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-01 05:41:49,801 INFO [train.py:904] (2/8) Epoch 21, batch 2650, loss[loss=0.1597, simple_loss=0.2539, pruned_loss=0.03274, over 17244.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2588, pruned_loss=0.04148, over 3330237.36 frames. ], batch size: 45, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:41:54,586 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205655.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:42:40,190 INFO [optim.py:368] (2/8) 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 05:43:00,020 INFO [train.py:904] (2/8) Epoch 21, batch 2700, loss[loss=0.1721, simple_loss=0.268, pruned_loss=0.03814, over 16659.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2595, pruned_loss=0.04156, over 3323725.16 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:43:18,631 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:43:49,122 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4981, 4.6140, 4.4169, 4.1804, 3.8338, 4.6861, 4.5832, 4.2794], device='cuda:2'), covar=tensor([0.1308, 0.1408, 0.0607, 0.0609, 0.1914, 0.0785, 0.0549, 0.0940], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0441, 0.0356, 0.0353, 0.0366, 0.0409, 0.0245, 0.0430], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:43:59,556 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:44:09,872 INFO [train.py:904] (2/8) Epoch 21, batch 2750, loss[loss=0.1615, simple_loss=0.2593, pruned_loss=0.03185, over 16732.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2598, pruned_loss=0.041, over 3334609.08 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:44:10,165 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8282, 5.1553, 5.5118, 5.4757, 5.4781, 5.1157, 4.7823, 4.8318], device='cuda:2'), covar=tensor([0.0613, 0.0644, 0.0537, 0.0630, 0.0778, 0.0609, 0.1673, 0.0647], device='cuda:2'), in_proj_covar=tensor([0.0414, 0.0457, 0.0444, 0.0414, 0.0491, 0.0466, 0.0556, 0.0373], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 05:44:13,631 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6075, 3.7497, 3.9186, 2.8138, 3.6097, 4.0053, 3.6441, 2.2456], device='cuda:2'), covar=tensor([0.0469, 0.0177, 0.0054, 0.0349, 0.0102, 0.0082, 0.0091, 0.0458], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0110, 0.0095, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 05:44:29,210 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205765.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:02,031 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 21, batch 2800, loss[loss=0.1643, simple_loss=0.26, pruned_loss=0.03432, over 17104.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2599, pruned_loss=0.04064, over 3333021.83 frames. ], batch size: 47, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:45:24,473 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205805.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:35,278 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205813.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:55,996 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7276, 2.9848, 3.0550, 4.9429, 4.1768, 4.3779, 1.4260, 3.6555], device='cuda:2'), covar=tensor([0.1371, 0.0691, 0.0996, 0.0214, 0.0194, 0.0317, 0.1667, 0.0577], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0192, 0.0208, 0.0217, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:46:29,085 INFO [train.py:904] (2/8) Epoch 21, batch 2850, loss[loss=0.1789, simple_loss=0.278, pruned_loss=0.03988, over 17127.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2589, pruned_loss=0.03999, over 3335685.96 frames. ], batch size: 48, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:46:31,706 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:46:32,185 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 05:46:43,197 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 05:46:57,572 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6723, 4.7137, 5.0967, 5.0897, 5.1133, 4.8030, 4.7611, 4.6016], device='cuda:2'), covar=tensor([0.0381, 0.0583, 0.0367, 0.0372, 0.0501, 0.0426, 0.0918, 0.0563], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0463, 0.0450, 0.0419, 0.0497, 0.0471, 0.0563, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 05:47:09,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2376, 3.5268, 3.8581, 2.2967, 3.0303, 2.4755, 3.8079, 3.7261], device='cuda:2'), covar=tensor([0.0306, 0.0873, 0.0483, 0.1876, 0.0845, 0.0943, 0.0539, 0.0985], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0168, 0.0153, 0.0145, 0.0130, 0.0144, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:47:17,820 INFO [optim.py:368] (2/8) 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 05:47:19,359 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:29,328 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0008, 3.0451, 2.9137, 5.1747, 4.2988, 4.5001, 1.7108, 3.4268], device='cuda:2'), covar=tensor([0.1287, 0.0761, 0.1094, 0.0182, 0.0220, 0.0388, 0.1617, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0175, 0.0195, 0.0192, 0.0208, 0.0217, 0.0202, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:47:36,338 INFO [train.py:904] (2/8) Epoch 21, batch 2900, loss[loss=0.1698, simple_loss=0.2415, pruned_loss=0.04903, over 16807.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2577, pruned_loss=0.04073, over 3338650.82 frames. ], batch size: 102, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:47:36,636 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205902.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:48:44,777 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:48:45,570 INFO [train.py:904] (2/8) Epoch 21, batch 2950, loss[loss=0.1655, simple_loss=0.2493, pruned_loss=0.0408, over 16241.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2576, pruned_loss=0.04175, over 3329340.03 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:13,839 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3765, 5.3609, 5.1739, 4.6511, 5.1874, 2.0505, 4.9507, 5.1370], device='cuda:2'), covar=tensor([0.0081, 0.0073, 0.0190, 0.0379, 0.0110, 0.2576, 0.0133, 0.0185], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0157, 0.0199, 0.0180, 0.0178, 0.0210, 0.0190, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:49:36,276 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7509, 2.6561, 2.2069, 2.5268, 2.9805, 2.7231, 3.3482, 3.2389], device='cuda:2'), covar=tensor([0.0148, 0.0459, 0.0574, 0.0506, 0.0310, 0.0461, 0.0247, 0.0286], device='cuda:2'), in_proj_covar=tensor([0.0214, 0.0241, 0.0230, 0.0231, 0.0241, 0.0241, 0.0244, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:49:36,948 INFO [optim.py:368] (2/8) 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 05:49:58,057 INFO [train.py:904] (2/8) Epoch 21, batch 3000, loss[loss=0.1455, simple_loss=0.2323, pruned_loss=0.02933, over 16964.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2575, pruned_loss=0.04234, over 3321597.06 frames. ], batch size: 41, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:58,057 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 05:50:06,470 INFO [train.py:938] (2/8) Epoch 21, validation: loss=0.1356, simple_loss=0.2408, pruned_loss=0.01521, over 944034.00 frames. 2023-05-01 05:50:06,470 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 05:50:18,436 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:50:55,659 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-01 05:51:14,522 INFO [train.py:904] (2/8) Epoch 21, batch 3050, loss[loss=0.1712, simple_loss=0.2449, pruned_loss=0.04877, over 16764.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2575, pruned_loss=0.04187, over 3321229.60 frames. ], batch size: 124, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:51:21,083 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206056.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:51:38,268 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7763, 4.8433, 5.0332, 4.8558, 4.8889, 5.4908, 5.0187, 4.6995], device='cuda:2'), covar=tensor([0.1406, 0.2103, 0.2266, 0.2205, 0.2722, 0.1056, 0.1659, 0.2715], device='cuda:2'), in_proj_covar=tensor([0.0412, 0.0604, 0.0662, 0.0500, 0.0663, 0.0694, 0.0520, 0.0667], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 05:52:05,512 INFO [optim.py:368] (2/8) 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 05:52:21,876 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206100.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:52:24,073 INFO [train.py:904] (2/8) Epoch 21, batch 3100, loss[loss=0.171, simple_loss=0.2617, pruned_loss=0.0402, over 17146.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2574, pruned_loss=0.04205, over 3315220.89 frames. ], batch size: 48, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:52:43,668 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206117.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:52:53,120 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-01 05:53:30,990 INFO [train.py:904] (2/8) Epoch 21, batch 3150, loss[loss=0.1917, simple_loss=0.2712, pruned_loss=0.0561, over 16736.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2568, pruned_loss=0.04231, over 3314667.78 frames. ], batch size: 124, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:54:21,467 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-05-01 05:54:22,972 INFO [optim.py:368] (2/8) 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 05:54:34,398 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206197.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:54:41,549 INFO [train.py:904] (2/8) Epoch 21, batch 3200, loss[loss=0.1692, simple_loss=0.2479, pruned_loss=0.04521, over 16496.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2558, pruned_loss=0.04198, over 3311235.86 frames. ], batch size: 146, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:54:56,918 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8471, 2.7616, 2.5733, 4.3316, 3.5466, 4.1588, 1.5969, 3.0557], device='cuda:2'), covar=tensor([0.1356, 0.0797, 0.1213, 0.0217, 0.0241, 0.0403, 0.1672, 0.0818], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0192, 0.0207, 0.0217, 0.0202, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:55:12,665 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7855, 2.8799, 2.5646, 4.8324, 3.9156, 4.2367, 1.5594, 3.2443], device='cuda:2'), covar=tensor([0.1319, 0.0779, 0.1233, 0.0211, 0.0228, 0.0398, 0.1653, 0.0729], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0192, 0.0207, 0.0217, 0.0202, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:55:37,890 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6521, 4.0091, 3.9163, 2.2390, 3.0195, 2.3504, 3.9207, 4.1595], device='cuda:2'), covar=tensor([0.0229, 0.0784, 0.0612, 0.2141, 0.1015, 0.1144, 0.0631, 0.0908], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0169, 0.0153, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 05:55:42,475 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:55:51,281 INFO [train.py:904] (2/8) Epoch 21, batch 3250, loss[loss=0.2072, simple_loss=0.2832, pruned_loss=0.06562, over 16688.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2561, pruned_loss=0.04209, over 3320233.12 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:59,399 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:56:39,011 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6975, 4.6311, 4.5764, 3.9779, 4.6202, 1.7237, 4.3690, 4.3128], device='cuda:2'), covar=tensor([0.0137, 0.0110, 0.0214, 0.0375, 0.0111, 0.2857, 0.0170, 0.0224], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0158, 0.0201, 0.0181, 0.0179, 0.0210, 0.0190, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 05:56:42,634 INFO [optim.py:368] (2/8) 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 05:57:00,581 INFO [train.py:904] (2/8) Epoch 21, batch 3300, loss[loss=0.1772, simple_loss=0.2699, pruned_loss=0.0423, over 16626.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2568, pruned_loss=0.04228, over 3318844.84 frames. ], batch size: 62, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:57:13,826 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206311.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:58:09,885 INFO [train.py:904] (2/8) Epoch 21, batch 3350, loss[loss=0.1824, simple_loss=0.2582, pruned_loss=0.05333, over 16860.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2575, pruned_loss=0.04216, over 3309514.03 frames. ], batch size: 116, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:58:20,159 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:00,090 INFO [optim.py:368] (2/8) 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 05:59:16,563 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:18,364 INFO [train.py:904] (2/8) Epoch 21, batch 3400, loss[loss=0.1579, simple_loss=0.2549, pruned_loss=0.03047, over 17247.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2578, pruned_loss=0.0424, over 3310810.88 frames. ], batch size: 52, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:59:32,334 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:34,950 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9539, 4.6237, 3.3713, 2.4927, 2.8066, 2.8158, 4.9240, 3.7819], device='cuda:2'), covar=tensor([0.2816, 0.0483, 0.1601, 0.2726, 0.2944, 0.1922, 0.0319, 0.1320], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0272, 0.0306, 0.0311, 0.0299, 0.0260, 0.0296, 0.0339], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 06:00:20,856 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-05-01 06:00:21,656 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:00:26,882 INFO [train.py:904] (2/8) Epoch 21, batch 3450, loss[loss=0.167, simple_loss=0.2409, pruned_loss=0.04654, over 16790.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2556, pruned_loss=0.0418, over 3318166.79 frames. ], batch size: 124, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:01:17,176 INFO [optim.py:368] (2/8) 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 06:01:36,857 INFO [train.py:904] (2/8) Epoch 21, batch 3500, loss[loss=0.1592, simple_loss=0.2639, pruned_loss=0.02721, over 17139.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2548, pruned_loss=0.04138, over 3325384.08 frames. ], batch size: 47, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:01:38,491 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2434, 2.9304, 3.1031, 1.9108, 3.2709, 3.2760, 2.6620, 2.5420], device='cuda:2'), covar=tensor([0.0775, 0.0265, 0.0253, 0.1064, 0.0126, 0.0249, 0.0480, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0098, 0.0139, 0.0080, 0.0126, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:02:37,250 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:02:39,134 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8733, 5.2573, 5.0176, 5.0621, 4.8107, 4.8397, 4.7083, 5.3457], device='cuda:2'), covar=tensor([0.1360, 0.0897, 0.1049, 0.0814, 0.0842, 0.0955, 0.1156, 0.0904], device='cuda:2'), in_proj_covar=tensor([0.0681, 0.0840, 0.0690, 0.0632, 0.0535, 0.0538, 0.0702, 0.0651], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:02:44,994 INFO [train.py:904] (2/8) Epoch 21, batch 3550, loss[loss=0.1843, simple_loss=0.266, pruned_loss=0.0513, over 16468.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2542, pruned_loss=0.0408, over 3330342.66 frames. ], batch size: 146, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:47,119 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:09,715 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6521, 3.6739, 3.8926, 2.8666, 3.5476, 3.9910, 3.6289, 2.3391], device='cuda:2'), covar=tensor([0.0514, 0.0208, 0.0068, 0.0357, 0.0112, 0.0102, 0.0096, 0.0518], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0084, 0.0083, 0.0133, 0.0098, 0.0109, 0.0094, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 06:03:37,042 INFO [optim.py:368] (2/8) 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 06:03:44,190 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:55,945 INFO [train.py:904] (2/8) Epoch 21, batch 3600, loss[loss=0.1544, simple_loss=0.2399, pruned_loss=0.03446, over 17207.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2532, pruned_loss=0.04031, over 3332371.25 frames. ], batch size: 44, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:05:06,976 INFO [train.py:904] (2/8) Epoch 21, batch 3650, loss[loss=0.1402, simple_loss=0.2217, pruned_loss=0.02934, over 16795.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2523, pruned_loss=0.04131, over 3317253.18 frames. ], batch size: 39, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:05:07,540 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7244, 3.7450, 2.9206, 2.3140, 2.5095, 2.4742, 3.8915, 3.3675], device='cuda:2'), covar=tensor([0.2500, 0.0568, 0.1593, 0.2899, 0.2721, 0.2009, 0.0453, 0.1309], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0272, 0.0306, 0.0312, 0.0300, 0.0260, 0.0296, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 06:05:59,786 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 21, batch 3700, loss[loss=0.173, simple_loss=0.2459, pruned_loss=0.05002, over 16305.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2513, pruned_loss=0.04279, over 3294949.91 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:06:34,648 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:07:32,563 INFO [train.py:904] (2/8) Epoch 21, batch 3750, loss[loss=0.1908, simple_loss=0.2583, pruned_loss=0.06159, over 16852.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2511, pruned_loss=0.04402, over 3270981.17 frames. ], batch size: 109, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:07:37,722 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4368, 4.4108, 4.3627, 4.1118, 4.1144, 4.4229, 4.1306, 4.1868], device='cuda:2'), covar=tensor([0.0633, 0.0667, 0.0297, 0.0259, 0.0711, 0.0540, 0.0637, 0.0632], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0442, 0.0358, 0.0352, 0.0366, 0.0409, 0.0244, 0.0429], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 06:07:45,098 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 21, batch 3800, loss[loss=0.1882, simple_loss=0.2676, pruned_loss=0.05441, over 16306.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2525, pruned_loss=0.04544, over 3262075.41 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:09:32,843 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206834.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:09:43,611 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2718, 5.3790, 5.1569, 4.7904, 4.5796, 5.2610, 5.0827, 4.8303], device='cuda:2'), covar=tensor([0.0679, 0.0391, 0.0340, 0.0340, 0.1286, 0.0487, 0.0323, 0.0727], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0441, 0.0358, 0.0351, 0.0366, 0.0409, 0.0245, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 06:09:58,315 INFO [train.py:904] (2/8) Epoch 21, batch 3850, loss[loss=0.1653, simple_loss=0.2415, pruned_loss=0.04457, over 16306.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2534, pruned_loss=0.04623, over 3251013.73 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:10:00,769 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:10:52,982 INFO [optim.py:368] (2/8) 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 06:11:00,939 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206895.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:11:05,279 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6288, 4.9480, 4.7460, 4.7500, 4.5133, 4.4477, 4.4245, 5.0429], device='cuda:2'), covar=tensor([0.1275, 0.0801, 0.0893, 0.0768, 0.0806, 0.1191, 0.1058, 0.0792], device='cuda:2'), in_proj_covar=tensor([0.0676, 0.0828, 0.0682, 0.0627, 0.0528, 0.0531, 0.0696, 0.0643], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:11:09,938 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206901.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:11:10,710 INFO [train.py:904] (2/8) Epoch 21, batch 3900, loss[loss=0.1792, simple_loss=0.2565, pruned_loss=0.05093, over 16819.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2535, pruned_loss=0.04693, over 3254302.46 frames. ], batch size: 102, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:12:17,948 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 2023-05-01 06:12:24,775 INFO [train.py:904] (2/8) Epoch 21, batch 3950, loss[loss=0.1736, simple_loss=0.2471, pruned_loss=0.05006, over 16392.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2535, pruned_loss=0.04762, over 3258103.10 frames. ], batch size: 146, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:15,781 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0960, 2.1019, 2.4091, 3.8744, 2.0711, 2.3575, 2.2145, 2.2386], device='cuda:2'), covar=tensor([0.1878, 0.4337, 0.3032, 0.0860, 0.4843, 0.3163, 0.4072, 0.3891], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0452, 0.0371, 0.0334, 0.0438, 0.0521, 0.0420, 0.0529], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:13:16,330 INFO [optim.py:368] (2/8) 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 06:13:35,018 INFO [train.py:904] (2/8) Epoch 21, batch 4000, loss[loss=0.1874, simple_loss=0.2695, pruned_loss=0.05269, over 12136.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2535, pruned_loss=0.04774, over 3269136.81 frames. ], batch size: 245, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:36,246 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8509, 4.9277, 4.7405, 4.4326, 4.4593, 4.8500, 4.6103, 4.5985], device='cuda:2'), covar=tensor([0.0717, 0.0588, 0.0284, 0.0271, 0.0843, 0.0496, 0.0454, 0.0561], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0439, 0.0356, 0.0349, 0.0364, 0.0408, 0.0243, 0.0427], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 06:13:57,628 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5953, 3.5796, 3.5421, 2.8447, 3.4308, 1.9535, 3.2180, 2.8617], device='cuda:2'), covar=tensor([0.0125, 0.0109, 0.0180, 0.0244, 0.0095, 0.2618, 0.0137, 0.0274], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0158, 0.0201, 0.0182, 0.0180, 0.0210, 0.0191, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:14:45,543 INFO [train.py:904] (2/8) Epoch 21, batch 4050, loss[loss=0.1735, simple_loss=0.2599, pruned_loss=0.04355, over 16694.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2542, pruned_loss=0.04686, over 3272171.00 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:15:09,577 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8597, 4.9566, 4.7427, 4.4214, 4.4304, 4.8599, 4.5653, 4.5757], device='cuda:2'), covar=tensor([0.0618, 0.0376, 0.0286, 0.0273, 0.0836, 0.0379, 0.0506, 0.0539], device='cuda:2'), in_proj_covar=tensor([0.0306, 0.0438, 0.0356, 0.0349, 0.0364, 0.0407, 0.0243, 0.0426], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 06:15:37,538 INFO [optim.py:368] (2/8) 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 06:15:55,998 INFO [train.py:904] (2/8) Epoch 21, batch 4100, loss[loss=0.1915, simple_loss=0.2786, pruned_loss=0.05221, over 16918.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2554, pruned_loss=0.04641, over 3266178.64 frames. ], batch size: 109, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:17:10,031 INFO [train.py:904] (2/8) Epoch 21, batch 4150, loss[loss=0.2173, simple_loss=0.2964, pruned_loss=0.06911, over 16795.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2618, pruned_loss=0.04849, over 3235478.54 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:17:28,273 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 2023-05-01 06:17:45,687 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:18:04,520 INFO [optim.py:368] (2/8) 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 06:18:06,770 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207190.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 06:18:24,668 INFO [train.py:904] (2/8) Epoch 21, batch 4200, loss[loss=0.2079, simple_loss=0.3015, pruned_loss=0.05714, over 15243.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2692, pruned_loss=0.05024, over 3205548.47 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:18:45,800 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 06:18:56,378 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7083, 4.8111, 5.0849, 5.0803, 5.1374, 4.8156, 4.7212, 4.5530], device='cuda:2'), covar=tensor([0.0300, 0.0451, 0.0436, 0.0407, 0.0385, 0.0309, 0.1105, 0.0512], device='cuda:2'), in_proj_covar=tensor([0.0411, 0.0458, 0.0443, 0.0411, 0.0489, 0.0465, 0.0553, 0.0372], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 06:18:59,987 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2940, 4.4042, 4.1639, 3.9009, 3.8775, 4.3159, 4.0675, 3.9652], device='cuda:2'), covar=tensor([0.0644, 0.0540, 0.0302, 0.0299, 0.0833, 0.0390, 0.0602, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0428, 0.0348, 0.0342, 0.0356, 0.0398, 0.0237, 0.0416], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:19:01,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0832, 4.0477, 3.8975, 3.1544, 3.9484, 1.8775, 3.7428, 3.5124], device='cuda:2'), covar=tensor([0.0149, 0.0183, 0.0291, 0.0391, 0.0156, 0.2793, 0.0220, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0157, 0.0199, 0.0181, 0.0178, 0.0210, 0.0189, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:19:04,542 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-05-01 06:19:18,508 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:19:40,048 INFO [train.py:904] (2/8) Epoch 21, batch 4250, loss[loss=0.1893, simple_loss=0.2803, pruned_loss=0.04912, over 16479.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2721, pruned_loss=0.04987, over 3187316.95 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:20:25,290 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3476, 5.4221, 5.2361, 4.8881, 4.8982, 5.3210, 5.1138, 4.9502], device='cuda:2'), covar=tensor([0.0590, 0.0352, 0.0255, 0.0257, 0.0845, 0.0383, 0.0280, 0.0641], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0427, 0.0347, 0.0341, 0.0355, 0.0397, 0.0236, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:20:35,819 INFO [optim.py:368] (2/8) 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 06:20:38,823 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6646, 4.7833, 4.5449, 4.2352, 4.1957, 4.6833, 4.4155, 4.3142], device='cuda:2'), covar=tensor([0.0646, 0.0380, 0.0332, 0.0317, 0.1005, 0.0388, 0.0494, 0.0673], device='cuda:2'), in_proj_covar=tensor([0.0298, 0.0426, 0.0346, 0.0340, 0.0354, 0.0396, 0.0236, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:20:55,788 INFO [train.py:904] (2/8) Epoch 21, batch 4300, loss[loss=0.1753, simple_loss=0.2701, pruned_loss=0.0402, over 16900.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.273, pruned_loss=0.04889, over 3195722.86 frames. ], batch size: 109, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:22:07,506 INFO [train.py:904] (2/8) Epoch 21, batch 4350, loss[loss=0.1944, simple_loss=0.2832, pruned_loss=0.05281, over 16316.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2763, pruned_loss=0.05031, over 3189524.83 frames. ], batch size: 165, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:23:02,764 INFO [optim.py:368] (2/8) 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 06:23:22,045 INFO [train.py:904] (2/8) Epoch 21, batch 4400, loss[loss=0.2002, simple_loss=0.2834, pruned_loss=0.0585, over 15477.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2783, pruned_loss=0.05098, over 3187803.03 frames. ], batch size: 191, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:23:55,867 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3516, 3.4910, 2.0371, 3.8710, 2.5733, 3.8608, 2.1986, 2.7276], device='cuda:2'), covar=tensor([0.0303, 0.0338, 0.1757, 0.0156, 0.0895, 0.0470, 0.1548, 0.0765], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0177, 0.0194, 0.0163, 0.0177, 0.0217, 0.0200, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:24:37,195 INFO [train.py:904] (2/8) Epoch 21, batch 4450, loss[loss=0.2153, simple_loss=0.2984, pruned_loss=0.06612, over 16641.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2815, pruned_loss=0.05239, over 3194577.25 frames. ], batch size: 57, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:25:10,192 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 06:25:15,870 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3781, 3.5234, 3.6637, 3.6387, 3.6364, 3.4870, 3.5097, 3.5080], device='cuda:2'), covar=tensor([0.0339, 0.0541, 0.0394, 0.0381, 0.0491, 0.0470, 0.0725, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0449, 0.0436, 0.0404, 0.0480, 0.0458, 0.0546, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 06:25:31,510 INFO [optim.py:368] (2/8) 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 06:25:33,864 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207490.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:25:50,491 INFO [train.py:904] (2/8) Epoch 21, batch 4500, loss[loss=0.1728, simple_loss=0.2608, pruned_loss=0.0424, over 16745.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2824, pruned_loss=0.0533, over 3195143.75 frames. ], batch size: 39, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:26:22,348 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8969, 3.5732, 4.0057, 2.0886, 4.2799, 4.2568, 3.1583, 3.1970], device='cuda:2'), covar=tensor([0.0671, 0.0254, 0.0193, 0.1146, 0.0057, 0.0123, 0.0407, 0.0430], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0097, 0.0138, 0.0080, 0.0125, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:26:35,635 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=207538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:26:59,857 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5549, 2.4486, 2.3231, 3.4511, 2.5907, 3.6440, 1.5419, 2.6248], device='cuda:2'), covar=tensor([0.1493, 0.0930, 0.1348, 0.0198, 0.0228, 0.0436, 0.1856, 0.0942], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0175, 0.0193, 0.0190, 0.0207, 0.0215, 0.0200, 0.0192], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:27:03,812 INFO [train.py:904] (2/8) Epoch 21, batch 4550, loss[loss=0.1959, simple_loss=0.2853, pruned_loss=0.05326, over 16516.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2831, pruned_loss=0.05413, over 3189119.34 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:27:57,547 INFO [optim.py:368] (2/8) 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 06:28:16,259 INFO [train.py:904] (2/8) Epoch 21, batch 4600, loss[loss=0.1849, simple_loss=0.2755, pruned_loss=0.04717, over 16587.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2845, pruned_loss=0.05423, over 3198908.37 frames. ], batch size: 75, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:29:29,137 INFO [train.py:904] (2/8) Epoch 21, batch 4650, loss[loss=0.2017, simple_loss=0.2881, pruned_loss=0.05764, over 16812.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2834, pruned_loss=0.05415, over 3207412.29 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:30:23,479 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 21, batch 4700, loss[loss=0.1585, simple_loss=0.2483, pruned_loss=0.03437, over 16909.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2804, pruned_loss=0.05274, over 3203817.96 frames. ], batch size: 96, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:31:49,518 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0743, 3.2490, 3.3285, 1.9249, 2.7624, 2.1618, 3.4932, 3.4953], device='cuda:2'), covar=tensor([0.0237, 0.0861, 0.0643, 0.2083, 0.0961, 0.1021, 0.0598, 0.0997], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0163, 0.0166, 0.0151, 0.0144, 0.0129, 0.0142, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:31:56,744 INFO [train.py:904] (2/8) Epoch 21, batch 4750, loss[loss=0.1979, simple_loss=0.2813, pruned_loss=0.05729, over 11798.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2759, pruned_loss=0.05063, over 3193850.25 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:32:00,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6571, 2.5691, 1.8216, 2.7402, 2.1374, 2.7689, 2.1186, 2.3605], device='cuda:2'), covar=tensor([0.0313, 0.0348, 0.1353, 0.0190, 0.0657, 0.0431, 0.1241, 0.0621], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0176, 0.0193, 0.0161, 0.0176, 0.0216, 0.0200, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:32:09,535 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0157, 2.5050, 2.5710, 1.8847, 2.7529, 2.8428, 2.4120, 2.3548], device='cuda:2'), covar=tensor([0.0760, 0.0261, 0.0258, 0.1022, 0.0119, 0.0248, 0.0498, 0.0495], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0109, 0.0098, 0.0140, 0.0081, 0.0126, 0.0130, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:32:32,860 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 06:32:50,130 INFO [optim.py:368] (2/8) 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 06:33:11,158 INFO [train.py:904] (2/8) Epoch 21, batch 4800, loss[loss=0.1583, simple_loss=0.2522, pruned_loss=0.03218, over 17234.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2722, pruned_loss=0.04866, over 3192475.92 frames. ], batch size: 52, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:33:34,784 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1851, 4.1755, 4.1230, 3.2389, 4.1033, 1.6098, 3.8566, 3.6570], device='cuda:2'), covar=tensor([0.0121, 0.0123, 0.0159, 0.0433, 0.0098, 0.2930, 0.0163, 0.0295], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0154, 0.0196, 0.0178, 0.0175, 0.0207, 0.0186, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:33:55,601 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:34:24,539 INFO [train.py:904] (2/8) Epoch 21, batch 4850, loss[loss=0.1839, simple_loss=0.2788, pruned_loss=0.04444, over 15467.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2729, pruned_loss=0.04781, over 3187659.91 frames. ], batch size: 191, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:34:31,604 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8591, 5.1494, 4.9251, 4.9820, 4.7222, 4.6628, 4.5377, 5.2252], device='cuda:2'), covar=tensor([0.1189, 0.0762, 0.0874, 0.0761, 0.0709, 0.0931, 0.1124, 0.0735], device='cuda:2'), in_proj_covar=tensor([0.0659, 0.0804, 0.0668, 0.0610, 0.0512, 0.0519, 0.0676, 0.0628], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:34:36,124 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 06:34:59,471 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1219, 2.9498, 3.1225, 1.7151, 3.2957, 3.3505, 2.6755, 2.4660], device='cuda:2'), covar=tensor([0.0914, 0.0290, 0.0205, 0.1303, 0.0093, 0.0171, 0.0497, 0.0576], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0110, 0.0098, 0.0140, 0.0081, 0.0126, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:35:08,160 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=207880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:35:22,068 INFO [optim.py:368] (2/8) 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 06:35:40,335 INFO [train.py:904] (2/8) Epoch 21, batch 4900, loss[loss=0.1858, simple_loss=0.2787, pruned_loss=0.04644, over 16901.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2729, pruned_loss=0.04692, over 3171233.74 frames. ], batch size: 109, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:36:29,999 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9380, 3.1018, 3.1257, 1.6606, 3.3099, 3.4472, 2.7673, 2.4164], device='cuda:2'), covar=tensor([0.1195, 0.0218, 0.0189, 0.1390, 0.0101, 0.0157, 0.0473, 0.0642], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0109, 0.0098, 0.0140, 0.0081, 0.0125, 0.0130, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:36:52,780 INFO [train.py:904] (2/8) Epoch 21, batch 4950, loss[loss=0.191, simple_loss=0.288, pruned_loss=0.04697, over 15175.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2725, pruned_loss=0.04608, over 3176580.22 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:36:56,218 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 06:37:47,806 INFO [optim.py:368] (2/8) 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 06:38:01,916 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 06:38:08,413 INFO [train.py:904] (2/8) Epoch 21, batch 5000, loss[loss=0.1767, simple_loss=0.277, pruned_loss=0.03815, over 16401.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2746, pruned_loss=0.04648, over 3155738.76 frames. ], batch size: 75, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:38:52,044 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208032.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:39:21,546 INFO [train.py:904] (2/8) Epoch 21, batch 5050, loss[loss=0.1787, simple_loss=0.272, pruned_loss=0.04266, over 16690.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2752, pruned_loss=0.04638, over 3163991.18 frames. ], batch size: 134, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:40:18,551 INFO [optim.py:368] (2/8) 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 06:40:23,275 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208093.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 06:40:35,284 INFO [train.py:904] (2/8) Epoch 21, batch 5100, loss[loss=0.1885, simple_loss=0.2772, pruned_loss=0.04985, over 16786.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2734, pruned_loss=0.04565, over 3191589.84 frames. ], batch size: 124, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:41:48,626 INFO [train.py:904] (2/8) Epoch 21, batch 5150, loss[loss=0.1796, simple_loss=0.2675, pruned_loss=0.04584, over 16583.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2737, pruned_loss=0.04514, over 3190385.32 frames. ], batch size: 62, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:42:43,673 INFO [optim.py:368] (2/8) 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 06:43:01,061 INFO [train.py:904] (2/8) Epoch 21, batch 5200, loss[loss=0.1852, simple_loss=0.2763, pruned_loss=0.04708, over 16691.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.272, pruned_loss=0.04471, over 3189232.74 frames. ], batch size: 134, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:11,865 INFO [train.py:904] (2/8) Epoch 21, batch 5250, loss[loss=0.1751, simple_loss=0.2656, pruned_loss=0.04231, over 15393.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2698, pruned_loss=0.04439, over 3199250.12 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:17,821 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 06:44:31,792 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:45:07,260 INFO [optim.py:368] (2/8) 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 06:45:25,337 INFO [train.py:904] (2/8) Epoch 21, batch 5300, loss[loss=0.1452, simple_loss=0.233, pruned_loss=0.02869, over 16858.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2666, pruned_loss=0.04336, over 3184765.35 frames. ], batch size: 102, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:45:27,979 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-05-01 06:46:00,440 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208326.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:46:37,966 INFO [train.py:904] (2/8) Epoch 21, batch 5350, loss[loss=0.1919, simple_loss=0.2816, pruned_loss=0.05109, over 16900.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2643, pruned_loss=0.04258, over 3196373.83 frames. ], batch size: 116, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:47:04,322 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.58 vs. limit=5.0 2023-05-01 06:47:32,335 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208388.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:47:34,960 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 21, batch 5400, loss[loss=0.1909, simple_loss=0.288, pruned_loss=0.04695, over 16723.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2673, pruned_loss=0.04352, over 3190036.46 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:49:10,973 INFO [train.py:904] (2/8) Epoch 21, batch 5450, loss[loss=0.2229, simple_loss=0.3049, pruned_loss=0.0705, over 16906.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2705, pruned_loss=0.04496, over 3188405.98 frames. ], batch size: 116, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:50:09,145 INFO [optim.py:368] (2/8) 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 06:50:28,299 INFO [train.py:904] (2/8) Epoch 21, batch 5500, loss[loss=0.2595, simple_loss=0.3252, pruned_loss=0.09687, over 11428.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2769, pruned_loss=0.04919, over 3157770.60 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:51:47,645 INFO [train.py:904] (2/8) Epoch 21, batch 5550, loss[loss=0.2359, simple_loss=0.317, pruned_loss=0.07746, over 16226.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2841, pruned_loss=0.05473, over 3116204.60 frames. ], batch size: 165, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:51:51,211 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2401, 5.5529, 5.2578, 5.3064, 5.0205, 4.9610, 4.9607, 5.6387], device='cuda:2'), covar=tensor([0.1160, 0.0770, 0.0892, 0.0906, 0.0810, 0.0759, 0.1117, 0.0819], device='cuda:2'), in_proj_covar=tensor([0.0654, 0.0796, 0.0664, 0.0603, 0.0506, 0.0512, 0.0667, 0.0622], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:52:27,805 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208576.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:52:49,283 INFO [optim.py:368] (2/8) 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 06:53:07,711 INFO [train.py:904] (2/8) Epoch 21, batch 5600, loss[loss=0.2509, simple_loss=0.3136, pruned_loss=0.09408, over 11155.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2893, pruned_loss=0.05897, over 3080318.18 frames. ], batch size: 249, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:53:22,834 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7963, 1.4394, 1.7611, 1.7294, 1.8636, 1.8751, 1.6778, 1.8357], device='cuda:2'), covar=tensor([0.0205, 0.0309, 0.0179, 0.0235, 0.0226, 0.0149, 0.0367, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0191, 0.0178, 0.0184, 0.0195, 0.0151, 0.0195, 0.0150], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:53:37,213 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7978, 2.0356, 2.4073, 3.0851, 2.1348, 2.2147, 2.2785, 2.1555], device='cuda:2'), covar=tensor([0.1667, 0.3799, 0.2276, 0.0789, 0.4495, 0.2691, 0.3338, 0.3503], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0440, 0.0361, 0.0323, 0.0428, 0.0508, 0.0410, 0.0514], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:53:39,476 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:06,475 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208637.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:29,865 INFO [train.py:904] (2/8) Epoch 21, batch 5650, loss[loss=0.2193, simple_loss=0.302, pruned_loss=0.06826, over 16923.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2947, pruned_loss=0.06347, over 3039520.39 frames. ], batch size: 109, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:54:30,486 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0129, 3.3567, 3.4540, 2.0519, 2.9163, 2.2756, 3.4427, 3.6383], device='cuda:2'), covar=tensor([0.0288, 0.0782, 0.0584, 0.2041, 0.0863, 0.0994, 0.0626, 0.0947], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0163, 0.0166, 0.0151, 0.0144, 0.0130, 0.0143, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 06:55:12,922 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2930, 4.3571, 4.1762, 3.9155, 3.9122, 4.2739, 3.9928, 4.0177], device='cuda:2'), covar=tensor([0.0598, 0.0479, 0.0310, 0.0303, 0.0767, 0.0496, 0.0739, 0.0621], device='cuda:2'), in_proj_covar=tensor([0.0288, 0.0418, 0.0337, 0.0334, 0.0347, 0.0389, 0.0231, 0.0405], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:55:28,096 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:55:30,665 INFO [optim.py:368] (2/8) 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 06:55:41,255 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 06:55:48,282 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7492, 4.8538, 5.2106, 5.1873, 5.2051, 4.8534, 4.8390, 4.6362], device='cuda:2'), covar=tensor([0.0325, 0.0508, 0.0378, 0.0392, 0.0479, 0.0416, 0.0968, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0445, 0.0431, 0.0400, 0.0474, 0.0454, 0.0541, 0.0363], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 06:55:50,854 INFO [train.py:904] (2/8) Epoch 21, batch 5700, loss[loss=0.2073, simple_loss=0.2945, pruned_loss=0.06002, over 16814.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2952, pruned_loss=0.06389, over 3062489.59 frames. ], batch size: 83, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:56:24,744 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-01 06:56:45,710 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:57:10,131 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8789, 4.1692, 3.9872, 4.0146, 3.6927, 3.7471, 3.8331, 4.1575], device='cuda:2'), covar=tensor([0.1110, 0.0837, 0.1039, 0.0871, 0.0760, 0.1671, 0.0905, 0.0907], device='cuda:2'), in_proj_covar=tensor([0.0649, 0.0789, 0.0660, 0.0599, 0.0502, 0.0509, 0.0663, 0.0617], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 06:57:10,969 INFO [train.py:904] (2/8) Epoch 21, batch 5750, loss[loss=0.24, simple_loss=0.3023, pruned_loss=0.08886, over 11299.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.298, pruned_loss=0.06531, over 3053913.58 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:58:13,921 INFO [optim.py:368] (2/8) 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 06:58:33,856 INFO [train.py:904] (2/8) Epoch 21, batch 5800, loss[loss=0.2068, simple_loss=0.2937, pruned_loss=0.05995, over 15280.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2977, pruned_loss=0.0648, over 3036744.13 frames. ], batch size: 191, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:59:00,632 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.98 vs. limit=5.0 2023-05-01 06:59:52,593 INFO [train.py:904] (2/8) Epoch 21, batch 5850, loss[loss=0.1967, simple_loss=0.2709, pruned_loss=0.06121, over 11252.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2949, pruned_loss=0.06264, over 3057902.20 frames. ], batch size: 247, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:00:53,525 INFO [optim.py:368] (2/8) 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 07:01:12,710 INFO [train.py:904] (2/8) Epoch 21, batch 5900, loss[loss=0.1911, simple_loss=0.2876, pruned_loss=0.04729, over 16504.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2947, pruned_loss=0.06202, over 3088909.89 frames. ], batch size: 68, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:01:48,466 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:04,489 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208932.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:36,220 INFO [train.py:904] (2/8) Epoch 21, batch 5950, loss[loss=0.1992, simple_loss=0.2881, pruned_loss=0.05511, over 16945.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2958, pruned_loss=0.06134, over 3073109.26 frames. ], batch size: 109, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:02,962 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:03:36,548 INFO [optim.py:368] (2/8) 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 07:03:56,982 INFO [train.py:904] (2/8) Epoch 21, batch 6000, loss[loss=0.1979, simple_loss=0.2856, pruned_loss=0.05508, over 15414.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.295, pruned_loss=0.06105, over 3069308.16 frames. ], batch size: 190, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:56,982 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 07:04:08,263 INFO [train.py:938] (2/8) Epoch 21, validation: loss=0.1512, simple_loss=0.2639, pruned_loss=0.01924, over 944034.00 frames. 2023-05-01 07:04:08,263 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 07:05:25,681 INFO [train.py:904] (2/8) Epoch 21, batch 6050, loss[loss=0.2176, simple_loss=0.2913, pruned_loss=0.07194, over 11407.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.293, pruned_loss=0.06031, over 3087900.15 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:05:46,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7534, 3.7087, 3.9365, 3.7237, 3.8796, 4.2650, 3.9162, 3.6588], device='cuda:2'), covar=tensor([0.2317, 0.2435, 0.2404, 0.2505, 0.2790, 0.1800, 0.1641, 0.2640], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0585, 0.0640, 0.0488, 0.0646, 0.0673, 0.0503, 0.0652], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:06:26,989 INFO [optim.py:368] (2/8) 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 07:06:45,920 INFO [train.py:904] (2/8) Epoch 21, batch 6100, loss[loss=0.2143, simple_loss=0.2985, pruned_loss=0.06507, over 17024.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2928, pruned_loss=0.05939, over 3103173.80 frames. ], batch size: 55, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:07:06,461 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 07:07:26,839 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6628, 2.5980, 1.8723, 2.7325, 2.2126, 2.7999, 2.1239, 2.3546], device='cuda:2'), covar=tensor([0.0307, 0.0362, 0.1242, 0.0244, 0.0606, 0.0476, 0.1133, 0.0579], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0174, 0.0192, 0.0160, 0.0175, 0.0213, 0.0199, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 07:08:04,716 INFO [train.py:904] (2/8) Epoch 21, batch 6150, loss[loss=0.2311, simple_loss=0.3031, pruned_loss=0.07951, over 11589.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.291, pruned_loss=0.0588, over 3103427.36 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:04,298 INFO [optim.py:368] (2/8) 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 07:09:23,293 INFO [train.py:904] (2/8) Epoch 21, batch 6200, loss[loss=0.1973, simple_loss=0.2891, pruned_loss=0.05273, over 16311.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2894, pruned_loss=0.05883, over 3089443.45 frames. ], batch size: 68, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:23,845 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:10,133 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 21, batch 6250, loss[loss=0.2034, simple_loss=0.2963, pruned_loss=0.05525, over 16883.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2887, pruned_loss=0.05831, over 3103800.41 frames. ], batch size: 116, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:10:45,643 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:57,621 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209263.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:22,155 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209280.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:32,664 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-01 07:11:35,896 INFO [optim.py:368] (2/8) 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 07:11:49,878 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209298.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:51,746 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3508, 3.3623, 2.0832, 3.6810, 2.5393, 3.7027, 2.1889, 2.7100], device='cuda:2'), covar=tensor([0.0295, 0.0390, 0.1584, 0.0260, 0.0817, 0.0636, 0.1570, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0161, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 07:11:54,988 INFO [train.py:904] (2/8) Epoch 21, batch 6300, loss[loss=0.1883, simple_loss=0.2736, pruned_loss=0.05153, over 16744.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2886, pruned_loss=0.05747, over 3118095.80 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:11:55,534 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0669, 2.4880, 2.5598, 1.9552, 2.7282, 2.7878, 2.4032, 2.3623], device='cuda:2'), covar=tensor([0.0687, 0.0246, 0.0243, 0.0888, 0.0106, 0.0270, 0.0456, 0.0448], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0107, 0.0096, 0.0137, 0.0079, 0.0123, 0.0127, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 07:12:18,023 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:13:13,581 INFO [train.py:904] (2/8) Epoch 21, batch 6350, loss[loss=0.185, simple_loss=0.2752, pruned_loss=0.04742, over 16790.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2894, pruned_loss=0.0586, over 3100263.68 frames. ], batch size: 102, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:13:24,838 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:13:41,244 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-05-01 07:14:13,986 INFO [optim.py:368] (2/8) 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 07:14:31,817 INFO [train.py:904] (2/8) Epoch 21, batch 6400, loss[loss=0.1911, simple_loss=0.2815, pruned_loss=0.0504, over 16692.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2894, pruned_loss=0.0591, over 3103561.08 frames. ], batch size: 124, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:15:33,809 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6978, 4.9683, 5.1158, 4.8942, 4.9096, 5.4872, 4.9724, 4.7395], device='cuda:2'), covar=tensor([0.1207, 0.1821, 0.2441, 0.1994, 0.2633, 0.0912, 0.1525, 0.2331], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0590, 0.0648, 0.0493, 0.0654, 0.0682, 0.0508, 0.0660], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:15:48,008 INFO [train.py:904] (2/8) Epoch 21, batch 6450, loss[loss=0.2041, simple_loss=0.3074, pruned_loss=0.05039, over 16856.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2903, pruned_loss=0.05864, over 3102595.68 frames. ], batch size: 102, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:16:10,759 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209466.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:16:15,973 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2023-05-01 07:16:52,760 INFO [optim.py:368] (2/8) 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 07:17:08,455 INFO [train.py:904] (2/8) Epoch 21, batch 6500, loss[loss=0.2145, simple_loss=0.2812, pruned_loss=0.07395, over 11440.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2883, pruned_loss=0.05805, over 3105161.65 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:17:20,787 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:17:48,025 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209527.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:17:58,990 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-05-01 07:18:26,785 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3116, 1.6325, 1.9140, 2.2230, 2.2939, 2.4827, 1.7309, 2.4729], device='cuda:2'), covar=tensor([0.0206, 0.0515, 0.0329, 0.0350, 0.0360, 0.0205, 0.0562, 0.0144], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0192, 0.0176, 0.0182, 0.0193, 0.0150, 0.0194, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:18:29,350 INFO [train.py:904] (2/8) Epoch 21, batch 6550, loss[loss=0.2019, simple_loss=0.3015, pruned_loss=0.05109, over 16688.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2911, pruned_loss=0.05909, over 3087682.99 frames. ], batch size: 76, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:18:40,051 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:00,537 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209571.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:33,744 INFO [optim.py:368] (2/8) 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 07:19:49,515 INFO [train.py:904] (2/8) Epoch 21, batch 6600, loss[loss=0.2221, simple_loss=0.2954, pruned_loss=0.07441, over 11610.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2927, pruned_loss=0.05939, over 3079946.09 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:19:55,979 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:20:03,683 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:20:48,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 07:21:07,995 INFO [train.py:904] (2/8) Epoch 21, batch 6650, loss[loss=0.1827, simple_loss=0.2722, pruned_loss=0.04665, over 16834.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2928, pruned_loss=0.05995, over 3106326.94 frames. ], batch size: 83, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:21:12,194 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209654.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:32,362 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209667.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:04,990 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:11,939 INFO [optim.py:368] (2/8) 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 07:22:25,641 INFO [train.py:904] (2/8) Epoch 21, batch 6700, loss[loss=0.1936, simple_loss=0.2844, pruned_loss=0.05144, over 16888.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2913, pruned_loss=0.05969, over 3110869.57 frames. ], batch size: 96, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:23:36,312 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 07:23:40,239 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209749.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:23:44,113 INFO [train.py:904] (2/8) Epoch 21, batch 6750, loss[loss=0.2354, simple_loss=0.3131, pruned_loss=0.07885, over 12003.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2896, pruned_loss=0.05964, over 3121533.33 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 2.0 2023-05-01 07:24:47,245 INFO [optim.py:368] (2/8) 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 07:25:01,553 INFO [train.py:904] (2/8) Epoch 21, batch 6800, loss[loss=0.225, simple_loss=0.3002, pruned_loss=0.07485, over 11396.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2899, pruned_loss=0.06012, over 3101854.14 frames. ], batch size: 247, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:25:11,599 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8049, 2.6554, 2.6199, 1.8712, 2.4998, 2.6713, 2.5340, 1.9138], device='cuda:2'), covar=tensor([0.0448, 0.0087, 0.0090, 0.0398, 0.0142, 0.0138, 0.0127, 0.0388], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0082, 0.0083, 0.0134, 0.0097, 0.0108, 0.0093, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:25:33,962 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:11,992 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 07:26:12,938 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6094, 2.5486, 1.8769, 2.6796, 2.1693, 2.7666, 2.1107, 2.3185], device='cuda:2'), covar=tensor([0.0333, 0.0397, 0.1450, 0.0260, 0.0707, 0.0503, 0.1272, 0.0645], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0161, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 07:26:20,406 INFO [train.py:904] (2/8) Epoch 21, batch 6850, loss[loss=0.2246, simple_loss=0.3071, pruned_loss=0.07107, over 15478.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2912, pruned_loss=0.06081, over 3086262.57 frames. ], batch size: 191, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:26:28,998 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:29,167 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6893, 3.0130, 3.1689, 1.8987, 2.7907, 2.1319, 3.2432, 3.2786], device='cuda:2'), covar=tensor([0.0275, 0.0846, 0.0663, 0.2150, 0.0895, 0.1094, 0.0672, 0.0894], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0164, 0.0168, 0.0153, 0.0145, 0.0129, 0.0144, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 07:26:42,201 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:51,354 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8417, 4.8149, 4.6988, 3.8835, 4.7000, 1.8436, 4.4249, 4.3436], device='cuda:2'), covar=tensor([0.0127, 0.0123, 0.0187, 0.0404, 0.0126, 0.2687, 0.0217, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0151, 0.0194, 0.0175, 0.0172, 0.0203, 0.0183, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:27:21,614 INFO [optim.py:368] (2/8) 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 07:27:25,265 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8094, 4.2521, 4.3240, 3.0995, 3.9872, 4.3282, 3.9544, 2.4322], device='cuda:2'), covar=tensor([0.0457, 0.0051, 0.0044, 0.0328, 0.0075, 0.0113, 0.0075, 0.0434], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0082, 0.0083, 0.0134, 0.0097, 0.0108, 0.0093, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:27:34,652 INFO [train.py:904] (2/8) Epoch 21, batch 6900, loss[loss=0.1984, simple_loss=0.2902, pruned_loss=0.05332, over 16499.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2929, pruned_loss=0.05941, over 3108925.63 frames. ], batch size: 68, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:27:41,303 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:48,440 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:28:32,284 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 07:28:51,082 INFO [train.py:904] (2/8) Epoch 21, batch 6950, loss[loss=0.2504, simple_loss=0.3086, pruned_loss=0.09605, over 11230.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2954, pruned_loss=0.06149, over 3096992.53 frames. ], batch size: 247, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:28:53,937 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:28:55,904 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4084, 2.1849, 1.8278, 1.9343, 2.4943, 2.1897, 2.2357, 2.6155], device='cuda:2'), covar=tensor([0.0218, 0.0402, 0.0528, 0.0474, 0.0243, 0.0371, 0.0202, 0.0247], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0229, 0.0222, 0.0222, 0.0231, 0.0228, 0.0229, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:29:02,106 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:51,930 INFO [optim.py:368] (2/8) 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 07:30:09,077 INFO [train.py:904] (2/8) Epoch 21, batch 7000, loss[loss=0.1775, simple_loss=0.2776, pruned_loss=0.03872, over 16812.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2954, pruned_loss=0.0606, over 3110214.98 frames. ], batch size: 102, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:30:09,334 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210002.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:31:11,191 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:31:22,808 INFO [train.py:904] (2/8) Epoch 21, batch 7050, loss[loss=0.2202, simple_loss=0.291, pruned_loss=0.0747, over 11287.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2963, pruned_loss=0.06085, over 3098158.67 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:31:29,223 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5239, 4.5423, 4.9242, 4.9216, 4.9364, 4.5857, 4.5479, 4.3832], device='cuda:2'), covar=tensor([0.0331, 0.0566, 0.0382, 0.0377, 0.0528, 0.0417, 0.0996, 0.0532], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0448, 0.0436, 0.0404, 0.0484, 0.0458, 0.0544, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 07:31:30,582 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:32:03,258 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5307, 4.1401, 4.1075, 2.7516, 3.6703, 4.1266, 3.6108, 2.2644], device='cuda:2'), covar=tensor([0.0483, 0.0040, 0.0044, 0.0383, 0.0102, 0.0094, 0.0095, 0.0439], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0082, 0.0082, 0.0133, 0.0096, 0.0108, 0.0093, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:32:20,599 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3281, 3.8602, 3.8359, 2.5261, 3.5297, 3.8644, 3.4597, 2.0969], device='cuda:2'), covar=tensor([0.0507, 0.0049, 0.0055, 0.0403, 0.0102, 0.0100, 0.0101, 0.0468], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0133, 0.0096, 0.0108, 0.0093, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:32:24,158 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 21, batch 7100, loss[loss=0.2203, simple_loss=0.3042, pruned_loss=0.06822, over 16757.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2953, pruned_loss=0.0613, over 3087491.80 frames. ], batch size: 124, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:33:03,221 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:09,280 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:55,430 INFO [train.py:904] (2/8) Epoch 21, batch 7150, loss[loss=0.2352, simple_loss=0.2965, pruned_loss=0.08692, over 11274.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2935, pruned_loss=0.06132, over 3080640.01 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:34:01,883 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1717, 1.5419, 1.8655, 2.0351, 2.2404, 2.3749, 1.7425, 2.3405], device='cuda:2'), covar=tensor([0.0229, 0.0470, 0.0315, 0.0352, 0.0295, 0.0204, 0.0488, 0.0147], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0181, 0.0193, 0.0151, 0.0193, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:34:16,394 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:21,260 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210170.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:53,945 INFO [optim.py:368] (2/8) 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 07:35:07,915 INFO [train.py:904] (2/8) Epoch 21, batch 7200, loss[loss=0.1914, simple_loss=0.2833, pruned_loss=0.04975, over 16835.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2912, pruned_loss=0.05968, over 3083108.36 frames. ], batch size: 116, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:35:26,243 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:36:28,284 INFO [train.py:904] (2/8) Epoch 21, batch 7250, loss[loss=0.2194, simple_loss=0.2855, pruned_loss=0.07665, over 11446.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2894, pruned_loss=0.05896, over 3060201.21 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:36:43,717 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210262.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:37:31,774 INFO [optim.py:368] (2/8) 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 07:37:41,337 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5003, 4.7185, 4.8830, 4.6776, 4.7390, 5.1931, 4.7047, 4.4708], device='cuda:2'), covar=tensor([0.1395, 0.1719, 0.1844, 0.1723, 0.2146, 0.0878, 0.1718, 0.2553], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0582, 0.0640, 0.0485, 0.0642, 0.0671, 0.0501, 0.0652], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:37:44,654 INFO [train.py:904] (2/8) Epoch 21, batch 7300, loss[loss=0.2024, simple_loss=0.2931, pruned_loss=0.05585, over 16683.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2884, pruned_loss=0.05868, over 3059269.50 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:37:58,791 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:10,836 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2940, 4.1714, 4.3785, 4.4756, 4.6061, 4.1265, 4.5686, 4.6260], device='cuda:2'), covar=tensor([0.1692, 0.1180, 0.1321, 0.0646, 0.0516, 0.1246, 0.0722, 0.0571], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0766, 0.0889, 0.0781, 0.0588, 0.0616, 0.0636, 0.0732], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:38:36,022 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2590, 2.0130, 1.7068, 1.6439, 2.2653, 1.8564, 2.0863, 2.3362], device='cuda:2'), covar=tensor([0.0205, 0.0353, 0.0493, 0.0480, 0.0222, 0.0348, 0.0195, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0201, 0.0229, 0.0221, 0.0222, 0.0229, 0.0228, 0.0228, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:38:43,526 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6028, 3.0375, 3.1767, 2.0726, 2.7287, 2.1147, 3.1816, 3.2755], device='cuda:2'), covar=tensor([0.0250, 0.0805, 0.0583, 0.1958, 0.0901, 0.1001, 0.0653, 0.0869], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0143, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 07:38:48,757 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210344.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:59,602 INFO [train.py:904] (2/8) Epoch 21, batch 7350, loss[loss=0.1941, simple_loss=0.2818, pruned_loss=0.0532, over 16699.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2897, pruned_loss=0.05913, over 3073370.06 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:39:36,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5401, 4.5534, 4.4071, 3.6675, 4.4649, 1.6772, 4.2386, 4.1376], device='cuda:2'), covar=tensor([0.0110, 0.0086, 0.0196, 0.0357, 0.0097, 0.2856, 0.0140, 0.0239], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0149, 0.0192, 0.0173, 0.0170, 0.0202, 0.0181, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:40:00,127 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:40:02,082 INFO [optim.py:368] (2/8) 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 07:40:14,853 INFO [train.py:904] (2/8) Epoch 21, batch 7400, loss[loss=0.1958, simple_loss=0.2914, pruned_loss=0.05004, over 16773.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2902, pruned_loss=0.05947, over 3062865.51 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:40:32,235 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210413.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:40:56,179 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2042, 2.1596, 2.2633, 3.9096, 2.1238, 2.6000, 2.2877, 2.3215], device='cuda:2'), covar=tensor([0.1342, 0.3514, 0.2861, 0.0507, 0.4221, 0.2335, 0.3378, 0.3370], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0445, 0.0362, 0.0324, 0.0435, 0.0511, 0.0414, 0.0518], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:41:32,223 INFO [train.py:904] (2/8) Epoch 21, batch 7450, loss[loss=0.2405, simple_loss=0.3066, pruned_loss=0.0872, over 11412.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.291, pruned_loss=0.06003, over 3061048.07 frames. ], batch size: 247, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:42:11,574 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4309, 2.8842, 3.0921, 1.9616, 2.7772, 2.0790, 3.0399, 3.0938], device='cuda:2'), covar=tensor([0.0315, 0.0830, 0.0574, 0.2057, 0.0821, 0.1036, 0.0684, 0.0876], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0143, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 07:42:42,641 INFO [optim.py:368] (2/8) 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 07:42:53,254 INFO [train.py:904] (2/8) Epoch 21, batch 7500, loss[loss=0.185, simple_loss=0.2692, pruned_loss=0.0504, over 16476.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2905, pruned_loss=0.05885, over 3069700.58 frames. ], batch size: 68, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:44:09,156 INFO [train.py:904] (2/8) Epoch 21, batch 7550, loss[loss=0.1892, simple_loss=0.2794, pruned_loss=0.04952, over 16827.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2897, pruned_loss=0.0592, over 3080478.11 frames. ], batch size: 102, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:44:16,116 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8968, 2.1106, 2.2361, 3.4174, 2.0600, 2.4171, 2.2245, 2.2682], device='cuda:2'), covar=tensor([0.1457, 0.3586, 0.2940, 0.0650, 0.4299, 0.2441, 0.3495, 0.3189], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0446, 0.0363, 0.0325, 0.0436, 0.0512, 0.0415, 0.0519], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:44:54,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6648, 2.6748, 2.2632, 4.3762, 3.1746, 3.8940, 1.5633, 2.8715], device='cuda:2'), covar=tensor([0.1437, 0.0823, 0.1485, 0.0184, 0.0300, 0.0444, 0.1718, 0.0914], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0196, 0.0190, 0.0207, 0.0216, 0.0202, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 07:45:11,347 INFO [optim.py:368] (2/8) 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 07:45:13,437 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0595, 5.0829, 4.9010, 4.4943, 4.5671, 4.9829, 4.8565, 4.6505], device='cuda:2'), covar=tensor([0.0626, 0.0465, 0.0302, 0.0320, 0.1052, 0.0474, 0.0326, 0.0606], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0406, 0.0326, 0.0322, 0.0337, 0.0376, 0.0226, 0.0392], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:45:23,227 INFO [train.py:904] (2/8) Epoch 21, batch 7600, loss[loss=0.228, simple_loss=0.3053, pruned_loss=0.07532, over 15202.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2895, pruned_loss=0.05965, over 3081948.41 frames. ], batch size: 190, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:46:08,277 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210632.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:46:27,010 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6447, 2.7341, 2.2211, 2.5051, 3.1475, 2.7848, 3.3044, 3.3384], device='cuda:2'), covar=tensor([0.0100, 0.0417, 0.0550, 0.0444, 0.0241, 0.0349, 0.0190, 0.0226], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0230, 0.0223, 0.0222, 0.0230, 0.0228, 0.0229, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:46:37,228 INFO [train.py:904] (2/8) Epoch 21, batch 7650, loss[loss=0.2146, simple_loss=0.3043, pruned_loss=0.0624, over 16797.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2903, pruned_loss=0.06005, over 3099744.59 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:46:59,455 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210666.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:01,938 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1013, 2.1487, 2.6088, 2.9724, 2.9092, 3.5318, 2.1600, 3.4420], device='cuda:2'), covar=tensor([0.0194, 0.0442, 0.0313, 0.0285, 0.0279, 0.0139, 0.0509, 0.0120], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0182, 0.0194, 0.0150, 0.0194, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:47:25,372 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,068 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0198, 4.0845, 3.9395, 3.6775, 3.6781, 4.0278, 3.7014, 3.7828], device='cuda:2'), covar=tensor([0.0670, 0.0722, 0.0323, 0.0296, 0.0809, 0.0609, 0.1039, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0408, 0.0327, 0.0322, 0.0337, 0.0377, 0.0227, 0.0393], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:47:42,105 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,757 INFO [optim.py:368] (2/8) 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 07:47:55,351 INFO [train.py:904] (2/8) Epoch 21, batch 7700, loss[loss=0.1892, simple_loss=0.28, pruned_loss=0.04922, over 16265.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2893, pruned_loss=0.05957, over 3109070.98 frames. ], batch size: 165, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:48:12,295 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:17,263 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3243, 3.4731, 3.6172, 3.5843, 3.6090, 3.4309, 3.4501, 3.4928], device='cuda:2'), covar=tensor([0.0423, 0.0711, 0.0486, 0.0498, 0.0535, 0.0547, 0.0845, 0.0595], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0444, 0.0432, 0.0402, 0.0478, 0.0453, 0.0539, 0.0363], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 07:48:34,098 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:46,546 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1135, 5.1761, 5.5725, 5.5297, 5.5718, 5.2247, 5.1665, 4.9037], device='cuda:2'), covar=tensor([0.0332, 0.0477, 0.0331, 0.0404, 0.0481, 0.0371, 0.0963, 0.0501], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0445, 0.0432, 0.0403, 0.0479, 0.0454, 0.0540, 0.0364], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 07:48:53,637 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-01 07:48:58,959 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:12,551 INFO [train.py:904] (2/8) Epoch 21, batch 7750, loss[loss=0.1894, simple_loss=0.2743, pruned_loss=0.05222, over 16775.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2892, pruned_loss=0.05907, over 3111480.62 frames. ], batch size: 39, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:49:27,324 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210761.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:29,639 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7900, 1.3457, 1.7066, 1.6526, 1.7970, 1.9183, 1.6265, 1.8034], device='cuda:2'), covar=tensor([0.0269, 0.0390, 0.0225, 0.0304, 0.0249, 0.0177, 0.0423, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0181, 0.0193, 0.0150, 0.0194, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:49:32,420 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210764.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:50:16,713 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7528, 3.6196, 3.8211, 3.9372, 4.0054, 3.6555, 3.9646, 4.0344], device='cuda:2'), covar=tensor([0.1756, 0.1360, 0.1529, 0.0846, 0.0799, 0.2182, 0.1076, 0.0924], device='cuda:2'), in_proj_covar=tensor([0.0620, 0.0763, 0.0885, 0.0778, 0.0585, 0.0614, 0.0634, 0.0731], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:50:18,619 INFO [optim.py:368] (2/8) 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 07:50:25,319 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1713, 3.6540, 3.6169, 2.2067, 3.3829, 3.7223, 3.4585, 1.6291], device='cuda:2'), covar=tensor([0.0673, 0.0104, 0.0111, 0.0576, 0.0141, 0.0196, 0.0147, 0.0764], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0082, 0.0082, 0.0133, 0.0096, 0.0108, 0.0093, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:50:31,005 INFO [train.py:904] (2/8) Epoch 21, batch 7800, loss[loss=0.2031, simple_loss=0.2941, pruned_loss=0.056, over 16649.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2901, pruned_loss=0.05975, over 3111348.55 frames. ], batch size: 57, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:51:07,575 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210825.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:51:48,910 INFO [train.py:904] (2/8) Epoch 21, batch 7850, loss[loss=0.18, simple_loss=0.2813, pruned_loss=0.03934, over 16784.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2914, pruned_loss=0.06018, over 3096907.62 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:52:18,955 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3992, 3.3599, 3.4452, 3.5144, 3.5462, 3.2915, 3.5055, 3.5958], device='cuda:2'), covar=tensor([0.1300, 0.0987, 0.1026, 0.0680, 0.0724, 0.2402, 0.1181, 0.0816], device='cuda:2'), in_proj_covar=tensor([0.0617, 0.0759, 0.0881, 0.0774, 0.0582, 0.0613, 0.0631, 0.0729], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:52:54,098 INFO [optim.py:368] (2/8) 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 07:53:03,423 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 07:53:05,678 INFO [train.py:904] (2/8) Epoch 21, batch 7900, loss[loss=0.1844, simple_loss=0.2805, pruned_loss=0.0442, over 16900.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2909, pruned_loss=0.05999, over 3090108.40 frames. ], batch size: 96, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:53:28,602 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2781, 2.2662, 2.8580, 3.0941, 3.0864, 3.8802, 2.4787, 3.6535], device='cuda:2'), covar=tensor([0.0194, 0.0463, 0.0285, 0.0292, 0.0267, 0.0121, 0.0476, 0.0125], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0191, 0.0175, 0.0181, 0.0194, 0.0151, 0.0194, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:54:24,395 INFO [train.py:904] (2/8) Epoch 21, batch 7950, loss[loss=0.2519, simple_loss=0.3188, pruned_loss=0.09249, over 11493.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2918, pruned_loss=0.06117, over 3072572.65 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:54:30,111 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.49 vs. limit=2.0 2023-05-01 07:55:14,547 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:17,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7476, 4.0130, 4.2424, 4.1855, 4.2063, 3.9436, 3.6925, 3.9538], device='cuda:2'), covar=tensor([0.0582, 0.0798, 0.0510, 0.0663, 0.0632, 0.0662, 0.1367, 0.0643], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0444, 0.0432, 0.0403, 0.0479, 0.0454, 0.0540, 0.0362], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 07:55:20,872 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:21,033 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4461, 2.1563, 1.7519, 1.8419, 2.4267, 2.1020, 2.1715, 2.5359], device='cuda:2'), covar=tensor([0.0195, 0.0355, 0.0524, 0.0461, 0.0262, 0.0366, 0.0232, 0.0258], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0229, 0.0222, 0.0222, 0.0231, 0.0228, 0.0230, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:55:29,111 INFO [optim.py:368] (2/8) 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 07:55:41,323 INFO [train.py:904] (2/8) Epoch 21, batch 8000, loss[loss=0.2204, simple_loss=0.308, pruned_loss=0.06641, over 16674.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2923, pruned_loss=0.06148, over 3070156.55 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:56:12,636 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:15,857 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5816, 4.1690, 4.1690, 2.7110, 3.6818, 4.1517, 3.6468, 2.4089], device='cuda:2'), covar=tensor([0.0486, 0.0051, 0.0048, 0.0395, 0.0103, 0.0102, 0.0094, 0.0425], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0082, 0.0083, 0.0134, 0.0096, 0.0108, 0.0093, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 07:56:35,335 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:45,453 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:56,539 INFO [train.py:904] (2/8) Epoch 21, batch 8050, loss[loss=0.1844, simple_loss=0.2757, pruned_loss=0.04652, over 16576.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2919, pruned_loss=0.06106, over 3070956.41 frames. ], batch size: 62, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:56:58,694 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9055, 4.1813, 3.9847, 4.0340, 3.7337, 3.8401, 3.8330, 4.1548], device='cuda:2'), covar=tensor([0.1170, 0.0838, 0.1005, 0.0861, 0.0817, 0.1623, 0.0941, 0.1051], device='cuda:2'), in_proj_covar=tensor([0.0657, 0.0793, 0.0665, 0.0607, 0.0506, 0.0517, 0.0671, 0.0625], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 07:57:59,279 INFO [optim.py:368] (2/8) 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 07:58:10,501 INFO [train.py:904] (2/8) Epoch 21, batch 8100, loss[loss=0.1707, simple_loss=0.2687, pruned_loss=0.03636, over 16896.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2912, pruned_loss=0.06002, over 3086215.17 frames. ], batch size: 96, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:58:13,856 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 07:58:38,220 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211120.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:59:22,860 INFO [train.py:904] (2/8) Epoch 21, batch 8150, loss[loss=0.1606, simple_loss=0.2421, pruned_loss=0.03956, over 17245.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2882, pruned_loss=0.05821, over 3108992.41 frames. ], batch size: 45, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:00:27,468 INFO [optim.py:368] (2/8) 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 08:00:40,744 INFO [train.py:904] (2/8) Epoch 21, batch 8200, loss[loss=0.2235, simple_loss=0.3066, pruned_loss=0.07017, over 15415.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2857, pruned_loss=0.0579, over 3121376.47 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:01:59,027 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7226, 4.7188, 4.5223, 3.8060, 4.5793, 1.7192, 4.3531, 4.3625], device='cuda:2'), covar=tensor([0.0108, 0.0105, 0.0201, 0.0448, 0.0134, 0.2853, 0.0145, 0.0237], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0151, 0.0195, 0.0174, 0.0172, 0.0204, 0.0182, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:01:59,677 INFO [train.py:904] (2/8) Epoch 21, batch 8250, loss[loss=0.1759, simple_loss=0.2691, pruned_loss=0.04132, over 16844.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2846, pruned_loss=0.05567, over 3113410.29 frames. ], batch size: 96, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:02:09,429 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 08:02:56,787 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211288.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:03:06,691 INFO [optim.py:368] (2/8) 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 08:03:18,674 INFO [train.py:904] (2/8) Epoch 21, batch 8300, loss[loss=0.1794, simple_loss=0.2753, pruned_loss=0.04176, over 16778.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2823, pruned_loss=0.05286, over 3108519.89 frames. ], batch size: 124, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:03:32,491 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2320, 2.1160, 2.0990, 3.8286, 2.0311, 2.4751, 2.2199, 2.2981], device='cuda:2'), covar=tensor([0.1239, 0.4012, 0.3214, 0.0529, 0.4658, 0.2796, 0.3906, 0.3605], device='cuda:2'), in_proj_covar=tensor([0.0391, 0.0439, 0.0360, 0.0319, 0.0431, 0.0504, 0.0409, 0.0512], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:03:51,805 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211322.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:12,347 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211336.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:16,012 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:18,455 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:33,927 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7071, 2.5698, 2.4361, 3.7303, 2.2093, 3.8183, 1.4792, 2.8966], device='cuda:2'), covar=tensor([0.1430, 0.0767, 0.1192, 0.0211, 0.0121, 0.0380, 0.1790, 0.0731], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0189, 0.0206, 0.0215, 0.0201, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 08:04:36,926 INFO [train.py:904] (2/8) Epoch 21, batch 8350, loss[loss=0.1919, simple_loss=0.2894, pruned_loss=0.04721, over 16663.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2814, pruned_loss=0.05077, over 3094480.88 frames. ], batch size: 76, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:05:05,458 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:10,764 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3916, 2.3007, 2.2664, 4.1971, 2.1751, 2.6415, 2.3723, 2.4561], device='cuda:2'), covar=tensor([0.1191, 0.3563, 0.3081, 0.0439, 0.4470, 0.2684, 0.3829, 0.3322], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0436, 0.0357, 0.0316, 0.0428, 0.0499, 0.0406, 0.0509], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:05:30,340 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,435 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:43,486 INFO [optim.py:368] (2/8) 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 08:05:55,504 INFO [train.py:904] (2/8) Epoch 21, batch 8400, loss[loss=0.18, simple_loss=0.2766, pruned_loss=0.04173, over 16714.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2786, pruned_loss=0.04874, over 3071568.52 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:06:10,873 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:06:22,484 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:03,728 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:10,965 INFO [train.py:904] (2/8) Epoch 21, batch 8450, loss[loss=0.1919, simple_loss=0.2693, pruned_loss=0.05719, over 12177.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2771, pruned_loss=0.04721, over 3067743.65 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:07:36,102 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:44,872 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211473.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:08:18,590 INFO [optim.py:368] (2/8) 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 08:08:29,990 INFO [train.py:904] (2/8) Epoch 21, batch 8500, loss[loss=0.1673, simple_loss=0.2606, pruned_loss=0.03697, over 16719.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2737, pruned_loss=0.0451, over 3073233.90 frames. ], batch size: 83, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:09:54,385 INFO [train.py:904] (2/8) Epoch 21, batch 8550, loss[loss=0.1863, simple_loss=0.2829, pruned_loss=0.04486, over 15333.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.272, pruned_loss=0.04418, over 3073557.53 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:11:18,417 INFO [optim.py:368] (2/8) 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 08:11:32,287 INFO [train.py:904] (2/8) Epoch 21, batch 8600, loss[loss=0.1786, simple_loss=0.2767, pruned_loss=0.04032, over 15320.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2719, pruned_loss=0.04316, over 3064528.83 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:12:48,779 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:13:09,686 INFO [train.py:904] (2/8) Epoch 21, batch 8650, loss[loss=0.1718, simple_loss=0.2873, pruned_loss=0.02814, over 16671.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.27, pruned_loss=0.04157, over 3047385.75 frames. ], batch size: 89, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:14:30,914 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:14:43,483 INFO [optim.py:368] (2/8) 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 08:14:51,333 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2448, 4.2317, 4.0727, 3.3907, 4.1531, 1.8843, 3.9131, 3.8357], device='cuda:2'), covar=tensor([0.0106, 0.0113, 0.0179, 0.0283, 0.0110, 0.2574, 0.0141, 0.0255], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0147, 0.0189, 0.0169, 0.0167, 0.0199, 0.0177, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:14:57,320 INFO [train.py:904] (2/8) Epoch 21, batch 8700, loss[loss=0.1854, simple_loss=0.2833, pruned_loss=0.04379, over 16658.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2678, pruned_loss=0.04066, over 3056136.15 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:16:09,882 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:16:31,418 INFO [train.py:904] (2/8) Epoch 21, batch 8750, loss[loss=0.1482, simple_loss=0.2417, pruned_loss=0.02738, over 12467.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2677, pruned_loss=0.04004, over 3067292.68 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:17:13,746 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:17:50,975 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1417, 2.1234, 2.0968, 3.8249, 2.1171, 2.4365, 2.2829, 2.2527], device='cuda:2'), covar=tensor([0.1314, 0.3868, 0.3251, 0.0526, 0.4470, 0.2805, 0.3704, 0.3603], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0435, 0.0358, 0.0316, 0.0427, 0.0498, 0.0405, 0.0507], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:18:09,822 INFO [optim.py:368] (2/8) 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 08:18:23,762 INFO [train.py:904] (2/8) Epoch 21, batch 8800, loss[loss=0.1756, simple_loss=0.2702, pruned_loss=0.04053, over 16639.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2663, pruned_loss=0.03906, over 3069341.77 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:18:48,204 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6723, 3.9845, 3.0588, 2.2253, 2.4750, 2.5521, 4.2141, 3.3467], device='cuda:2'), covar=tensor([0.2832, 0.0586, 0.1709, 0.2987, 0.2804, 0.2038, 0.0390, 0.1305], device='cuda:2'), in_proj_covar=tensor([0.0320, 0.0263, 0.0300, 0.0305, 0.0289, 0.0253, 0.0288, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 08:20:07,833 INFO [train.py:904] (2/8) Epoch 21, batch 8850, loss[loss=0.1755, simple_loss=0.2774, pruned_loss=0.03679, over 16366.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2679, pruned_loss=0.03844, over 3058963.03 frames. ], batch size: 146, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:21:38,908 INFO [optim.py:368] (2/8) 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 08:21:54,319 INFO [train.py:904] (2/8) Epoch 21, batch 8900, loss[loss=0.1761, simple_loss=0.2684, pruned_loss=0.04186, over 15306.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2679, pruned_loss=0.03767, over 3055265.64 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:23:57,812 INFO [train.py:904] (2/8) Epoch 21, batch 8950, loss[loss=0.1472, simple_loss=0.243, pruned_loss=0.02568, over 16438.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2667, pruned_loss=0.0375, over 3052812.40 frames. ], batch size: 68, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:29,295 INFO [optim.py:368] (2/8) 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 08:25:46,899 INFO [train.py:904] (2/8) Epoch 21, batch 9000, loss[loss=0.1779, simple_loss=0.2561, pruned_loss=0.04986, over 12399.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2636, pruned_loss=0.03657, over 3046244.69 frames. ], batch size: 250, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:46,900 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 08:25:57,426 INFO [train.py:938] (2/8) Epoch 21, validation: loss=0.1457, simple_loss=0.2498, pruned_loss=0.02077, over 944034.00 frames. 2023-05-01 08:25:57,427 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 08:27:19,761 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212042.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:27:39,214 INFO [train.py:904] (2/8) Epoch 21, batch 9050, loss[loss=0.151, simple_loss=0.2423, pruned_loss=0.02991, over 16527.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2647, pruned_loss=0.03703, over 3068879.00 frames. ], batch size: 62, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:28:14,809 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212090.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:29:10,423 INFO [optim.py:368] (2/8) 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 08:29:26,247 INFO [train.py:904] (2/8) Epoch 21, batch 9100, loss[loss=0.2036, simple_loss=0.2978, pruned_loss=0.05468, over 16299.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2643, pruned_loss=0.03754, over 3059170.16 frames. ], batch size: 146, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:29:30,847 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212104.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:29:53,443 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:31:22,739 INFO [train.py:904] (2/8) Epoch 21, batch 9150, loss[loss=0.1591, simple_loss=0.2617, pruned_loss=0.02823, over 16475.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2641, pruned_loss=0.03683, over 3045503.86 frames. ], batch size: 68, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:31:23,327 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4911, 5.8442, 5.5412, 5.6409, 5.2937, 5.3483, 5.1469, 5.9261], device='cuda:2'), covar=tensor([0.1340, 0.0805, 0.0929, 0.0722, 0.0835, 0.0611, 0.1269, 0.0847], device='cuda:2'), in_proj_covar=tensor([0.0645, 0.0783, 0.0650, 0.0595, 0.0498, 0.0508, 0.0657, 0.0613], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:31:52,346 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212165.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:32:57,277 INFO [optim.py:368] (2/8) 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 08:33:00,909 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7493, 2.6493, 2.4190, 4.5265, 2.4619, 2.9486, 2.6708, 2.7322], device='cuda:2'), covar=tensor([0.1049, 0.3333, 0.3045, 0.0410, 0.4073, 0.2566, 0.3100, 0.3671], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0436, 0.0359, 0.0316, 0.0428, 0.0498, 0.0407, 0.0508], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:33:09,937 INFO [train.py:904] (2/8) Epoch 21, batch 9200, loss[loss=0.162, simple_loss=0.2437, pruned_loss=0.04016, over 11988.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2602, pruned_loss=0.036, over 3045201.51 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:48,818 INFO [train.py:904] (2/8) Epoch 21, batch 9250, loss[loss=0.153, simple_loss=0.2412, pruned_loss=0.03236, over 12246.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2599, pruned_loss=0.0362, over 3023616.12 frames. ], batch size: 250, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:51,586 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:34:55,757 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5952, 3.6013, 2.1959, 4.0591, 2.7276, 3.9701, 2.4166, 2.9696], device='cuda:2'), covar=tensor([0.0256, 0.0337, 0.1593, 0.0177, 0.0793, 0.0503, 0.1477, 0.0748], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0169, 0.0188, 0.0153, 0.0171, 0.0207, 0.0196, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 08:35:51,853 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 08:36:25,635 INFO [optim.py:368] (2/8) 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 08:36:39,743 INFO [train.py:904] (2/8) Epoch 21, batch 9300, loss[loss=0.1617, simple_loss=0.2541, pruned_loss=0.03465, over 15244.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2585, pruned_loss=0.03592, over 3011832.29 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:37:07,754 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212314.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:37:34,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6159, 4.0282, 4.3251, 1.8111, 4.5466, 4.7200, 3.3551, 3.2675], device='cuda:2'), covar=tensor([0.0971, 0.0180, 0.0204, 0.1386, 0.0067, 0.0097, 0.0378, 0.0498], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0103, 0.0091, 0.0132, 0.0075, 0.0117, 0.0123, 0.0123], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 08:38:00,333 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-01 08:38:22,843 INFO [train.py:904] (2/8) Epoch 21, batch 9350, loss[loss=0.1963, simple_loss=0.2923, pruned_loss=0.0502, over 16966.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2582, pruned_loss=0.03537, over 3044838.39 frames. ], batch size: 109, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:39:47,852 INFO [optim.py:368] (2/8) 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 08:40:02,994 INFO [train.py:904] (2/8) Epoch 21, batch 9400, loss[loss=0.1792, simple_loss=0.2805, pruned_loss=0.03895, over 16368.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.258, pruned_loss=0.03547, over 3018914.59 frames. ], batch size: 146, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:18,510 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:41:42,917 INFO [train.py:904] (2/8) Epoch 21, batch 9450, loss[loss=0.1642, simple_loss=0.2533, pruned_loss=0.03756, over 13054.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2596, pruned_loss=0.03547, over 3014786.18 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:58,671 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212460.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 08:43:11,245 INFO [optim.py:368] (2/8) 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 08:43:13,486 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7184, 2.9579, 2.7169, 4.6736, 3.3365, 4.2773, 1.7383, 3.1656], device='cuda:2'), covar=tensor([0.1338, 0.0707, 0.1027, 0.0195, 0.0159, 0.0355, 0.1504, 0.0680], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0169, 0.0190, 0.0181, 0.0197, 0.0209, 0.0198, 0.0189], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 08:43:21,239 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:43:25,063 INFO [train.py:904] (2/8) Epoch 21, batch 9500, loss[loss=0.1589, simple_loss=0.258, pruned_loss=0.02985, over 16434.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2593, pruned_loss=0.03515, over 3042052.51 frames. ], batch size: 146, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:45:10,798 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8132, 4.8912, 4.7083, 4.3244, 4.4257, 4.8087, 4.6783, 4.4719], device='cuda:2'), covar=tensor([0.0701, 0.0668, 0.0337, 0.0348, 0.0870, 0.0618, 0.0406, 0.0788], device='cuda:2'), in_proj_covar=tensor([0.0273, 0.0395, 0.0321, 0.0316, 0.0327, 0.0367, 0.0222, 0.0383], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-05-01 08:45:12,071 INFO [train.py:904] (2/8) Epoch 21, batch 9550, loss[loss=0.1548, simple_loss=0.2512, pruned_loss=0.0292, over 16653.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2592, pruned_loss=0.03508, over 3057266.70 frames. ], batch size: 83, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:46:37,322 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 08:46:39,906 INFO [optim.py:368] (2/8) 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 08:46:50,960 INFO [train.py:904] (2/8) Epoch 21, batch 9600, loss[loss=0.1897, simple_loss=0.2868, pruned_loss=0.0463, over 16223.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2603, pruned_loss=0.03575, over 3040652.61 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:47:06,242 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212609.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:48:38,610 INFO [train.py:904] (2/8) Epoch 21, batch 9650, loss[loss=0.1528, simple_loss=0.2549, pruned_loss=0.0254, over 16378.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2627, pruned_loss=0.03617, over 3058833.70 frames. ], batch size: 146, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:11,919 INFO [optim.py:368] (2/8) 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 08:50:20,789 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212698.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:50:28,029 INFO [train.py:904] (2/8) Epoch 21, batch 9700, loss[loss=0.1661, simple_loss=0.2574, pruned_loss=0.03745, over 12563.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2621, pruned_loss=0.03597, over 3070450.33 frames. ], batch size: 250, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:46,044 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:51:31,325 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6992, 3.7579, 2.2241, 4.2116, 2.8402, 4.1278, 2.5036, 3.0463], device='cuda:2'), covar=tensor([0.0254, 0.0316, 0.1721, 0.0209, 0.0869, 0.0503, 0.1537, 0.0753], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0168, 0.0188, 0.0152, 0.0171, 0.0206, 0.0195, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 08:52:10,270 INFO [train.py:904] (2/8) Epoch 21, batch 9750, loss[loss=0.1733, simple_loss=0.2573, pruned_loss=0.04464, over 12337.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2609, pruned_loss=0.03618, over 3065461.77 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:52:24,883 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:26,961 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212760.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:52:29,531 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5864, 4.5327, 4.9269, 4.8963, 4.9159, 4.6431, 4.6094, 4.5042], device='cuda:2'), covar=tensor([0.0305, 0.0667, 0.0384, 0.0395, 0.0413, 0.0363, 0.0825, 0.0420], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0430, 0.0417, 0.0388, 0.0464, 0.0437, 0.0516, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 08:52:49,223 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:53,750 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212774.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:55,413 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2249, 5.5260, 5.2844, 5.2963, 5.0376, 4.9803, 4.9300, 5.6004], device='cuda:2'), covar=tensor([0.1325, 0.0856, 0.1019, 0.0896, 0.0897, 0.0797, 0.1232, 0.0860], device='cuda:2'), in_proj_covar=tensor([0.0637, 0.0774, 0.0642, 0.0587, 0.0493, 0.0503, 0.0647, 0.0607], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:52:57,035 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0878, 4.0713, 4.4453, 4.4244, 4.4236, 4.1493, 4.1533, 4.1732], device='cuda:2'), covar=tensor([0.0371, 0.0844, 0.0390, 0.0409, 0.0507, 0.0430, 0.0929, 0.0461], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0430, 0.0417, 0.0388, 0.0464, 0.0437, 0.0515, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 08:53:36,937 INFO [optim.py:368] (2/8) 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 08:53:37,559 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:44,466 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6883, 2.5926, 1.8699, 2.8320, 2.0751, 2.7999, 2.1506, 2.4002], device='cuda:2'), covar=tensor([0.0322, 0.0329, 0.1296, 0.0272, 0.0704, 0.0502, 0.1327, 0.0616], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0168, 0.0188, 0.0152, 0.0171, 0.0206, 0.0195, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 08:53:48,982 INFO [train.py:904] (2/8) Epoch 21, batch 9800, loss[loss=0.1723, simple_loss=0.272, pruned_loss=0.03636, over 16312.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2608, pruned_loss=0.03534, over 3064760.75 frames. ], batch size: 146, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:53:59,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7036, 1.8135, 2.2327, 2.7186, 2.6043, 3.1265, 2.0706, 3.0610], device='cuda:2'), covar=tensor([0.0267, 0.0585, 0.0421, 0.0327, 0.0348, 0.0169, 0.0538, 0.0170], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0189, 0.0173, 0.0177, 0.0190, 0.0147, 0.0190, 0.0143], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:54:01,428 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212808.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:54:46,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3150, 4.2018, 4.3732, 4.4872, 4.6231, 4.1612, 4.6434, 4.6448], device='cuda:2'), covar=tensor([0.1703, 0.1175, 0.1455, 0.0719, 0.0590, 0.1143, 0.0552, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0596, 0.0731, 0.0855, 0.0748, 0.0568, 0.0594, 0.0612, 0.0709], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 08:54:51,506 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212835.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:55:31,551 INFO [train.py:904] (2/8) Epoch 21, batch 9850, loss[loss=0.1661, simple_loss=0.2656, pruned_loss=0.03328, over 16628.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2629, pruned_loss=0.03551, over 3067784.46 frames. ], batch size: 62, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:57:07,912 INFO [optim.py:368] (2/8) 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 08:57:21,370 INFO [train.py:904] (2/8) Epoch 21, batch 9900, loss[loss=0.1791, simple_loss=0.2802, pruned_loss=0.03905, over 16169.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2629, pruned_loss=0.0357, over 3049335.40 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:57:28,789 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-01 08:57:38,613 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:58:40,890 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1985, 3.2249, 1.9364, 3.4754, 2.3928, 3.4635, 2.1025, 2.6526], device='cuda:2'), covar=tensor([0.0321, 0.0388, 0.1671, 0.0290, 0.0866, 0.0586, 0.1598, 0.0768], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0170, 0.0189, 0.0153, 0.0172, 0.0207, 0.0197, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 08:59:17,572 INFO [train.py:904] (2/8) Epoch 21, batch 9950, loss[loss=0.1601, simple_loss=0.2668, pruned_loss=0.02673, over 16680.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2647, pruned_loss=0.03558, over 3068805.93 frames. ], batch size: 134, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:59:29,578 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:00:58,645 INFO [optim.py:368] (2/8) 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 09:01:17,070 INFO [train.py:904] (2/8) Epoch 21, batch 10000, loss[loss=0.1597, simple_loss=0.2595, pruned_loss=0.02998, over 16772.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.263, pruned_loss=0.03525, over 3077413.68 frames. ], batch size: 76, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:01:48,180 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 09:02:00,746 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6676, 3.7726, 1.9408, 4.1307, 2.8065, 4.0812, 2.1543, 2.9509], device='cuda:2'), covar=tensor([0.0240, 0.0285, 0.1929, 0.0233, 0.0793, 0.0461, 0.1835, 0.0713], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0169, 0.0188, 0.0152, 0.0171, 0.0206, 0.0196, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 09:02:36,509 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6060, 3.6747, 2.7351, 2.1935, 2.2733, 2.3886, 3.8813, 3.1550], device='cuda:2'), covar=tensor([0.2890, 0.0553, 0.1840, 0.2939, 0.2989, 0.2140, 0.0423, 0.1312], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0258, 0.0295, 0.0300, 0.0281, 0.0249, 0.0282, 0.0320], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:02:58,019 INFO [train.py:904] (2/8) Epoch 21, batch 10050, loss[loss=0.1723, simple_loss=0.273, pruned_loss=0.03578, over 16911.00 frames. ], tot_loss[loss=0.167, simple_loss=0.263, pruned_loss=0.03555, over 3049551.70 frames. ], batch size: 116, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:03:02,065 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213054.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:03:18,748 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6040, 4.8013, 4.9391, 4.7552, 4.7175, 5.2904, 4.8183, 4.5201], device='cuda:2'), covar=tensor([0.1172, 0.1744, 0.1867, 0.1833, 0.2483, 0.0871, 0.1517, 0.2362], device='cuda:2'), in_proj_covar=tensor([0.0381, 0.0554, 0.0613, 0.0463, 0.0611, 0.0646, 0.0484, 0.0619], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:03:28,731 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:04:21,049 INFO [optim.py:368] (2/8) 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 09:04:21,648 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:04:32,156 INFO [train.py:904] (2/8) Epoch 21, batch 10100, loss[loss=0.1569, simple_loss=0.2438, pruned_loss=0.035, over 12718.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2638, pruned_loss=0.03574, over 3065193.43 frames. ], batch size: 248, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:05:27,296 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213130.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:30,050 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9200, 4.8998, 4.6438, 4.3175, 4.7268, 1.7813, 4.5416, 4.5562], device='cuda:2'), covar=tensor([0.0076, 0.0073, 0.0181, 0.0251, 0.0099, 0.2598, 0.0120, 0.0202], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0145, 0.0185, 0.0163, 0.0164, 0.0198, 0.0174, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:05:42,670 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:51,220 INFO [train.py:904] (2/8) Epoch 21, batch 10150, loss[loss=0.1605, simple_loss=0.2477, pruned_loss=0.03662, over 12425.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2622, pruned_loss=0.03554, over 3038796.54 frames. ], batch size: 250, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:06:16,171 INFO [train.py:904] (2/8) Epoch 22, batch 0, loss[loss=0.2033, simple_loss=0.2949, pruned_loss=0.05579, over 17134.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2949, pruned_loss=0.05579, over 17134.00 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:06:16,171 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 09:06:23,629 INFO [train.py:938] (2/8) Epoch 22, validation: loss=0.1457, simple_loss=0.2489, pruned_loss=0.0212, over 944034.00 frames. 2023-05-01 09:06:23,630 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 09:06:32,005 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-01 09:07:26,339 INFO [optim.py:368] (2/8) 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 09:07:34,021 INFO [train.py:904] (2/8) Epoch 22, batch 50, loss[loss=0.1771, simple_loss=0.2455, pruned_loss=0.05435, over 16913.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2697, pruned_loss=0.04895, over 751533.76 frames. ], batch size: 90, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:07:42,198 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5405, 4.5144, 4.4707, 4.1048, 4.4668, 1.8143, 4.2721, 4.1928], device='cuda:2'), covar=tensor([0.0128, 0.0103, 0.0211, 0.0261, 0.0132, 0.2507, 0.0162, 0.0239], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0146, 0.0186, 0.0163, 0.0165, 0.0198, 0.0174, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:08:02,956 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 09:08:26,375 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4337, 5.8046, 5.5432, 5.5959, 5.2017, 5.2019, 5.1908, 5.9225], device='cuda:2'), covar=tensor([0.1370, 0.0968, 0.1199, 0.0867, 0.0879, 0.0737, 0.1220, 0.0953], device='cuda:2'), in_proj_covar=tensor([0.0638, 0.0777, 0.0643, 0.0588, 0.0494, 0.0504, 0.0650, 0.0608], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:08:26,477 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6864, 3.7100, 4.0646, 1.8297, 4.2901, 4.4457, 3.1631, 3.3232], device='cuda:2'), covar=tensor([0.0891, 0.0298, 0.0256, 0.1415, 0.0090, 0.0191, 0.0497, 0.0458], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0104, 0.0092, 0.0135, 0.0077, 0.0119, 0.0125, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 09:08:41,847 INFO [train.py:904] (2/8) Epoch 22, batch 100, loss[loss=0.1893, simple_loss=0.2821, pruned_loss=0.04829, over 17123.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2659, pruned_loss=0.04609, over 1327952.54 frames. ], batch size: 48, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:09:44,707 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 22, batch 150, loss[loss=0.1477, simple_loss=0.2309, pruned_loss=0.03224, over 16751.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2654, pruned_loss=0.04632, over 1760809.03 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:10:01,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5787, 4.4848, 4.8543, 4.8690, 4.9226, 4.5687, 4.5537, 4.4387], device='cuda:2'), covar=tensor([0.0317, 0.0715, 0.0475, 0.0459, 0.0499, 0.0465, 0.0977, 0.0661], device='cuda:2'), in_proj_covar=tensor([0.0386, 0.0431, 0.0418, 0.0392, 0.0465, 0.0440, 0.0517, 0.0353], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 09:10:24,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0732, 4.6842, 3.2689, 2.4718, 2.8342, 2.6276, 4.9571, 3.7221], device='cuda:2'), covar=tensor([0.2578, 0.0492, 0.1666, 0.2764, 0.2873, 0.2040, 0.0316, 0.1409], device='cuda:2'), in_proj_covar=tensor([0.0320, 0.0262, 0.0299, 0.0304, 0.0285, 0.0253, 0.0286, 0.0325], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:10:39,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4421, 5.3734, 5.2809, 4.7344, 4.9088, 5.2924, 5.2690, 4.8369], device='cuda:2'), covar=tensor([0.0610, 0.0518, 0.0338, 0.0391, 0.1274, 0.0488, 0.0270, 0.0929], device='cuda:2'), in_proj_covar=tensor([0.0280, 0.0406, 0.0327, 0.0322, 0.0334, 0.0375, 0.0226, 0.0391], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:10:44,843 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 09:11:00,679 INFO [train.py:904] (2/8) Epoch 22, batch 200, loss[loss=0.1802, simple_loss=0.2581, pruned_loss=0.05113, over 16770.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2646, pruned_loss=0.04622, over 2102048.57 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:02,277 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:11:20,361 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213367.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:11:33,312 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2535, 4.6534, 3.4122, 2.6687, 3.0806, 2.8314, 4.9259, 3.7662], device='cuda:2'), covar=tensor([0.2573, 0.0544, 0.1744, 0.2836, 0.2597, 0.1981, 0.0388, 0.1331], device='cuda:2'), in_proj_covar=tensor([0.0321, 0.0263, 0.0300, 0.0305, 0.0286, 0.0253, 0.0287, 0.0326], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:12:00,918 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213402.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:08,112 INFO [train.py:904] (2/8) Epoch 22, batch 250, loss[loss=0.1666, simple_loss=0.2565, pruned_loss=0.03837, over 17058.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.263, pruned_loss=0.04472, over 2383874.55 frames. ], batch size: 53, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:12:25,570 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213415.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:40,463 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4100, 2.4585, 1.9455, 2.1453, 2.7860, 2.5167, 3.1396, 3.0507], device='cuda:2'), covar=tensor([0.0258, 0.0567, 0.0789, 0.0684, 0.0385, 0.0532, 0.0334, 0.0356], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0234, 0.0226, 0.0226, 0.0235, 0.0233, 0.0232, 0.0228], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:12:46,319 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:13:00,872 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-01 09:13:17,488 INFO [train.py:904] (2/8) Epoch 22, batch 300, loss[loss=0.194, simple_loss=0.288, pruned_loss=0.04996, over 17050.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2595, pruned_loss=0.0438, over 2596921.84 frames. ], batch size: 55, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:13:41,806 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-01 09:13:52,064 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:14:19,526 INFO [optim.py:368] (2/8) 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 09:14:25,241 INFO [train.py:904] (2/8) Epoch 22, batch 350, loss[loss=0.182, simple_loss=0.2682, pruned_loss=0.04787, over 17016.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2576, pruned_loss=0.04278, over 2758511.27 frames. ], batch size: 55, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:15:31,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8097, 3.0007, 3.1866, 1.9652, 2.7118, 2.1799, 3.2915, 3.3079], device='cuda:2'), covar=tensor([0.0266, 0.0876, 0.0613, 0.2018, 0.0926, 0.1029, 0.0612, 0.0902], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0159, 0.0165, 0.0152, 0.0143, 0.0129, 0.0141, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 09:15:34,171 INFO [train.py:904] (2/8) Epoch 22, batch 400, loss[loss=0.1603, simple_loss=0.2414, pruned_loss=0.03962, over 16398.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2564, pruned_loss=0.04259, over 2893846.02 frames. ], batch size: 36, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:15:52,107 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:16:36,629 INFO [optim.py:368] (2/8) 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 09:16:43,116 INFO [train.py:904] (2/8) Epoch 22, batch 450, loss[loss=0.1533, simple_loss=0.2549, pruned_loss=0.02584, over 17068.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2543, pruned_loss=0.04146, over 2988200.01 frames. ], batch size: 50, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:17:16,801 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213627.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:17:47,072 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0430, 2.2593, 2.3475, 2.6538, 2.0514, 3.1084, 1.8368, 2.7406], device='cuda:2'), covar=tensor([0.1091, 0.0708, 0.1067, 0.0186, 0.0110, 0.0352, 0.1356, 0.0709], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0186, 0.0201, 0.0213, 0.0202, 0.0192], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 09:17:48,778 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 09:17:53,028 INFO [train.py:904] (2/8) Epoch 22, batch 500, loss[loss=0.1726, simple_loss=0.2451, pruned_loss=0.05001, over 16773.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2538, pruned_loss=0.041, over 3060654.98 frames. ], batch size: 124, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:18:11,682 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1160, 2.1275, 2.6980, 3.0345, 2.8337, 3.4791, 2.1681, 3.5306], device='cuda:2'), covar=tensor([0.0220, 0.0559, 0.0316, 0.0290, 0.0330, 0.0175, 0.0673, 0.0149], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0192, 0.0178, 0.0182, 0.0196, 0.0151, 0.0195, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:18:54,887 INFO [optim.py:368] (2/8) 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 09:19:01,576 INFO [train.py:904] (2/8) Epoch 22, batch 550, loss[loss=0.158, simple_loss=0.2392, pruned_loss=0.03846, over 16862.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2524, pruned_loss=0.04057, over 3119930.52 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:19:47,646 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8623, 4.5594, 3.2435, 2.3426, 2.7253, 2.6991, 4.8672, 3.6688], device='cuda:2'), covar=tensor([0.2823, 0.0481, 0.1638, 0.2848, 0.2936, 0.2076, 0.0288, 0.1365], device='cuda:2'), in_proj_covar=tensor([0.0325, 0.0267, 0.0304, 0.0311, 0.0293, 0.0258, 0.0292, 0.0333], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:20:10,673 INFO [train.py:904] (2/8) Epoch 22, batch 600, loss[loss=0.1684, simple_loss=0.2377, pruned_loss=0.04952, over 16689.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2522, pruned_loss=0.04082, over 3165926.70 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:20:20,314 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8390, 3.6932, 4.2577, 2.0612, 4.4319, 4.5603, 3.1717, 3.4140], device='cuda:2'), covar=tensor([0.0779, 0.0295, 0.0232, 0.1191, 0.0101, 0.0196, 0.0504, 0.0409], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0096, 0.0140, 0.0080, 0.0125, 0.0129, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 09:20:40,624 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1994, 5.1637, 5.0334, 4.4773, 4.6740, 5.0647, 5.0157, 4.6653], device='cuda:2'), covar=tensor([0.0582, 0.0471, 0.0355, 0.0400, 0.1110, 0.0488, 0.0371, 0.0895], device='cuda:2'), in_proj_covar=tensor([0.0288, 0.0417, 0.0336, 0.0333, 0.0343, 0.0387, 0.0231, 0.0404], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:21:13,515 INFO [optim.py:368] (2/8) 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 09:21:21,578 INFO [train.py:904] (2/8) Epoch 22, batch 650, loss[loss=0.1514, simple_loss=0.245, pruned_loss=0.0289, over 17164.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2508, pruned_loss=0.04076, over 3187306.24 frames. ], batch size: 46, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:22:30,249 INFO [train.py:904] (2/8) Epoch 22, batch 700, loss[loss=0.1717, simple_loss=0.259, pruned_loss=0.04218, over 16730.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2504, pruned_loss=0.04005, over 3217508.46 frames. ], batch size: 57, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:22:41,504 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7850, 3.7925, 2.3362, 4.0300, 3.0078, 4.0285, 2.4071, 3.0472], device='cuda:2'), covar=tensor([0.0257, 0.0383, 0.1551, 0.0408, 0.0769, 0.0707, 0.1497, 0.0725], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0176, 0.0195, 0.0162, 0.0178, 0.0216, 0.0203, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 09:23:35,483 INFO [optim.py:368] (2/8) 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 09:23:41,789 INFO [train.py:904] (2/8) Epoch 22, batch 750, loss[loss=0.1555, simple_loss=0.2387, pruned_loss=0.03614, over 16503.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2507, pruned_loss=0.04032, over 3238450.22 frames. ], batch size: 68, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:24:07,130 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9527, 4.2353, 4.0682, 4.1367, 3.8402, 3.8763, 3.8390, 4.2331], device='cuda:2'), covar=tensor([0.1287, 0.1066, 0.1036, 0.0854, 0.0822, 0.1641, 0.1034, 0.1173], device='cuda:2'), in_proj_covar=tensor([0.0679, 0.0827, 0.0681, 0.0629, 0.0528, 0.0535, 0.0693, 0.0649], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:24:08,343 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:24:35,535 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 09:24:53,400 INFO [train.py:904] (2/8) Epoch 22, batch 800, loss[loss=0.1929, simple_loss=0.2656, pruned_loss=0.06007, over 12106.00 frames. ], tot_loss[loss=0.165, simple_loss=0.25, pruned_loss=0.03994, over 3257012.52 frames. ], batch size: 248, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:25:26,420 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7046, 4.0754, 4.2187, 2.9407, 3.5575, 4.1338, 3.8312, 2.5502], device='cuda:2'), covar=tensor([0.0490, 0.0113, 0.0057, 0.0371, 0.0127, 0.0107, 0.0094, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0083, 0.0083, 0.0135, 0.0098, 0.0107, 0.0094, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:25:56,894 INFO [optim.py:368] (2/8) 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 09:26:06,616 INFO [train.py:904] (2/8) Epoch 22, batch 850, loss[loss=0.1802, simple_loss=0.2571, pruned_loss=0.05163, over 16869.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2494, pruned_loss=0.03964, over 3269571.65 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:27:17,240 INFO [train.py:904] (2/8) Epoch 22, batch 900, loss[loss=0.1704, simple_loss=0.2567, pruned_loss=0.04207, over 17169.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2493, pruned_loss=0.03951, over 3273622.88 frames. ], batch size: 46, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:28:19,798 INFO [optim.py:368] (2/8) 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 09:28:24,250 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7005, 6.1262, 5.8543, 5.9742, 5.4826, 5.5749, 5.6142, 6.2538], device='cuda:2'), covar=tensor([0.1519, 0.1002, 0.1158, 0.0802, 0.0958, 0.0626, 0.1130, 0.0943], device='cuda:2'), in_proj_covar=tensor([0.0676, 0.0821, 0.0679, 0.0626, 0.0525, 0.0532, 0.0690, 0.0646], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:28:27,527 INFO [train.py:904] (2/8) Epoch 22, batch 950, loss[loss=0.1461, simple_loss=0.233, pruned_loss=0.02957, over 16718.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2489, pruned_loss=0.0396, over 3288054.78 frames. ], batch size: 83, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:29:03,349 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214129.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:29:35,587 INFO [train.py:904] (2/8) Epoch 22, batch 1000, loss[loss=0.1509, simple_loss=0.2459, pruned_loss=0.02788, over 17150.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2481, pruned_loss=0.03932, over 3297418.82 frames. ], batch size: 48, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:30:15,709 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1374, 2.0458, 1.7284, 1.8311, 2.2581, 2.0421, 2.1094, 2.3864], device='cuda:2'), covar=tensor([0.0293, 0.0440, 0.0567, 0.0497, 0.0262, 0.0348, 0.0225, 0.0249], device='cuda:2'), in_proj_covar=tensor([0.0212, 0.0238, 0.0228, 0.0229, 0.0239, 0.0237, 0.0238, 0.0233], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:30:20,601 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2329, 5.2332, 4.9563, 4.4772, 5.0582, 1.9432, 4.7508, 4.7479], device='cuda:2'), covar=tensor([0.0080, 0.0069, 0.0198, 0.0369, 0.0105, 0.2798, 0.0144, 0.0235], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0155, 0.0197, 0.0174, 0.0175, 0.0208, 0.0186, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:30:29,067 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214190.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:30:39,709 INFO [optim.py:368] (2/8) 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 09:30:46,634 INFO [train.py:904] (2/8) Epoch 22, batch 1050, loss[loss=0.1649, simple_loss=0.2383, pruned_loss=0.04574, over 16843.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2485, pruned_loss=0.03892, over 3297018.25 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:31:07,762 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:13,739 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:56,093 INFO [train.py:904] (2/8) Epoch 22, batch 1100, loss[loss=0.1559, simple_loss=0.2463, pruned_loss=0.03277, over 17132.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2478, pruned_loss=0.03822, over 3309575.53 frames. ], batch size: 48, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:32:19,670 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214270.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:31,195 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:37,437 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2815, 4.6496, 4.1898, 4.5213, 4.2457, 4.1993, 4.1996, 4.7185], device='cuda:2'), covar=tensor([0.2365, 0.1655, 0.2400, 0.1474, 0.1596, 0.2361, 0.2381, 0.1966], device='cuda:2'), in_proj_covar=tensor([0.0678, 0.0829, 0.0683, 0.0630, 0.0527, 0.0535, 0.0694, 0.0651], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:32:57,752 INFO [optim.py:368] (2/8) 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 09:33:03,837 INFO [train.py:904] (2/8) Epoch 22, batch 1150, loss[loss=0.1734, simple_loss=0.2545, pruned_loss=0.04617, over 16968.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2475, pruned_loss=0.0379, over 3323147.53 frames. ], batch size: 41, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:34:15,635 INFO [train.py:904] (2/8) Epoch 22, batch 1200, loss[loss=0.1785, simple_loss=0.2784, pruned_loss=0.03936, over 17127.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2469, pruned_loss=0.03797, over 3323069.63 frames. ], batch size: 48, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:34:40,558 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5303, 3.6300, 3.8709, 2.7449, 3.4891, 3.8880, 3.6294, 2.2236], device='cuda:2'), covar=tensor([0.0489, 0.0249, 0.0054, 0.0364, 0.0122, 0.0090, 0.0099, 0.0513], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0083, 0.0084, 0.0134, 0.0098, 0.0108, 0.0094, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:35:18,108 INFO [optim.py:368] (2/8) 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 09:35:19,982 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 09:35:25,095 INFO [train.py:904] (2/8) Epoch 22, batch 1250, loss[loss=0.1725, simple_loss=0.2486, pruned_loss=0.04823, over 15492.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2469, pruned_loss=0.03889, over 3318596.57 frames. ], batch size: 190, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:35:38,817 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-01 09:36:35,043 INFO [train.py:904] (2/8) Epoch 22, batch 1300, loss[loss=0.1448, simple_loss=0.2402, pruned_loss=0.02469, over 17126.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2475, pruned_loss=0.03898, over 3325822.23 frames. ], batch size: 48, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:37:18,496 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 22, batch 1350, loss[loss=0.1492, simple_loss=0.234, pruned_loss=0.03223, over 15417.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2481, pruned_loss=0.03922, over 3326425.04 frames. ], batch size: 190, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:37:55,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8330, 4.0178, 2.9782, 2.3868, 2.6798, 2.5811, 4.2466, 3.4901], device='cuda:2'), covar=tensor([0.2630, 0.0609, 0.1787, 0.2769, 0.2626, 0.2068, 0.0457, 0.1480], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0271, 0.0307, 0.0313, 0.0297, 0.0261, 0.0295, 0.0338], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:38:48,216 INFO [train.py:904] (2/8) Epoch 22, batch 1400, loss[loss=0.1608, simple_loss=0.2449, pruned_loss=0.03841, over 16810.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2479, pruned_loss=0.03901, over 3319073.03 frames. ], batch size: 96, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:39:17,566 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:39:50,086 INFO [optim.py:368] (2/8) 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 09:39:57,056 INFO [train.py:904] (2/8) Epoch 22, batch 1450, loss[loss=0.1591, simple_loss=0.2354, pruned_loss=0.04139, over 12111.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2469, pruned_loss=0.03866, over 3311906.23 frames. ], batch size: 246, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:41:07,195 INFO [train.py:904] (2/8) Epoch 22, batch 1500, loss[loss=0.1651, simple_loss=0.2569, pruned_loss=0.0367, over 17080.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2464, pruned_loss=0.03863, over 3309395.16 frames. ], batch size: 53, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:42:10,190 INFO [optim.py:368] (2/8) 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 09:42:16,354 INFO [train.py:904] (2/8) Epoch 22, batch 1550, loss[loss=0.1755, simple_loss=0.2581, pruned_loss=0.0464, over 15453.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2478, pruned_loss=0.03923, over 3317306.10 frames. ], batch size: 190, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:42:24,099 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-05-01 09:42:59,687 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6633, 3.8270, 2.0035, 4.3929, 2.8838, 4.3680, 2.2286, 3.1591], device='cuda:2'), covar=tensor([0.0363, 0.0440, 0.2138, 0.0317, 0.0929, 0.0426, 0.2071, 0.0804], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0181, 0.0199, 0.0168, 0.0181, 0.0223, 0.0207, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 09:43:06,156 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7178, 2.5339, 2.5036, 4.5876, 2.5447, 2.9668, 2.5448, 2.7140], device='cuda:2'), covar=tensor([0.1130, 0.3416, 0.2988, 0.0470, 0.3795, 0.2312, 0.3443, 0.3301], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0452, 0.0372, 0.0332, 0.0442, 0.0518, 0.0423, 0.0530], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:43:27,999 INFO [train.py:904] (2/8) Epoch 22, batch 1600, loss[loss=0.1977, simple_loss=0.2949, pruned_loss=0.0502, over 16732.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2499, pruned_loss=0.03981, over 3312672.21 frames. ], batch size: 57, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:44:12,580 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:44:32,065 INFO [optim.py:368] (2/8) 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 09:44:37,881 INFO [train.py:904] (2/8) Epoch 22, batch 1650, loss[loss=0.2055, simple_loss=0.2901, pruned_loss=0.06045, over 11996.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2511, pruned_loss=0.04076, over 3313911.78 frames. ], batch size: 247, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:45:02,933 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-01 09:45:20,147 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:45:39,500 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-01 09:45:47,648 INFO [train.py:904] (2/8) Epoch 22, batch 1700, loss[loss=0.1788, simple_loss=0.2746, pruned_loss=0.04154, over 17032.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2529, pruned_loss=0.04117, over 3324369.61 frames. ], batch size: 53, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:46:18,480 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:46:53,103 INFO [optim.py:368] (2/8) 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 09:46:58,548 INFO [train.py:904] (2/8) Epoch 22, batch 1750, loss[loss=0.1883, simple_loss=0.2675, pruned_loss=0.05456, over 16502.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2542, pruned_loss=0.04132, over 3317247.03 frames. ], batch size: 146, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:47:06,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6042, 3.8401, 2.3794, 4.3425, 2.8356, 4.2945, 2.4187, 3.0521], device='cuda:2'), covar=tensor([0.0362, 0.0372, 0.1632, 0.0335, 0.0937, 0.0565, 0.1635, 0.0856], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0198, 0.0168, 0.0181, 0.0223, 0.0206, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 09:47:25,706 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:48:06,728 INFO [train.py:904] (2/8) Epoch 22, batch 1800, loss[loss=0.163, simple_loss=0.2507, pruned_loss=0.0377, over 16683.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2549, pruned_loss=0.04103, over 3311769.28 frames. ], batch size: 134, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:48:31,682 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0783, 3.2409, 3.3746, 2.1755, 3.0832, 3.4331, 3.1721, 2.1000], device='cuda:2'), covar=tensor([0.0532, 0.0126, 0.0059, 0.0452, 0.0111, 0.0098, 0.0100, 0.0420], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0083, 0.0084, 0.0134, 0.0098, 0.0109, 0.0095, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:49:03,137 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 09:49:13,316 INFO [optim.py:368] (2/8) 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 09:49:18,206 INFO [train.py:904] (2/8) Epoch 22, batch 1850, loss[loss=0.147, simple_loss=0.2341, pruned_loss=0.02992, over 15926.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2557, pruned_loss=0.0413, over 3307810.98 frames. ], batch size: 35, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:49:34,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7375, 2.6853, 2.3043, 2.5839, 3.0124, 2.8047, 3.2658, 3.2089], device='cuda:2'), covar=tensor([0.0191, 0.0436, 0.0561, 0.0481, 0.0329, 0.0400, 0.0332, 0.0303], device='cuda:2'), in_proj_covar=tensor([0.0218, 0.0241, 0.0232, 0.0232, 0.0242, 0.0241, 0.0244, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:49:54,749 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 09:50:27,231 INFO [train.py:904] (2/8) Epoch 22, batch 1900, loss[loss=0.1677, simple_loss=0.2485, pruned_loss=0.04341, over 16527.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2556, pruned_loss=0.04114, over 3309727.58 frames. ], batch size: 146, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:50:41,422 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 09:51:31,793 INFO [optim.py:368] (2/8) 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 09:51:36,123 INFO [train.py:904] (2/8) Epoch 22, batch 1950, loss[loss=0.1592, simple_loss=0.2543, pruned_loss=0.03203, over 17260.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2556, pruned_loss=0.04064, over 3308745.28 frames. ], batch size: 52, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:52:44,782 INFO [train.py:904] (2/8) Epoch 22, batch 2000, loss[loss=0.187, simple_loss=0.274, pruned_loss=0.04999, over 12458.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2554, pruned_loss=0.04085, over 3304053.82 frames. ], batch size: 248, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:53:48,737 INFO [optim.py:368] (2/8) 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 09:53:52,746 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5105, 3.4391, 2.6705, 2.1350, 2.2861, 2.2392, 3.5348, 3.0202], device='cuda:2'), covar=tensor([0.2874, 0.0726, 0.1797, 0.3039, 0.2733, 0.2244, 0.0602, 0.1597], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0271, 0.0307, 0.0314, 0.0297, 0.0262, 0.0295, 0.0339], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 09:53:53,919 INFO [train.py:904] (2/8) Epoch 22, batch 2050, loss[loss=0.1745, simple_loss=0.2616, pruned_loss=0.04373, over 16317.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2547, pruned_loss=0.04083, over 3295794.70 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:54:03,871 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7186, 2.8542, 2.7131, 4.9455, 4.0265, 4.4499, 1.5784, 3.2021], device='cuda:2'), covar=tensor([0.1462, 0.0832, 0.1262, 0.0212, 0.0207, 0.0381, 0.1788, 0.0800], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0192, 0.0204, 0.0218, 0.0203, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 09:54:06,190 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8794, 2.8029, 2.4329, 2.7601, 3.1007, 2.9165, 3.5455, 3.3717], device='cuda:2'), covar=tensor([0.0165, 0.0482, 0.0572, 0.0435, 0.0337, 0.0431, 0.0251, 0.0306], device='cuda:2'), in_proj_covar=tensor([0.0218, 0.0241, 0.0231, 0.0231, 0.0241, 0.0241, 0.0243, 0.0236], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:54:29,561 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 09:54:31,421 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215230.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:55:01,197 INFO [train.py:904] (2/8) Epoch 22, batch 2100, loss[loss=0.1631, simple_loss=0.2556, pruned_loss=0.03533, over 17107.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2548, pruned_loss=0.04035, over 3305811.76 frames. ], batch size: 55, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:55:54,360 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215291.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:56:04,629 INFO [optim.py:368] (2/8) 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 09:56:09,028 INFO [train.py:904] (2/8) Epoch 22, batch 2150, loss[loss=0.1696, simple_loss=0.2713, pruned_loss=0.03401, over 17263.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2565, pruned_loss=0.04121, over 3304418.89 frames. ], batch size: 52, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:56:39,660 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215325.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:56:45,687 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 09:56:51,001 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 22, batch 2200, loss[loss=0.1533, simple_loss=0.2466, pruned_loss=0.03002, over 17128.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2578, pruned_loss=0.04165, over 3305620.19 frames. ], batch size: 49, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:58:02,279 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215386.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 09:58:14,369 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215394.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 09:58:20,777 INFO [optim.py:368] (2/8) 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 09:58:21,497 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9068, 5.2778, 5.0600, 5.0419, 4.8402, 4.7520, 4.7047, 5.3959], device='cuda:2'), covar=tensor([0.1251, 0.0857, 0.0910, 0.0916, 0.0764, 0.1016, 0.1157, 0.0811], device='cuda:2'), in_proj_covar=tensor([0.0685, 0.0842, 0.0695, 0.0641, 0.0534, 0.0539, 0.0707, 0.0657], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 09:58:24,662 INFO [train.py:904] (2/8) Epoch 22, batch 2250, loss[loss=0.1589, simple_loss=0.2471, pruned_loss=0.03531, over 17222.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2578, pruned_loss=0.04204, over 3312394.09 frames. ], batch size: 43, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:58:59,961 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2489, 3.4748, 3.6056, 2.3881, 3.3349, 3.7028, 3.4563, 2.1005], device='cuda:2'), covar=tensor([0.0538, 0.0153, 0.0070, 0.0448, 0.0125, 0.0115, 0.0106, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0083, 0.0084, 0.0134, 0.0098, 0.0109, 0.0095, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 09:59:36,620 INFO [train.py:904] (2/8) Epoch 22, batch 2300, loss[loss=0.1527, simple_loss=0.2333, pruned_loss=0.03606, over 16848.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.258, pruned_loss=0.04215, over 3310570.46 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:59:38,684 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-01 10:00:42,879 INFO [optim.py:368] (2/8) 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 10:00:46,496 INFO [train.py:904] (2/8) Epoch 22, batch 2350, loss[loss=0.1921, simple_loss=0.2778, pruned_loss=0.0532, over 15663.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2584, pruned_loss=0.04245, over 3300974.69 frames. ], batch size: 191, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:00:46,887 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:01:05,447 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 10:01:30,696 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4126, 2.9465, 2.6560, 2.3143, 2.2665, 2.2862, 2.9533, 2.8187], device='cuda:2'), covar=tensor([0.2448, 0.0679, 0.1618, 0.2403, 0.2335, 0.1985, 0.0515, 0.1121], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0272, 0.0308, 0.0315, 0.0299, 0.0262, 0.0296, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:01:48,837 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:01:55,002 INFO [train.py:904] (2/8) Epoch 22, batch 2400, loss[loss=0.1762, simple_loss=0.2577, pruned_loss=0.0474, over 16356.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2583, pruned_loss=0.0421, over 3310624.54 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:02:11,593 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215564.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:12,658 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215565.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:30,112 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5531, 2.5352, 2.2281, 2.3723, 2.8827, 2.6383, 3.1335, 3.0756], device='cuda:2'), covar=tensor([0.0178, 0.0460, 0.0507, 0.0469, 0.0292, 0.0416, 0.0289, 0.0288], device='cuda:2'), in_proj_covar=tensor([0.0218, 0.0241, 0.0230, 0.0232, 0.0241, 0.0240, 0.0243, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:02:41,691 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215586.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:02:59,708 INFO [optim.py:368] (2/8) 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 10:03:04,614 INFO [train.py:904] (2/8) Epoch 22, batch 2450, loss[loss=0.1575, simple_loss=0.2448, pruned_loss=0.03511, over 16951.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.259, pruned_loss=0.04224, over 3315376.10 frames. ], batch size: 41, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:03:11,980 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215609.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 10:03:35,715 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215626.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:03:38,622 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 10:04:11,703 INFO [train.py:904] (2/8) Epoch 22, batch 2500, loss[loss=0.1394, simple_loss=0.2264, pruned_loss=0.02623, over 16978.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2589, pruned_loss=0.042, over 3320631.37 frames. ], batch size: 41, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:04:13,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2097, 5.6395, 5.7672, 5.4581, 5.5009, 6.1399, 5.5722, 5.2734], device='cuda:2'), covar=tensor([0.0910, 0.2007, 0.2164, 0.2049, 0.2814, 0.0988, 0.1534, 0.2502], device='cuda:2'), in_proj_covar=tensor([0.0424, 0.0620, 0.0683, 0.0516, 0.0687, 0.0717, 0.0537, 0.0687], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:04:38,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7694, 4.3149, 4.3598, 3.0454, 3.6849, 4.3217, 3.9422, 2.4477], device='cuda:2'), covar=tensor([0.0512, 0.0073, 0.0045, 0.0380, 0.0141, 0.0104, 0.0093, 0.0495], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0134, 0.0099, 0.0109, 0.0095, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:04:51,429 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215681.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:05:02,670 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215689.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:05:11,793 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4177, 4.3971, 4.3050, 3.7804, 4.3489, 1.6779, 4.1285, 3.9274], device='cuda:2'), covar=tensor([0.0132, 0.0113, 0.0198, 0.0294, 0.0100, 0.2976, 0.0149, 0.0235], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0158, 0.0201, 0.0179, 0.0179, 0.0210, 0.0191, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:05:15,511 INFO [optim.py:368] (2/8) 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 10:05:20,562 INFO [train.py:904] (2/8) Epoch 22, batch 2550, loss[loss=0.1951, simple_loss=0.2849, pruned_loss=0.05268, over 17053.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2591, pruned_loss=0.04185, over 3327272.11 frames. ], batch size: 53, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:05:23,500 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0231, 4.4949, 3.2427, 2.5048, 2.8670, 2.7216, 4.8665, 3.7774], device='cuda:2'), covar=tensor([0.2706, 0.0596, 0.1771, 0.2817, 0.3039, 0.2057, 0.0385, 0.1424], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0272, 0.0308, 0.0316, 0.0299, 0.0262, 0.0296, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:06:25,128 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5116, 3.1813, 3.5763, 1.8926, 3.6409, 3.6990, 2.9676, 2.7605], device='cuda:2'), covar=tensor([0.0761, 0.0285, 0.0173, 0.1190, 0.0118, 0.0206, 0.0425, 0.0451], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0140, 0.0081, 0.0128, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:06:30,386 INFO [train.py:904] (2/8) Epoch 22, batch 2600, loss[loss=0.1701, simple_loss=0.2576, pruned_loss=0.04126, over 17215.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2588, pruned_loss=0.0418, over 3323927.37 frames. ], batch size: 45, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:07:01,373 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 10:07:36,117 INFO [optim.py:368] (2/8) 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 10:07:39,998 INFO [train.py:904] (2/8) Epoch 22, batch 2650, loss[loss=0.1671, simple_loss=0.2604, pruned_loss=0.03689, over 17245.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.259, pruned_loss=0.04141, over 3322526.93 frames. ], batch size: 45, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:08:16,609 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215829.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:08:48,040 INFO [train.py:904] (2/8) Epoch 22, batch 2700, loss[loss=0.1618, simple_loss=0.242, pruned_loss=0.04078, over 16812.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2587, pruned_loss=0.04094, over 3330240.52 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:08:57,510 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:22,072 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 10:09:34,780 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215886.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:09:39,271 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:53,923 INFO [optim.py:368] (2/8) 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 10:09:57,374 INFO [train.py:904] (2/8) Epoch 22, batch 2750, loss[loss=0.1721, simple_loss=0.2516, pruned_loss=0.0463, over 16499.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2589, pruned_loss=0.04074, over 3335250.64 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:09:59,458 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215904.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:10:20,152 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7670, 2.6819, 2.4305, 2.6212, 3.0399, 2.7804, 3.1953, 3.1691], device='cuda:2'), covar=tensor([0.0198, 0.0443, 0.0496, 0.0448, 0.0295, 0.0389, 0.0313, 0.0280], device='cuda:2'), in_proj_covar=tensor([0.0218, 0.0240, 0.0230, 0.0231, 0.0241, 0.0240, 0.0243, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:10:21,930 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:10:38,798 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=215934.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:11:05,009 INFO [train.py:904] (2/8) Epoch 22, batch 2800, loss[loss=0.1713, simple_loss=0.2648, pruned_loss=0.03889, over 16669.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2586, pruned_loss=0.04053, over 3336365.92 frames. ], batch size: 62, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:11:42,292 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215981.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:11:53,840 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215989.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:12:07,418 INFO [optim.py:368] (2/8) 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 10:12:14,945 INFO [train.py:904] (2/8) Epoch 22, batch 2850, loss[loss=0.1464, simple_loss=0.2292, pruned_loss=0.03175, over 16845.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2582, pruned_loss=0.04063, over 3332579.31 frames. ], batch size: 102, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:12:51,478 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216028.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:12:52,495 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216029.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:13:03,977 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:13:20,031 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 10:13:26,222 INFO [train.py:904] (2/8) Epoch 22, batch 2900, loss[loss=0.1517, simple_loss=0.2487, pruned_loss=0.02737, over 17153.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2569, pruned_loss=0.04067, over 3336624.97 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:13:33,140 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7664, 2.6775, 2.3757, 2.6052, 3.0618, 2.8315, 3.3818, 3.2784], device='cuda:2'), covar=tensor([0.0160, 0.0440, 0.0510, 0.0460, 0.0303, 0.0408, 0.0225, 0.0279], device='cuda:2'), in_proj_covar=tensor([0.0218, 0.0240, 0.0230, 0.0230, 0.0241, 0.0240, 0.0243, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:13:38,347 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8477, 5.0340, 5.2268, 4.9916, 4.9578, 5.6819, 5.2220, 4.9051], device='cuda:2'), covar=tensor([0.1359, 0.2198, 0.2608, 0.2240, 0.2880, 0.1046, 0.1855, 0.2764], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0615, 0.0676, 0.0513, 0.0682, 0.0712, 0.0533, 0.0681], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:14:17,775 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:14:31,570 INFO [optim.py:368] (2/8) 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 10:14:35,749 INFO [train.py:904] (2/8) Epoch 22, batch 2950, loss[loss=0.1543, simple_loss=0.2498, pruned_loss=0.02941, over 17191.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2569, pruned_loss=0.04124, over 3330799.06 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:45,385 INFO [train.py:904] (2/8) Epoch 22, batch 3000, loss[loss=0.1599, simple_loss=0.2407, pruned_loss=0.0396, over 16745.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2565, pruned_loss=0.0413, over 3326678.50 frames. ], batch size: 134, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:45,385 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 10:15:50,147 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9442, 3.9027, 3.7930, 3.2067, 3.8253, 2.0770, 3.6919, 3.1971], device='cuda:2'), covar=tensor([0.0110, 0.0096, 0.0158, 0.0204, 0.0093, 0.2640, 0.0102, 0.0284], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0159, 0.0202, 0.0181, 0.0181, 0.0211, 0.0192, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:15:54,101 INFO [train.py:938] (2/8) Epoch 22, validation: loss=0.1347, simple_loss=0.2399, pruned_loss=0.0148, over 944034.00 frames. 2023-05-01 10:15:54,103 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 10:16:02,788 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:59,839 INFO [optim.py:368] (2/8) 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 10:17:03,730 INFO [train.py:904] (2/8) Epoch 22, batch 3050, loss[loss=0.1731, simple_loss=0.2714, pruned_loss=0.03737, over 17108.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2564, pruned_loss=0.0416, over 3330348.53 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:17:05,250 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216204.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:17:09,668 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216207.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:29,296 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:57,820 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4373, 3.2166, 3.5052, 2.0287, 3.5752, 3.6059, 2.8885, 2.6914], device='cuda:2'), covar=tensor([0.0780, 0.0246, 0.0181, 0.1053, 0.0112, 0.0217, 0.0452, 0.0446], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0140, 0.0081, 0.0129, 0.0131, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:18:10,763 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216252.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:18:12,222 INFO [train.py:904] (2/8) Epoch 22, batch 3100, loss[loss=0.1663, simple_loss=0.2445, pruned_loss=0.04409, over 16477.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2563, pruned_loss=0.04131, over 3328804.48 frames. ], batch size: 75, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:18:33,052 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0792, 2.1487, 2.2959, 3.6390, 2.2068, 2.4140, 2.2713, 2.2872], device='cuda:2'), covar=tensor([0.1490, 0.3595, 0.2946, 0.0717, 0.3926, 0.2725, 0.3588, 0.3527], device='cuda:2'), in_proj_covar=tensor([0.0406, 0.0452, 0.0372, 0.0332, 0.0439, 0.0520, 0.0423, 0.0530], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:18:33,891 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216269.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:18:58,762 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2625, 5.2089, 5.1221, 4.6800, 4.7627, 5.1571, 5.0685, 4.7695], device='cuda:2'), covar=tensor([0.0572, 0.0473, 0.0310, 0.0357, 0.1083, 0.0460, 0.0323, 0.0774], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0448, 0.0358, 0.0358, 0.0369, 0.0416, 0.0246, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:19:16,924 INFO [optim.py:368] (2/8) 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 10:19:21,077 INFO [train.py:904] (2/8) Epoch 22, batch 3150, loss[loss=0.2051, simple_loss=0.283, pruned_loss=0.06359, over 12310.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2559, pruned_loss=0.04187, over 3313543.20 frames. ], batch size: 246, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:19:44,876 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3811, 5.3762, 5.2664, 4.8100, 4.8484, 5.3233, 5.2675, 4.9416], device='cuda:2'), covar=tensor([0.0659, 0.0551, 0.0335, 0.0377, 0.1168, 0.0516, 0.0279, 0.0788], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0449, 0.0359, 0.0359, 0.0370, 0.0416, 0.0247, 0.0434], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:20:29,459 INFO [train.py:904] (2/8) Epoch 22, batch 3200, loss[loss=0.1653, simple_loss=0.2598, pruned_loss=0.03542, over 16747.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.255, pruned_loss=0.04134, over 3313398.94 frames. ], batch size: 62, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:21:13,058 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 10:21:13,818 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 10:21:40,319 INFO [train.py:904] (2/8) Epoch 22, batch 3250, loss[loss=0.1666, simple_loss=0.2435, pruned_loss=0.0449, over 16280.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2551, pruned_loss=0.04195, over 3317619.08 frames. ], batch size: 36, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:22:37,784 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 10:22:49,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6252, 2.5429, 2.1610, 2.3011, 2.9059, 2.6008, 3.2450, 3.1537], device='cuda:2'), covar=tensor([0.0199, 0.0486, 0.0644, 0.0557, 0.0351, 0.0475, 0.0302, 0.0336], device='cuda:2'), in_proj_covar=tensor([0.0219, 0.0240, 0.0230, 0.0230, 0.0240, 0.0240, 0.0244, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:22:52,220 INFO [train.py:904] (2/8) Epoch 22, batch 3300, loss[loss=0.1675, simple_loss=0.2551, pruned_loss=0.03998, over 16787.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2559, pruned_loss=0.04143, over 3328291.59 frames. ], batch size: 102, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:23:33,527 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3768, 4.2256, 4.4217, 4.5641, 4.6863, 4.2791, 4.4647, 4.6805], device='cuda:2'), covar=tensor([0.1634, 0.1167, 0.1256, 0.0665, 0.0585, 0.1164, 0.2235, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0667, 0.0828, 0.0960, 0.0839, 0.0632, 0.0663, 0.0683, 0.0792], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:23:36,592 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216485.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:23:56,628 INFO [optim.py:368] (2/8) 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 10:24:00,075 INFO [train.py:904] (2/8) Epoch 22, batch 3350, loss[loss=0.16, simple_loss=0.2454, pruned_loss=0.0373, over 16724.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2557, pruned_loss=0.041, over 3329667.94 frames. ], batch size: 83, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:24:42,589 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:25:10,644 INFO [train.py:904] (2/8) Epoch 22, batch 3400, loss[loss=0.1515, simple_loss=0.2345, pruned_loss=0.03419, over 16992.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2555, pruned_loss=0.04097, over 3330610.50 frames. ], batch size: 41, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:25:20,890 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7302, 2.3934, 2.3852, 3.5485, 2.6804, 3.7333, 1.5720, 2.7301], device='cuda:2'), covar=tensor([0.1479, 0.0830, 0.1238, 0.0235, 0.0202, 0.0406, 0.1767, 0.0866], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0194, 0.0206, 0.0219, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:25:47,209 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9732, 2.1186, 2.5485, 2.8394, 2.8743, 3.0807, 2.1048, 3.1883], device='cuda:2'), covar=tensor([0.0200, 0.0486, 0.0369, 0.0274, 0.0324, 0.0253, 0.0589, 0.0158], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0197, 0.0183, 0.0189, 0.0201, 0.0157, 0.0199, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:25:48,617 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-01 10:26:02,833 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:26:15,741 INFO [optim.py:368] (2/8) 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 10:26:19,664 INFO [train.py:904] (2/8) Epoch 22, batch 3450, loss[loss=0.1449, simple_loss=0.2407, pruned_loss=0.02457, over 17216.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2546, pruned_loss=0.04031, over 3336129.84 frames. ], batch size: 45, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:26:51,280 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4761, 4.4327, 4.3818, 3.9074, 4.4373, 1.6285, 4.2359, 4.0143], device='cuda:2'), covar=tensor([0.0123, 0.0116, 0.0193, 0.0283, 0.0097, 0.2942, 0.0140, 0.0233], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0160, 0.0205, 0.0183, 0.0183, 0.0212, 0.0194, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:27:29,237 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 22, batch 3500, loss[loss=0.174, simple_loss=0.2672, pruned_loss=0.04038, over 17017.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2539, pruned_loss=0.03999, over 3336009.38 frames. ], batch size: 50, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:27:32,805 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216655.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:27:38,577 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 10:28:06,691 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1014, 2.2726, 2.7634, 3.0179, 2.8823, 3.5966, 2.6029, 3.4954], device='cuda:2'), covar=tensor([0.0268, 0.0500, 0.0340, 0.0338, 0.0358, 0.0185, 0.0443, 0.0181], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0198, 0.0184, 0.0190, 0.0202, 0.0158, 0.0200, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:28:12,552 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216684.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:35,770 INFO [optim.py:368] (2/8) 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 10:28:39,276 INFO [train.py:904] (2/8) Epoch 22, batch 3550, loss[loss=0.1667, simple_loss=0.2591, pruned_loss=0.03715, over 16699.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.254, pruned_loss=0.03963, over 3325329.10 frames. ], batch size: 62, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:28:42,529 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:57,361 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216716.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:29:20,498 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216732.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:29:45,294 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8746, 2.7172, 2.6955, 4.6412, 3.4711, 4.2295, 1.9433, 3.0654], device='cuda:2'), covar=tensor([0.1403, 0.0872, 0.1211, 0.0232, 0.0246, 0.0433, 0.1554, 0.0844], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0195, 0.0206, 0.0219, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:29:49,493 INFO [train.py:904] (2/8) Epoch 22, batch 3600, loss[loss=0.1431, simple_loss=0.2313, pruned_loss=0.02749, over 16795.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2528, pruned_loss=0.03934, over 3326633.67 frames. ], batch size: 39, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:29:52,323 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8712, 2.7892, 2.3637, 2.7368, 3.1624, 2.9511, 3.4430, 3.3948], device='cuda:2'), covar=tensor([0.0139, 0.0463, 0.0563, 0.0438, 0.0305, 0.0378, 0.0271, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0221, 0.0242, 0.0232, 0.0232, 0.0242, 0.0243, 0.0247, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:30:08,399 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:30:51,305 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 10:31:00,614 INFO [optim.py:368] (2/8) 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 10:31:03,547 INFO [train.py:904] (2/8) Epoch 22, batch 3650, loss[loss=0.1868, simple_loss=0.2613, pruned_loss=0.05617, over 11250.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2507, pruned_loss=0.03969, over 3313353.52 frames. ], batch size: 245, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:32:13,484 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-01 10:32:18,424 INFO [train.py:904] (2/8) Epoch 22, batch 3700, loss[loss=0.1897, simple_loss=0.2712, pruned_loss=0.05414, over 16557.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2498, pruned_loss=0.04144, over 3294334.67 frames. ], batch size: 62, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:32:52,379 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 10:32:53,744 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6592, 4.6261, 4.5771, 4.1046, 4.6233, 1.8061, 4.4368, 4.2795], device='cuda:2'), covar=tensor([0.0121, 0.0120, 0.0189, 0.0308, 0.0100, 0.2726, 0.0140, 0.0212], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0161, 0.0205, 0.0182, 0.0183, 0.0212, 0.0194, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:33:31,442 INFO [optim.py:368] (2/8) 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 10:33:32,645 INFO [train.py:904] (2/8) Epoch 22, batch 3750, loss[loss=0.1714, simple_loss=0.2646, pruned_loss=0.03914, over 17250.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2504, pruned_loss=0.04275, over 3270425.99 frames. ], batch size: 45, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:34:11,434 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5538, 4.5887, 4.9283, 4.9193, 4.9691, 4.6146, 4.6251, 4.4461], device='cuda:2'), covar=tensor([0.0384, 0.0697, 0.0378, 0.0422, 0.0495, 0.0432, 0.0874, 0.0672], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0467, 0.0453, 0.0420, 0.0502, 0.0477, 0.0562, 0.0382], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 10:34:31,446 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3443, 2.5321, 2.4460, 4.3453, 2.3329, 2.8371, 2.5583, 2.7114], device='cuda:2'), covar=tensor([0.1331, 0.3193, 0.2680, 0.0441, 0.3765, 0.2273, 0.3364, 0.2841], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0455, 0.0374, 0.0334, 0.0442, 0.0525, 0.0426, 0.0534], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:34:36,639 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216947.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:34:44,479 INFO [train.py:904] (2/8) Epoch 22, batch 3800, loss[loss=0.1821, simple_loss=0.2595, pruned_loss=0.05236, over 16836.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2528, pruned_loss=0.04433, over 3262864.67 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:35:03,071 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:35:55,035 INFO [optim.py:368] (2/8) 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 10:35:56,838 INFO [train.py:904] (2/8) Epoch 22, batch 3850, loss[loss=0.1599, simple_loss=0.2397, pruned_loss=0.04001, over 16795.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2535, pruned_loss=0.04513, over 3251828.19 frames. ], batch size: 83, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:36:08,428 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217011.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:36:29,873 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:37:09,496 INFO [train.py:904] (2/8) Epoch 22, batch 3900, loss[loss=0.1604, simple_loss=0.247, pruned_loss=0.03688, over 17042.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2527, pruned_loss=0.04531, over 3253289.17 frames. ], batch size: 50, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:37:22,071 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217061.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:37:40,295 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0385, 5.3622, 5.1159, 5.1509, 4.9271, 4.7599, 4.8298, 5.4590], device='cuda:2'), covar=tensor([0.1197, 0.0797, 0.0888, 0.0887, 0.0744, 0.1042, 0.1163, 0.0850], device='cuda:2'), in_proj_covar=tensor([0.0696, 0.0853, 0.0701, 0.0647, 0.0540, 0.0548, 0.0714, 0.0667], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:38:21,558 INFO [optim.py:368] (2/8) 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 10:38:22,849 INFO [train.py:904] (2/8) Epoch 22, batch 3950, loss[loss=0.201, simple_loss=0.2773, pruned_loss=0.06232, over 17069.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2519, pruned_loss=0.04588, over 3270067.73 frames. ], batch size: 55, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:38:40,778 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8697, 5.1632, 5.3688, 5.1150, 5.1981, 5.7627, 5.2375, 4.9442], device='cuda:2'), covar=tensor([0.1150, 0.1897, 0.2207, 0.1911, 0.2521, 0.0987, 0.1501, 0.2374], device='cuda:2'), in_proj_covar=tensor([0.0420, 0.0607, 0.0665, 0.0504, 0.0672, 0.0699, 0.0522, 0.0671], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:38:49,527 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-05-01 10:39:35,767 INFO [train.py:904] (2/8) Epoch 22, batch 4000, loss[loss=0.1729, simple_loss=0.2529, pruned_loss=0.04647, over 17040.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2519, pruned_loss=0.04629, over 3278139.09 frames. ], batch size: 55, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:39:38,884 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-01 10:40:48,067 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 22, batch 4050, loss[loss=0.1608, simple_loss=0.2506, pruned_loss=0.03555, over 15410.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2524, pruned_loss=0.04551, over 3272129.35 frames. ], batch size: 190, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:41:49,060 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 10:41:55,314 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:42:04,919 INFO [train.py:904] (2/8) Epoch 22, batch 4100, loss[loss=0.2096, simple_loss=0.2853, pruned_loss=0.067, over 12228.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2545, pruned_loss=0.04544, over 3263485.45 frames. ], batch size: 248, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:43:11,027 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 10:43:23,203 INFO [train.py:904] (2/8) Epoch 22, batch 4150, loss[loss=0.1607, simple_loss=0.2623, pruned_loss=0.02958, over 16859.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2615, pruned_loss=0.04773, over 3232112.88 frames. ], batch size: 96, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:43:28,282 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 10:43:36,177 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217311.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:43:52,136 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:44:27,293 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 10:44:28,727 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0589, 2.1800, 2.6527, 2.9790, 3.0011, 3.6256, 2.2643, 3.4344], device='cuda:2'), covar=tensor([0.0228, 0.0488, 0.0317, 0.0306, 0.0279, 0.0143, 0.0523, 0.0135], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0195, 0.0181, 0.0187, 0.0198, 0.0156, 0.0197, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:44:39,672 INFO [train.py:904] (2/8) Epoch 22, batch 4200, loss[loss=0.1981, simple_loss=0.2946, pruned_loss=0.05074, over 16751.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2685, pruned_loss=0.04921, over 3207684.37 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:44:50,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217359.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:44:52,721 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:32,805 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217388.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:42,708 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-01 10:45:53,906 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 22, batch 4250, loss[loss=0.1759, simple_loss=0.2703, pruned_loss=0.04074, over 16762.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2713, pruned_loss=0.04858, over 3210162.50 frames. ], batch size: 39, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:46:04,662 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217409.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:46:44,499 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7195, 1.5097, 2.2778, 2.5692, 2.6503, 2.9530, 1.7006, 2.8518], device='cuda:2'), covar=tensor([0.0181, 0.0638, 0.0306, 0.0286, 0.0275, 0.0167, 0.0707, 0.0128], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0186, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 10:47:04,240 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217449.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:47:09,154 INFO [train.py:904] (2/8) Epoch 22, batch 4300, loss[loss=0.1914, simple_loss=0.2859, pruned_loss=0.04851, over 17051.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2726, pruned_loss=0.04741, over 3227940.52 frames. ], batch size: 50, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:48:07,829 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217492.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:48:23,053 INFO [optim.py:368] (2/8) 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 10:48:24,299 INFO [train.py:904] (2/8) Epoch 22, batch 4350, loss[loss=0.2129, simple_loss=0.2963, pruned_loss=0.06469, over 11340.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2754, pruned_loss=0.04866, over 3205793.20 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:39,977 INFO [train.py:904] (2/8) Epoch 22, batch 4400, loss[loss=0.1942, simple_loss=0.2803, pruned_loss=0.05405, over 16748.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2781, pruned_loss=0.05033, over 3190081.40 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:41,102 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:50:00,612 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 10:50:13,926 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6670, 3.7217, 3.9345, 2.2299, 3.1775, 2.4504, 3.9316, 4.0452], device='cuda:2'), covar=tensor([0.0167, 0.0724, 0.0510, 0.2044, 0.0801, 0.0969, 0.0493, 0.0838], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0166, 0.0153, 0.0145, 0.0130, 0.0143, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:50:52,393 INFO [optim.py:368] (2/8) 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 10:50:53,484 INFO [train.py:904] (2/8) Epoch 22, batch 4450, loss[loss=0.198, simple_loss=0.2792, pruned_loss=0.05845, over 11915.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2812, pruned_loss=0.05143, over 3205541.52 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:50:59,495 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5926, 2.9457, 3.1263, 1.9019, 2.6734, 2.0516, 3.1558, 3.1969], device='cuda:2'), covar=tensor([0.0258, 0.0846, 0.0595, 0.2199, 0.0904, 0.1117, 0.0637, 0.0900], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0166, 0.0152, 0.0144, 0.0130, 0.0143, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:51:20,493 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:51:22,643 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 10:51:48,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7146, 3.8446, 2.4465, 4.6067, 3.0102, 4.4780, 2.6228, 3.0938], device='cuda:2'), covar=tensor([0.0293, 0.0332, 0.1572, 0.0133, 0.0776, 0.0497, 0.1352, 0.0774], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0177, 0.0194, 0.0164, 0.0177, 0.0219, 0.0202, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:52:08,355 INFO [train.py:904] (2/8) Epoch 22, batch 4500, loss[loss=0.1906, simple_loss=0.273, pruned_loss=0.05412, over 16858.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2813, pruned_loss=0.05215, over 3194394.55 frames. ], batch size: 42, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:52:24,050 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:52:32,189 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 10:53:19,301 INFO [train.py:904] (2/8) Epoch 22, batch 4550, loss[loss=0.2238, simple_loss=0.3048, pruned_loss=0.07137, over 16695.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2826, pruned_loss=0.05326, over 3213212.67 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:53:19,777 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6883, 2.7590, 2.4510, 4.1285, 3.1800, 3.9410, 1.4827, 2.9263], device='cuda:2'), covar=tensor([0.1356, 0.0787, 0.1301, 0.0145, 0.0307, 0.0392, 0.1791, 0.0829], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0192, 0.0206, 0.0216, 0.0202, 0.0192], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:53:30,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9009, 4.2200, 3.1398, 2.5298, 3.0005, 2.5933, 4.8257, 3.7639], device='cuda:2'), covar=tensor([0.2737, 0.0561, 0.1682, 0.2352, 0.2433, 0.2032, 0.0348, 0.1113], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0271, 0.0306, 0.0316, 0.0300, 0.0261, 0.0296, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:53:52,637 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:13,402 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 10:54:19,809 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 22, batch 4600, loss[loss=0.1746, simple_loss=0.2648, pruned_loss=0.04216, over 16709.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2834, pruned_loss=0.05347, over 3225038.22 frames. ], batch size: 76, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:54:52,671 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8732, 4.1199, 3.1239, 2.5137, 2.8198, 2.5240, 4.6514, 3.6798], device='cuda:2'), covar=tensor([0.2694, 0.0559, 0.1642, 0.2289, 0.2381, 0.1987, 0.0341, 0.1065], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0270, 0.0305, 0.0315, 0.0299, 0.0260, 0.0295, 0.0339], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:55:20,851 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6170, 2.4594, 1.8417, 2.6834, 2.1094, 2.7491, 2.1404, 2.3648], device='cuda:2'), covar=tensor([0.0372, 0.0358, 0.1359, 0.0203, 0.0755, 0.0434, 0.1232, 0.0661], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0164, 0.0178, 0.0221, 0.0203, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:55:41,528 INFO [optim.py:368] (2/8) 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 10:55:42,813 INFO [train.py:904] (2/8) Epoch 22, batch 4650, loss[loss=0.1688, simple_loss=0.2606, pruned_loss=0.03848, over 16655.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2833, pruned_loss=0.05374, over 3215830.97 frames. ], batch size: 76, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:56:46,084 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217848.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:56:53,004 INFO [train.py:904] (2/8) Epoch 22, batch 4700, loss[loss=0.1806, simple_loss=0.2651, pruned_loss=0.04806, over 16584.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2805, pruned_loss=0.05234, over 3220318.56 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:57:01,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6147, 4.3743, 4.3149, 2.9221, 3.7360, 4.3775, 3.7528, 2.7546], device='cuda:2'), covar=tensor([0.0535, 0.0039, 0.0043, 0.0382, 0.0106, 0.0078, 0.0111, 0.0375], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0083, 0.0084, 0.0133, 0.0099, 0.0110, 0.0095, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 10:57:16,765 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7304, 3.6803, 2.8853, 2.3660, 2.5820, 2.3884, 4.1825, 3.3707], device='cuda:2'), covar=tensor([0.2909, 0.0800, 0.1918, 0.2674, 0.2573, 0.2211, 0.0440, 0.1261], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0270, 0.0306, 0.0316, 0.0299, 0.0261, 0.0296, 0.0340], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 10:57:38,776 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8758, 3.9792, 2.4314, 4.8203, 3.1400, 4.6672, 2.5772, 3.1394], device='cuda:2'), covar=tensor([0.0304, 0.0367, 0.1833, 0.0123, 0.0880, 0.0436, 0.1568, 0.0877], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0164, 0.0178, 0.0220, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 10:57:59,177 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217898.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:58:04,666 INFO [optim.py:368] (2/8) 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 10:58:05,922 INFO [train.py:904] (2/8) Epoch 22, batch 4750, loss[loss=0.184, simple_loss=0.279, pruned_loss=0.0445, over 16450.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2762, pruned_loss=0.05002, over 3235875.92 frames. ], batch size: 146, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:58:49,643 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 10:59:17,330 INFO [train.py:904] (2/8) Epoch 22, batch 4800, loss[loss=0.1697, simple_loss=0.252, pruned_loss=0.04364, over 16632.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2728, pruned_loss=0.04831, over 3219907.40 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:59:27,593 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217959.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:00:36,339 INFO [optim.py:368] (2/8) 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 11:00:36,354 INFO [train.py:904] (2/8) Epoch 22, batch 4850, loss[loss=0.1919, simple_loss=0.2831, pruned_loss=0.05035, over 16765.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2734, pruned_loss=0.04786, over 3195554.96 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:01:01,606 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218020.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:38,581 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:51,558 INFO [train.py:904] (2/8) Epoch 22, batch 4900, loss[loss=0.1713, simple_loss=0.2638, pruned_loss=0.03939, over 16771.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2718, pruned_loss=0.04613, over 3186113.63 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:02:49,275 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218092.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:01,570 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 11:03:05,463 INFO [train.py:904] (2/8) Epoch 22, batch 4950, loss[loss=0.1745, simple_loss=0.2719, pruned_loss=0.03858, over 16456.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.272, pruned_loss=0.04576, over 3200975.71 frames. ], batch size: 68, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:03:20,838 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 11:04:11,470 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 22, batch 5000, loss[loss=0.2118, simple_loss=0.2968, pruned_loss=0.06342, over 12224.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2737, pruned_loss=0.04587, over 3193004.91 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:04:30,118 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:04:30,264 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 11:04:56,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9976, 3.8670, 3.8287, 2.5234, 3.4297, 3.9115, 3.4112, 1.8563], device='cuda:2'), covar=tensor([0.0632, 0.0048, 0.0050, 0.0399, 0.0106, 0.0078, 0.0114, 0.0559], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0082, 0.0084, 0.0132, 0.0098, 0.0109, 0.0094, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 11:05:00,898 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218182.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:21,357 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218196.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:32,359 INFO [optim.py:368] (2/8) 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 11:05:32,374 INFO [train.py:904] (2/8) Epoch 22, batch 5050, loss[loss=0.2082, simple_loss=0.2895, pruned_loss=0.06349, over 11873.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2747, pruned_loss=0.04583, over 3197039.60 frames. ], batch size: 247, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:05:35,929 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3704, 4.3361, 4.2933, 3.4689, 4.3082, 1.5948, 4.0147, 3.9620], device='cuda:2'), covar=tensor([0.0116, 0.0112, 0.0172, 0.0476, 0.0111, 0.2967, 0.0153, 0.0265], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0155, 0.0198, 0.0178, 0.0176, 0.0207, 0.0188, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:05:42,358 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:06:30,538 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:06:44,148 INFO [train.py:904] (2/8) Epoch 22, batch 5100, loss[loss=0.1789, simple_loss=0.2738, pruned_loss=0.04198, over 16642.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2733, pruned_loss=0.04536, over 3192142.80 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:06:45,786 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218254.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:07:10,589 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218271.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:07:22,624 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9809, 3.8422, 4.0516, 4.1534, 4.2874, 3.8643, 4.2233, 4.3110], device='cuda:2'), covar=tensor([0.1479, 0.1222, 0.1235, 0.0669, 0.0465, 0.1410, 0.0733, 0.0589], device='cuda:2'), in_proj_covar=tensor([0.0631, 0.0784, 0.0907, 0.0795, 0.0598, 0.0630, 0.0649, 0.0752], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:07:24,309 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6048, 3.0586, 3.0890, 1.9592, 2.7416, 2.1713, 3.0791, 3.2860], device='cuda:2'), covar=tensor([0.0366, 0.0784, 0.0680, 0.2010, 0.0902, 0.0996, 0.0773, 0.0853], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0164, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 11:07:46,732 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9846, 2.2808, 1.9259, 2.1439, 2.7143, 2.3757, 2.5073, 2.8957], device='cuda:2'), covar=tensor([0.0161, 0.0525, 0.0640, 0.0528, 0.0272, 0.0433, 0.0250, 0.0297], device='cuda:2'), in_proj_covar=tensor([0.0210, 0.0233, 0.0225, 0.0225, 0.0234, 0.0234, 0.0236, 0.0231], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:07:57,331 INFO [optim.py:368] (2/8) 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 11:07:57,354 INFO [train.py:904] (2/8) Epoch 22, batch 5150, loss[loss=0.1831, simple_loss=0.2837, pruned_loss=0.04125, over 16137.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2731, pruned_loss=0.04486, over 3179729.66 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:08:08,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3985, 2.3820, 2.3402, 4.0644, 2.2859, 2.7419, 2.4175, 2.5551], device='cuda:2'), covar=tensor([0.1351, 0.3466, 0.2969, 0.0543, 0.4126, 0.2494, 0.3667, 0.3137], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0449, 0.0368, 0.0328, 0.0436, 0.0517, 0.0421, 0.0526], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:08:23,140 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:08:58,537 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7945, 2.5660, 2.3023, 3.1620, 2.0065, 3.5640, 1.6257, 2.8479], device='cuda:2'), covar=tensor([0.1313, 0.0745, 0.1262, 0.0139, 0.0124, 0.0341, 0.1643, 0.0746], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0189, 0.0204, 0.0213, 0.0200, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 11:08:58,579 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5082, 3.3187, 2.7813, 2.2472, 2.2425, 2.3515, 3.4351, 2.9828], device='cuda:2'), covar=tensor([0.2699, 0.0600, 0.1723, 0.2738, 0.2539, 0.2068, 0.0549, 0.1276], device='cuda:2'), in_proj_covar=tensor([0.0325, 0.0269, 0.0304, 0.0314, 0.0298, 0.0259, 0.0295, 0.0338], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 11:08:59,843 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-05-01 11:09:11,242 INFO [train.py:904] (2/8) Epoch 22, batch 5200, loss[loss=0.2161, simple_loss=0.2967, pruned_loss=0.06781, over 12102.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2715, pruned_loss=0.04443, over 3186960.62 frames. ], batch size: 246, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:09:24,732 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2543, 4.3271, 4.5884, 4.5674, 4.5653, 4.3398, 4.3015, 4.2520], device='cuda:2'), covar=tensor([0.0300, 0.0508, 0.0357, 0.0363, 0.0412, 0.0357, 0.0785, 0.0448], device='cuda:2'), in_proj_covar=tensor([0.0395, 0.0438, 0.0425, 0.0393, 0.0470, 0.0444, 0.0530, 0.0356], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 11:09:28,259 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-05-01 11:09:33,041 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) Epoch 22, batch 5250, loss[loss=0.1866, simple_loss=0.2702, pruned_loss=0.05155, over 12522.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2691, pruned_loss=0.04414, over 3182802.45 frames. ], batch size: 246, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:37,150 INFO [train.py:904] (2/8) Epoch 22, batch 5300, loss[loss=0.143, simple_loss=0.2325, pruned_loss=0.0268, over 16785.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2654, pruned_loss=0.04296, over 3191733.28 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:41,559 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218456.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:11:51,772 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1515, 1.5680, 1.9714, 2.1057, 2.2811, 2.4529, 1.8928, 2.3501], device='cuda:2'), covar=tensor([0.0228, 0.0554, 0.0279, 0.0407, 0.0329, 0.0256, 0.0508, 0.0155], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0196, 0.0181, 0.0186, 0.0199, 0.0154, 0.0199, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:12:51,254 INFO [optim.py:368] (2/8) 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 11:12:51,269 INFO [train.py:904] (2/8) Epoch 22, batch 5350, loss[loss=0.1931, simple_loss=0.2845, pruned_loss=0.05084, over 15340.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2649, pruned_loss=0.04273, over 3180268.51 frames. ], batch size: 190, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:13:00,650 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:13:21,803 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-01 11:13:43,237 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:03,263 INFO [train.py:904] (2/8) Epoch 22, batch 5400, loss[loss=0.1952, simple_loss=0.2758, pruned_loss=0.05725, over 16828.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2665, pruned_loss=0.04244, over 3200375.93 frames. ], batch size: 42, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:14:06,171 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218554.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:14:23,144 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218566.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:14:28,907 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:34,763 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:47,877 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 11:15:18,765 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218602.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:15:19,489 INFO [optim.py:368] (2/8) 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 11:15:19,508 INFO [train.py:904] (2/8) Epoch 22, batch 5450, loss[loss=0.2202, simple_loss=0.2975, pruned_loss=0.07145, over 11950.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2693, pruned_loss=0.04415, over 3170870.93 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:16:10,952 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218635.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:16:37,991 INFO [train.py:904] (2/8) Epoch 22, batch 5500, loss[loss=0.2062, simple_loss=0.2976, pruned_loss=0.05742, over 17098.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2756, pruned_loss=0.0478, over 3158950.78 frames. ], batch size: 49, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:17:57,929 INFO [optim.py:368] (2/8) 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 11:17:57,945 INFO [train.py:904] (2/8) Epoch 22, batch 5550, loss[loss=0.2036, simple_loss=0.2908, pruned_loss=0.05822, over 17027.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2829, pruned_loss=0.0532, over 3111472.59 frames. ], batch size: 55, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:13,561 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5130, 4.5186, 4.8820, 4.8413, 4.8705, 4.5864, 4.5633, 4.4287], device='cuda:2'), covar=tensor([0.0327, 0.0623, 0.0407, 0.0435, 0.0460, 0.0444, 0.0883, 0.0536], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0446, 0.0432, 0.0399, 0.0478, 0.0452, 0.0537, 0.0362], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 11:19:20,438 INFO [train.py:904] (2/8) Epoch 22, batch 5600, loss[loss=0.2456, simple_loss=0.3259, pruned_loss=0.08264, over 15253.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2871, pruned_loss=0.05704, over 3094398.73 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:25,770 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218756.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:20:41,872 INFO [optim.py:368] (2/8) 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 11:20:41,893 INFO [train.py:904] (2/8) Epoch 22, batch 5650, loss[loss=0.237, simple_loss=0.3202, pruned_loss=0.07685, over 16840.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2924, pruned_loss=0.06096, over 3071099.67 frames. ], batch size: 116, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:20:44,084 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218804.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:20:58,852 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-01 11:21:35,459 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218838.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:21:56,891 INFO [train.py:904] (2/8) Epoch 22, batch 5700, loss[loss=0.2479, simple_loss=0.3139, pruned_loss=0.09094, over 11439.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2943, pruned_loss=0.06276, over 3052111.27 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:22:16,177 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218865.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:17,594 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:17,849 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7150, 2.7329, 2.7642, 4.5912, 2.5810, 3.0853, 2.7620, 2.9466], device='cuda:2'), covar=tensor([0.1140, 0.2954, 0.2399, 0.0404, 0.3545, 0.2167, 0.2960, 0.2793], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0447, 0.0367, 0.0326, 0.0435, 0.0516, 0.0419, 0.0522], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:22:32,483 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 11:22:48,659 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218886.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:13,875 INFO [optim.py:368] (2/8) 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 11:23:13,891 INFO [train.py:904] (2/8) Epoch 22, batch 5750, loss[loss=0.196, simple_loss=0.2938, pruned_loss=0.04908, over 16935.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2965, pruned_loss=0.06383, over 3037003.44 frames. ], batch size: 96, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:23:25,653 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1150, 2.4580, 2.3723, 2.7115, 2.0423, 3.1541, 1.9322, 2.7619], device='cuda:2'), covar=tensor([0.1158, 0.0545, 0.1098, 0.0176, 0.0119, 0.0403, 0.1447, 0.0699], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0190, 0.0204, 0.0214, 0.0201, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 11:23:29,745 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1294, 2.4712, 2.3782, 2.7787, 2.0585, 3.1777, 1.9273, 2.7805], device='cuda:2'), covar=tensor([0.1204, 0.0543, 0.1155, 0.0185, 0.0128, 0.0396, 0.1477, 0.0714], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0190, 0.0204, 0.0214, 0.0201, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 11:23:32,117 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:57,588 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:24:34,390 INFO [train.py:904] (2/8) Epoch 22, batch 5800, loss[loss=0.2154, simple_loss=0.2888, pruned_loss=0.07103, over 11873.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2967, pruned_loss=0.06358, over 3012989.87 frames. ], batch size: 250, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:24:54,974 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:24,043 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:52,934 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4030, 3.2508, 2.6139, 2.1678, 2.2215, 2.2487, 3.4049, 2.9961], device='cuda:2'), covar=tensor([0.3104, 0.0709, 0.1953, 0.2913, 0.2646, 0.2212, 0.0578, 0.1341], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0269, 0.0305, 0.0314, 0.0297, 0.0259, 0.0296, 0.0337], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 11:25:53,553 INFO [optim.py:368] (2/8) 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 11:25:53,568 INFO [train.py:904] (2/8) Epoch 22, batch 5850, loss[loss=0.1976, simple_loss=0.2858, pruned_loss=0.05468, over 15321.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2947, pruned_loss=0.06214, over 3024058.15 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:26:03,872 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6825, 4.9544, 4.7194, 4.7428, 4.4964, 4.4851, 4.4079, 5.0400], device='cuda:2'), covar=tensor([0.1247, 0.0875, 0.1100, 0.0886, 0.0812, 0.1127, 0.1138, 0.0872], device='cuda:2'), in_proj_covar=tensor([0.0666, 0.0812, 0.0674, 0.0618, 0.0514, 0.0522, 0.0677, 0.0635], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:26:21,526 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:26:22,770 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2852, 5.5589, 5.2946, 5.3435, 5.0966, 4.9722, 4.9407, 5.6644], device='cuda:2'), covar=tensor([0.1166, 0.0758, 0.1090, 0.0928, 0.0771, 0.0804, 0.1220, 0.0862], device='cuda:2'), in_proj_covar=tensor([0.0665, 0.0811, 0.0674, 0.0618, 0.0514, 0.0522, 0.0676, 0.0634], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:26:30,003 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:02,414 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:14,927 INFO [train.py:904] (2/8) Epoch 22, batch 5900, loss[loss=0.2189, simple_loss=0.3006, pruned_loss=0.06858, over 16183.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2944, pruned_loss=0.06224, over 3030770.73 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:27:49,353 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:00,039 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 11:28:04,716 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:16,202 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 11:28:35,885 INFO [optim.py:368] (2/8) 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 11:28:35,900 INFO [train.py:904] (2/8) Epoch 22, batch 5950, loss[loss=0.1996, simple_loss=0.2922, pruned_loss=0.05346, over 16440.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2943, pruned_loss=0.0603, over 3054401.57 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:28:56,925 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 11:29:24,454 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:29:46,162 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 11:29:57,662 INFO [train.py:904] (2/8) Epoch 22, batch 6000, loss[loss=0.1948, simple_loss=0.284, pruned_loss=0.05282, over 16735.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2935, pruned_loss=0.05935, over 3086501.07 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:29:57,662 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 11:30:07,609 INFO [train.py:938] (2/8) Epoch 22, validation: loss=0.1507, simple_loss=0.2632, pruned_loss=0.01907, over 944034.00 frames. 2023-05-01 11:30:07,609 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 11:30:27,709 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219165.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:08,397 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-01 11:31:28,922 INFO [optim.py:368] (2/8) 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 11:31:28,937 INFO [train.py:904] (2/8) Epoch 22, batch 6050, loss[loss=0.1991, simple_loss=0.2919, pruned_loss=0.05309, over 16429.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2924, pruned_loss=0.05911, over 3096335.23 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:31:29,586 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:44,911 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:11,118 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219230.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:46,221 INFO [train.py:904] (2/8) Epoch 22, batch 6100, loss[loss=0.2398, simple_loss=0.3065, pruned_loss=0.08659, over 11510.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2916, pruned_loss=0.05805, over 3099184.84 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:33:05,449 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219264.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:25,445 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8673, 2.1838, 2.4599, 3.1175, 2.2389, 2.3568, 2.3534, 2.2925], device='cuda:2'), covar=tensor([0.1426, 0.3044, 0.2380, 0.0786, 0.3887, 0.2437, 0.3047, 0.3090], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0448, 0.0367, 0.0326, 0.0435, 0.0515, 0.0419, 0.0522], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:33:26,347 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:32,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7360, 1.8036, 1.6353, 1.4611, 1.9381, 1.6164, 1.6514, 1.9352], device='cuda:2'), covar=tensor([0.0195, 0.0280, 0.0383, 0.0346, 0.0215, 0.0275, 0.0183, 0.0201], device='cuda:2'), in_proj_covar=tensor([0.0208, 0.0230, 0.0223, 0.0223, 0.0232, 0.0231, 0.0232, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:34:04,107 INFO [train.py:904] (2/8) Epoch 22, batch 6150, loss[loss=0.1604, simple_loss=0.2539, pruned_loss=0.03342, over 17263.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2892, pruned_loss=0.05717, over 3114408.62 frames. ], batch size: 52, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:34:05,864 INFO [optim.py:368] (2/8) 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 11:34:33,693 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:34:41,759 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5579, 2.2574, 1.8348, 2.0257, 2.5694, 2.2146, 2.3946, 2.7258], device='cuda:2'), covar=tensor([0.0197, 0.0432, 0.0563, 0.0484, 0.0262, 0.0411, 0.0198, 0.0258], device='cuda:2'), in_proj_covar=tensor([0.0208, 0.0230, 0.0223, 0.0223, 0.0232, 0.0231, 0.0232, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:35:03,740 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:06,793 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219342.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:35:22,806 INFO [train.py:904] (2/8) Epoch 22, batch 6200, loss[loss=0.2223, simple_loss=0.2887, pruned_loss=0.07795, over 11325.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2869, pruned_loss=0.05621, over 3119278.16 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:02,464 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:36:42,473 INFO [train.py:904] (2/8) Epoch 22, batch 6250, loss[loss=0.1823, simple_loss=0.2864, pruned_loss=0.03909, over 16830.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2863, pruned_loss=0.05528, over 3148469.57 frames. ], batch size: 102, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:43,105 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219403.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:36:43,733 INFO [optim.py:368] (2/8) 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 11:36:48,735 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4068, 3.6239, 2.8196, 2.1456, 2.3587, 2.3906, 3.8408, 3.2471], device='cuda:2'), covar=tensor([0.3201, 0.0654, 0.1812, 0.3002, 0.2671, 0.2065, 0.0434, 0.1300], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0269, 0.0305, 0.0314, 0.0298, 0.0259, 0.0296, 0.0337], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 11:36:53,049 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 11:37:20,347 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:37:31,718 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.93 vs. limit=5.0 2023-05-01 11:37:57,562 INFO [train.py:904] (2/8) Epoch 22, batch 6300, loss[loss=0.1999, simple_loss=0.285, pruned_loss=0.05739, over 15208.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2861, pruned_loss=0.05517, over 3141527.95 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:37:58,167 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8351, 3.8680, 4.1196, 4.0917, 4.1201, 3.8788, 3.8883, 3.9029], device='cuda:2'), covar=tensor([0.0383, 0.0686, 0.0445, 0.0470, 0.0449, 0.0482, 0.0888, 0.0517], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0452, 0.0437, 0.0405, 0.0485, 0.0459, 0.0545, 0.0368], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 11:38:47,124 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0199, 2.1193, 2.2352, 3.5664, 2.0910, 2.4556, 2.2299, 2.2670], device='cuda:2'), covar=tensor([0.1385, 0.3469, 0.2817, 0.0598, 0.4055, 0.2444, 0.3413, 0.3071], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0448, 0.0367, 0.0326, 0.0436, 0.0516, 0.0420, 0.0522], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:38:55,966 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219490.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:39:15,298 INFO [train.py:904] (2/8) Epoch 22, batch 6350, loss[loss=0.2462, simple_loss=0.3119, pruned_loss=0.09027, over 11480.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2875, pruned_loss=0.05669, over 3113249.70 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:39:16,413 INFO [optim.py:368] (2/8) 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 11:40:28,605 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:40:31,084 INFO [train.py:904] (2/8) Epoch 22, batch 6400, loss[loss=0.181, simple_loss=0.2661, pruned_loss=0.04795, over 16887.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2879, pruned_loss=0.05801, over 3108225.68 frames. ], batch size: 116, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:40:40,475 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219559.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:41:45,410 INFO [train.py:904] (2/8) Epoch 22, batch 6450, loss[loss=0.1892, simple_loss=0.2768, pruned_loss=0.05078, over 15270.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.288, pruned_loss=0.05782, over 3091933.76 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:41:47,189 INFO [optim.py:368] (2/8) 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 11:41:51,311 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-01 11:42:13,384 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:42:13,428 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6307, 2.5263, 1.9448, 2.6532, 2.1651, 2.7621, 2.1868, 2.4235], device='cuda:2'), covar=tensor([0.0309, 0.0495, 0.1199, 0.0314, 0.0615, 0.0697, 0.1166, 0.0546], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0161, 0.0176, 0.0216, 0.0200, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 11:42:45,451 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:43:04,572 INFO [train.py:904] (2/8) Epoch 22, batch 6500, loss[loss=0.213, simple_loss=0.2886, pruned_loss=0.06868, over 11704.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2862, pruned_loss=0.05737, over 3087937.29 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:43:17,315 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1038, 4.4582, 3.2301, 2.7266, 3.1817, 2.8184, 4.9695, 3.8828], device='cuda:2'), covar=tensor([0.2592, 0.0588, 0.1700, 0.2386, 0.2448, 0.1842, 0.0340, 0.1149], device='cuda:2'), in_proj_covar=tensor([0.0323, 0.0267, 0.0302, 0.0312, 0.0295, 0.0257, 0.0293, 0.0334], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 11:43:24,439 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7560, 3.7416, 3.8987, 3.7018, 3.8515, 4.2508, 3.9083, 3.5952], device='cuda:2'), covar=tensor([0.2417, 0.2693, 0.2895, 0.2488, 0.2731, 0.1906, 0.1762, 0.2890], device='cuda:2'), in_proj_covar=tensor([0.0412, 0.0592, 0.0654, 0.0495, 0.0658, 0.0682, 0.0515, 0.0664], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 11:43:24,545 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5285, 3.5882, 3.3211, 2.9819, 3.1800, 3.4849, 3.3014, 3.3023], device='cuda:2'), covar=tensor([0.0617, 0.0737, 0.0285, 0.0280, 0.0513, 0.0474, 0.1416, 0.0515], device='cuda:2'), in_proj_covar=tensor([0.0291, 0.0429, 0.0341, 0.0338, 0.0347, 0.0393, 0.0235, 0.0407], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:43:30,546 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:43:42,826 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:00,772 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:19,767 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219698.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:44:26,639 INFO [train.py:904] (2/8) Epoch 22, batch 6550, loss[loss=0.2566, simple_loss=0.3244, pruned_loss=0.09439, over 11916.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2885, pruned_loss=0.0582, over 3088488.42 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:44:28,429 INFO [optim.py:368] (2/8) 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 11:45:03,111 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:08,245 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:46,181 INFO [train.py:904] (2/8) Epoch 22, batch 6600, loss[loss=0.2258, simple_loss=0.3005, pruned_loss=0.07554, over 11666.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2909, pruned_loss=0.05891, over 3076654.25 frames. ], batch size: 246, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:46:02,062 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219763.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:23,297 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:33,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1564, 1.5611, 1.9545, 2.1260, 2.2366, 2.4257, 1.7462, 2.2848], device='cuda:2'), covar=tensor([0.0223, 0.0485, 0.0267, 0.0339, 0.0304, 0.0198, 0.0496, 0.0143], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0193, 0.0179, 0.0183, 0.0197, 0.0153, 0.0196, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:47:02,885 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0956, 2.3704, 2.5718, 1.9634, 2.6689, 2.7590, 2.3870, 2.3256], device='cuda:2'), covar=tensor([0.0739, 0.0257, 0.0250, 0.0940, 0.0118, 0.0278, 0.0481, 0.0482], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0110, 0.0100, 0.0141, 0.0082, 0.0128, 0.0131, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 11:47:06,507 INFO [train.py:904] (2/8) Epoch 22, batch 6650, loss[loss=0.2535, simple_loss=0.3159, pruned_loss=0.09551, over 11800.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2912, pruned_loss=0.05957, over 3083755.33 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:47:07,640 INFO [optim.py:368] (2/8) 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 11:47:28,411 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 11:47:39,798 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219824.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:06,520 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6984, 4.7120, 5.0312, 5.0178, 5.0450, 4.7546, 4.7533, 4.5061], device='cuda:2'), covar=tensor([0.0307, 0.0521, 0.0359, 0.0376, 0.0446, 0.0377, 0.0832, 0.0501], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0452, 0.0437, 0.0406, 0.0486, 0.0460, 0.0544, 0.0368], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 11:48:10,734 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:21,241 INFO [train.py:904] (2/8) Epoch 22, batch 6700, loss[loss=0.2076, simple_loss=0.2865, pruned_loss=0.0643, over 16683.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2898, pruned_loss=0.05947, over 3094617.37 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:48:23,545 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2045, 4.0909, 4.2719, 4.4114, 4.5428, 4.1267, 4.4700, 4.5466], device='cuda:2'), covar=tensor([0.1829, 0.1266, 0.1486, 0.0721, 0.0559, 0.1238, 0.0837, 0.0704], device='cuda:2'), in_proj_covar=tensor([0.0628, 0.0780, 0.0903, 0.0786, 0.0598, 0.0622, 0.0650, 0.0751], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:48:27,549 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:30,112 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:44,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9619, 3.7740, 3.8872, 4.1709, 4.2425, 3.9446, 4.2656, 4.2841], device='cuda:2'), covar=tensor([0.1851, 0.1512, 0.2065, 0.0935, 0.0854, 0.1659, 0.1033, 0.0977], device='cuda:2'), in_proj_covar=tensor([0.0630, 0.0782, 0.0905, 0.0788, 0.0599, 0.0624, 0.0651, 0.0752], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:49:36,042 INFO [train.py:904] (2/8) Epoch 22, batch 6750, loss[loss=0.2243, simple_loss=0.3024, pruned_loss=0.07312, over 15349.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2891, pruned_loss=0.05978, over 3088917.91 frames. ], batch size: 191, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:49:37,873 INFO [optim.py:368] (2/8) 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 11:49:43,850 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:00,672 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:54,124 INFO [train.py:904] (2/8) Epoch 22, batch 6800, loss[loss=0.243, simple_loss=0.3154, pruned_loss=0.0853, over 12007.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2891, pruned_loss=0.05913, over 3107925.99 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:51:28,872 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2964, 2.3570, 2.4994, 4.0374, 2.3160, 2.6621, 2.4187, 2.4781], device='cuda:2'), covar=tensor([0.1376, 0.3312, 0.2693, 0.0531, 0.3897, 0.2369, 0.3287, 0.3112], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0446, 0.0364, 0.0324, 0.0435, 0.0513, 0.0417, 0.0520], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:52:04,946 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219998.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:52:15,003 INFO [train.py:904] (2/8) Epoch 22, batch 6850, loss[loss=0.2036, simple_loss=0.2999, pruned_loss=0.05366, over 16249.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2903, pruned_loss=0.05921, over 3119834.00 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:52:16,799 INFO [optim.py:368] (2/8) 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 11:52:37,039 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 11:53:20,649 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220046.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:53:31,114 INFO [train.py:904] (2/8) Epoch 22, batch 6900, loss[loss=0.2281, simple_loss=0.3084, pruned_loss=0.07388, over 16627.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2927, pruned_loss=0.05886, over 3130206.05 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:53:37,171 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 11:54:24,086 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220087.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:41,868 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220098.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:49,991 INFO [train.py:904] (2/8) Epoch 22, batch 6950, loss[loss=0.1887, simple_loss=0.2751, pruned_loss=0.05114, over 16382.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2947, pruned_loss=0.06033, over 3117641.72 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:54:51,079 INFO [optim.py:368] (2/8) 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 11:55:14,155 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220119.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:46,949 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4234, 5.3338, 5.1684, 4.3103, 5.2499, 1.7058, 4.9381, 4.9105], device='cuda:2'), covar=tensor([0.0109, 0.0115, 0.0203, 0.0484, 0.0115, 0.3007, 0.0212, 0.0220], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0156, 0.0198, 0.0179, 0.0175, 0.0208, 0.0187, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:55:54,598 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:56:01,188 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-01 11:56:03,773 INFO [train.py:904] (2/8) Epoch 22, batch 7000, loss[loss=0.1977, simple_loss=0.2927, pruned_loss=0.05139, over 17011.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2946, pruned_loss=0.06001, over 3097364.47 frames. ], batch size: 41, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 11:56:14,318 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:05,650 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220194.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:18,841 INFO [train.py:904] (2/8) Epoch 22, batch 7050, loss[loss=0.2706, simple_loss=0.3243, pruned_loss=0.1084, over 11440.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2952, pruned_loss=0.05973, over 3108299.67 frames. ], batch size: 247, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:57:21,954 INFO [optim.py:368] (2/8) 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 11:57:35,244 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:59,212 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:58:34,611 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1461, 2.4234, 1.9811, 2.2149, 2.8262, 2.4151, 2.7790, 2.9980], device='cuda:2'), covar=tensor([0.0165, 0.0413, 0.0590, 0.0475, 0.0248, 0.0402, 0.0238, 0.0242], device='cuda:2'), in_proj_covar=tensor([0.0207, 0.0231, 0.0223, 0.0223, 0.0232, 0.0230, 0.0231, 0.0227], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 11:58:36,504 INFO [train.py:904] (2/8) Epoch 22, batch 7100, loss[loss=0.1827, simple_loss=0.276, pruned_loss=0.04467, over 16903.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2935, pruned_loss=0.05942, over 3089963.44 frames. ], batch size: 96, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:58:45,418 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:35,774 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:54,481 INFO [train.py:904] (2/8) Epoch 22, batch 7150, loss[loss=0.2092, simple_loss=0.3002, pruned_loss=0.05907, over 16449.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2914, pruned_loss=0.05914, over 3093259.16 frames. ], batch size: 68, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:59:58,136 INFO [optim.py:368] (2/8) 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 12:00:20,902 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:01:04,964 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 12:01:07,949 INFO [train.py:904] (2/8) Epoch 22, batch 7200, loss[loss=0.178, simple_loss=0.268, pruned_loss=0.04407, over 16715.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2886, pruned_loss=0.0574, over 3082713.55 frames. ], batch size: 134, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:28,580 INFO [train.py:904] (2/8) Epoch 22, batch 7250, loss[loss=0.1851, simple_loss=0.267, pruned_loss=0.05157, over 12217.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2859, pruned_loss=0.05593, over 3095389.13 frames. ], batch size: 247, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:30,891 INFO [optim.py:368] (2/8) 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 12:02:47,333 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 12:02:53,054 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220419.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:27,827 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6371, 2.5272, 1.9638, 2.6700, 2.1171, 2.7798, 2.1599, 2.3585], device='cuda:2'), covar=tensor([0.0318, 0.0392, 0.1229, 0.0245, 0.0620, 0.0487, 0.1192, 0.0602], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0174, 0.0192, 0.0160, 0.0174, 0.0214, 0.0200, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:03:30,263 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220443.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:45,048 INFO [train.py:904] (2/8) Epoch 22, batch 7300, loss[loss=0.2593, simple_loss=0.3215, pruned_loss=0.09858, over 11638.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.286, pruned_loss=0.05628, over 3087498.42 frames. ], batch size: 246, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:03:46,868 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220454.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:04:07,576 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:04:44,043 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0548, 2.2556, 2.3142, 2.5203, 1.9978, 3.0933, 1.9057, 2.6888], device='cuda:2'), covar=tensor([0.1133, 0.0712, 0.1058, 0.0177, 0.0157, 0.0353, 0.1364, 0.0703], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0196, 0.0189, 0.0207, 0.0215, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:05:02,408 INFO [train.py:904] (2/8) Epoch 22, batch 7350, loss[loss=0.1994, simple_loss=0.2896, pruned_loss=0.05458, over 16462.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2873, pruned_loss=0.05747, over 3064413.16 frames. ], batch size: 75, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:05:05,571 INFO [optim.py:368] (2/8) 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 12:05:06,083 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0837, 4.0534, 3.9310, 3.1337, 3.9947, 1.7267, 3.7548, 3.4797], device='cuda:2'), covar=tensor([0.0126, 0.0106, 0.0209, 0.0332, 0.0102, 0.3002, 0.0138, 0.0283], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0155, 0.0197, 0.0177, 0.0174, 0.0207, 0.0186, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:05:17,921 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220513.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:20,489 INFO [train.py:904] (2/8) Epoch 22, batch 7400, loss[loss=0.2214, simple_loss=0.3038, pruned_loss=0.06952, over 15309.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2884, pruned_loss=0.05848, over 3047167.07 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:06:25,433 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:35,403 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:41,011 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 12:06:58,962 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6395, 2.8541, 2.5635, 4.9291, 3.5126, 4.1001, 1.6324, 2.9402], device='cuda:2'), covar=tensor([0.1497, 0.0849, 0.1311, 0.0146, 0.0402, 0.0446, 0.1789, 0.0912], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0176, 0.0197, 0.0191, 0.0208, 0.0217, 0.0204, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:07:02,588 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9103, 2.7906, 2.7310, 2.1391, 2.6200, 2.1376, 2.7456, 2.9931], device='cuda:2'), covar=tensor([0.0305, 0.0846, 0.0605, 0.1778, 0.0914, 0.1000, 0.0622, 0.0695], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0165, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:07:13,078 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:07:38,735 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7509, 3.8723, 2.4082, 4.5235, 2.8961, 4.3777, 2.4697, 3.0108], device='cuda:2'), covar=tensor([0.0291, 0.0374, 0.1620, 0.0215, 0.0861, 0.0571, 0.1597, 0.0850], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0160, 0.0174, 0.0215, 0.0201, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:07:41,486 INFO [train.py:904] (2/8) Epoch 22, batch 7450, loss[loss=0.223, simple_loss=0.3218, pruned_loss=0.06207, over 16526.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2899, pruned_loss=0.05949, over 3055482.43 frames. ], batch size: 75, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:07:43,937 INFO [optim.py:368] (2/8) 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 12:08:01,234 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:08:04,717 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220616.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 12:09:00,779 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 22, batch 7500, loss[loss=0.1896, simple_loss=0.2715, pruned_loss=0.05386, over 16957.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2901, pruned_loss=0.05869, over 3053508.04 frames. ], batch size: 109, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:21,150 INFO [train.py:904] (2/8) Epoch 22, batch 7550, loss[loss=0.244, simple_loss=0.3089, pruned_loss=0.08952, over 11824.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2892, pruned_loss=0.05887, over 3067698.23 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:24,492 INFO [optim.py:368] (2/8) 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 12:10:35,520 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:10:37,491 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-01 12:11:10,626 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4291, 1.7472, 2.1342, 2.4319, 2.5036, 2.7639, 1.9679, 2.6604], device='cuda:2'), covar=tensor([0.0212, 0.0490, 0.0305, 0.0314, 0.0314, 0.0182, 0.0475, 0.0145], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0189, 0.0176, 0.0180, 0.0193, 0.0150, 0.0193, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:11:23,178 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:11:38,533 INFO [train.py:904] (2/8) Epoch 22, batch 7600, loss[loss=0.1775, simple_loss=0.2695, pruned_loss=0.04277, over 17116.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2875, pruned_loss=0.05855, over 3074926.62 frames. ], batch size: 49, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:11:40,683 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:36,980 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:50,670 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3886, 3.3488, 3.4146, 3.4923, 3.5252, 3.2617, 3.4788, 3.5768], device='cuda:2'), covar=tensor([0.1299, 0.1028, 0.1043, 0.0645, 0.0684, 0.2631, 0.1247, 0.0829], device='cuda:2'), in_proj_covar=tensor([0.0623, 0.0769, 0.0890, 0.0779, 0.0593, 0.0617, 0.0645, 0.0742], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:12:55,415 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220802.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:56,365 INFO [train.py:904] (2/8) Epoch 22, batch 7650, loss[loss=0.2028, simple_loss=0.2861, pruned_loss=0.05977, over 16365.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2887, pruned_loss=0.05942, over 3068162.51 frames. ], batch size: 35, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:12:59,168 INFO [optim.py:368] (2/8) 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 12:13:06,087 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5886, 2.7128, 2.6680, 4.4951, 2.5726, 2.9884, 2.7272, 2.8562], device='cuda:2'), covar=tensor([0.1254, 0.3015, 0.2680, 0.0434, 0.3629, 0.2198, 0.3180, 0.2863], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0446, 0.0364, 0.0324, 0.0434, 0.0513, 0.0417, 0.0520], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:13:20,340 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4238, 1.4822, 2.0410, 2.2760, 2.3667, 2.5911, 1.5908, 2.5647], device='cuda:2'), covar=tensor([0.0208, 0.0599, 0.0309, 0.0320, 0.0290, 0.0197, 0.0696, 0.0143], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0189, 0.0176, 0.0180, 0.0193, 0.0150, 0.0193, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:13:27,053 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:14:11,876 INFO [train.py:904] (2/8) Epoch 22, batch 7700, loss[loss=0.1864, simple_loss=0.2764, pruned_loss=0.04816, over 16753.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.289, pruned_loss=0.05974, over 3074013.16 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:14:18,199 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1175, 2.0198, 2.5990, 2.9411, 2.8291, 3.6010, 2.2771, 3.4538], device='cuda:2'), covar=tensor([0.0224, 0.0551, 0.0335, 0.0360, 0.0340, 0.0132, 0.0530, 0.0161], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0190, 0.0177, 0.0181, 0.0194, 0.0151, 0.0194, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:14:47,081 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 12:14:58,245 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:01,455 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220885.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:13,433 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-05-01 12:15:29,148 INFO [train.py:904] (2/8) Epoch 22, batch 7750, loss[loss=0.1809, simple_loss=0.2708, pruned_loss=0.04553, over 17164.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2889, pruned_loss=0.05943, over 3079972.78 frames. ], batch size: 46, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:15:30,840 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220904.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:32,046 INFO [optim.py:368] (2/8) 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 12:15:40,070 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220911.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 12:15:45,238 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:15,123 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:42,219 INFO [train.py:904] (2/8) Epoch 22, batch 7800, loss[loss=0.2071, simple_loss=0.3007, pruned_loss=0.0568, over 16556.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2892, pruned_loss=0.05919, over 3094105.62 frames. ], batch size: 75, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:16:56,567 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:00,228 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:55,724 INFO [train.py:904] (2/8) Epoch 22, batch 7850, loss[loss=0.215, simple_loss=0.2912, pruned_loss=0.06946, over 11750.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2903, pruned_loss=0.0588, over 3095113.55 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:17:58,013 INFO [optim.py:368] (2/8) 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 12:18:02,915 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:18:09,865 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5762, 3.6537, 2.2746, 4.1250, 2.7894, 4.0925, 2.3523, 2.8996], device='cuda:2'), covar=tensor([0.0272, 0.0369, 0.1649, 0.0168, 0.0789, 0.0471, 0.1495, 0.0771], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0161, 0.0175, 0.0215, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:18:16,488 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3973, 4.6370, 4.9424, 4.8765, 4.9004, 4.5998, 4.2908, 4.3876], device='cuda:2'), covar=tensor([0.0588, 0.0676, 0.0532, 0.0692, 0.0754, 0.0626, 0.1683, 0.0665], device='cuda:2'), in_proj_covar=tensor([0.0410, 0.0456, 0.0440, 0.0409, 0.0488, 0.0465, 0.0549, 0.0371], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 12:18:24,406 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7507, 4.6340, 4.8197, 4.9629, 5.1720, 4.5941, 5.1480, 5.1409], device='cuda:2'), covar=tensor([0.1891, 0.1191, 0.1479, 0.0705, 0.0494, 0.0991, 0.0569, 0.0633], device='cuda:2'), in_proj_covar=tensor([0.0627, 0.0774, 0.0897, 0.0781, 0.0595, 0.0622, 0.0648, 0.0748], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:18:50,923 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-01 12:18:53,617 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4117, 3.4762, 2.1242, 3.8340, 2.6154, 3.8231, 2.1462, 2.7685], device='cuda:2'), covar=tensor([0.0280, 0.0390, 0.1671, 0.0236, 0.0907, 0.0674, 0.1629, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0175, 0.0193, 0.0161, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:19:09,447 INFO [train.py:904] (2/8) Epoch 22, batch 7900, loss[loss=0.1957, simple_loss=0.2848, pruned_loss=0.05331, over 16426.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2891, pruned_loss=0.05785, over 3102399.42 frames. ], batch size: 35, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:19:16,614 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 12:20:27,100 INFO [train.py:904] (2/8) Epoch 22, batch 7950, loss[loss=0.2082, simple_loss=0.2909, pruned_loss=0.06271, over 15458.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2897, pruned_loss=0.05831, over 3111393.44 frames. ], batch size: 191, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:20:32,011 INFO [optim.py:368] (2/8) 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 12:21:30,239 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4644, 2.4951, 2.5012, 4.2090, 2.2084, 2.7466, 2.5342, 2.6687], device='cuda:2'), covar=tensor([0.1251, 0.3414, 0.2658, 0.0492, 0.4125, 0.2476, 0.3257, 0.3066], device='cuda:2'), in_proj_covar=tensor([0.0398, 0.0446, 0.0363, 0.0325, 0.0436, 0.0514, 0.0418, 0.0519], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:21:41,830 INFO [train.py:904] (2/8) Epoch 22, batch 8000, loss[loss=0.1816, simple_loss=0.2759, pruned_loss=0.04364, over 16737.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2904, pruned_loss=0.05903, over 3110804.64 frames. ], batch size: 89, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:22:18,531 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:22:55,165 INFO [train.py:904] (2/8) Epoch 22, batch 8050, loss[loss=0.199, simple_loss=0.2828, pruned_loss=0.0576, over 16852.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2907, pruned_loss=0.05889, over 3121775.83 frames. ], batch size: 116, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:23:01,252 INFO [optim.py:368] (2/8) 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 12:23:07,226 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:23:26,662 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1320, 3.3547, 3.6032, 2.1639, 3.0556, 2.3945, 3.5117, 3.7470], device='cuda:2'), covar=tensor([0.0284, 0.0922, 0.0610, 0.2165, 0.0894, 0.1010, 0.0669, 0.1019], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0152, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:24:08,783 INFO [train.py:904] (2/8) Epoch 22, batch 8100, loss[loss=0.2304, simple_loss=0.3068, pruned_loss=0.07698, over 11256.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2903, pruned_loss=0.05868, over 3112512.22 frames. ], batch size: 250, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:24:17,302 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:24:19,325 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221260.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:25:24,333 INFO [train.py:904] (2/8) Epoch 22, batch 8150, loss[loss=0.2138, simple_loss=0.2898, pruned_loss=0.06891, over 11856.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2874, pruned_loss=0.05718, over 3123141.61 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:25:31,011 INFO [optim.py:368] (2/8) 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 12:25:31,357 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221307.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:26:30,196 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3132, 1.9252, 2.7317, 3.1163, 2.9717, 3.6463, 1.8776, 3.6145], device='cuda:2'), covar=tensor([0.0167, 0.0594, 0.0292, 0.0280, 0.0270, 0.0135, 0.0724, 0.0113], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0190, 0.0176, 0.0181, 0.0194, 0.0151, 0.0193, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:26:41,083 INFO [train.py:904] (2/8) Epoch 22, batch 8200, loss[loss=0.2118, simple_loss=0.2815, pruned_loss=0.07107, over 11427.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.286, pruned_loss=0.0575, over 3100119.77 frames. ], batch size: 247, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:26:44,115 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221355.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:27:58,675 INFO [train.py:904] (2/8) Epoch 22, batch 8250, loss[loss=0.1624, simple_loss=0.2674, pruned_loss=0.02872, over 16816.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2846, pruned_loss=0.05493, over 3094236.98 frames. ], batch size: 96, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:28:05,605 INFO [optim.py:368] (2/8) 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 12:29:17,814 INFO [train.py:904] (2/8) Epoch 22, batch 8300, loss[loss=0.1994, simple_loss=0.2942, pruned_loss=0.05233, over 16293.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2825, pruned_loss=0.0525, over 3073816.37 frames. ], batch size: 146, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:29:22,204 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6601, 2.7660, 2.4411, 3.8633, 2.4198, 3.9803, 1.4690, 2.9035], device='cuda:2'), covar=tensor([0.1458, 0.0731, 0.1231, 0.0176, 0.0141, 0.0393, 0.1815, 0.0742], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0189, 0.0207, 0.0215, 0.0202, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:29:30,057 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 12:29:57,605 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:30:33,850 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0365, 3.9904, 3.9173, 3.0640, 3.9272, 1.8005, 3.7594, 3.5750], device='cuda:2'), covar=tensor([0.0130, 0.0144, 0.0201, 0.0327, 0.0125, 0.2912, 0.0150, 0.0276], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0153, 0.0196, 0.0175, 0.0172, 0.0205, 0.0184, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:30:38,299 INFO [train.py:904] (2/8) Epoch 22, batch 8350, loss[loss=0.2112, simple_loss=0.3044, pruned_loss=0.05905, over 16928.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2822, pruned_loss=0.05081, over 3088317.87 frames. ], batch size: 116, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:30:43,698 INFO [optim.py:368] (2/8) 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 12:30:44,197 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:15,661 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221526.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:55,736 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 12:31:56,139 INFO [train.py:904] (2/8) Epoch 22, batch 8400, loss[loss=0.1699, simple_loss=0.265, pruned_loss=0.03738, over 15112.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2793, pruned_loss=0.04853, over 3070971.00 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:32:08,954 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:32:09,102 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9261, 2.0714, 2.5376, 2.9308, 2.7790, 3.3645, 2.4077, 3.3507], device='cuda:2'), covar=tensor([0.0209, 0.0517, 0.0359, 0.0288, 0.0300, 0.0197, 0.0439, 0.0140], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0190, 0.0176, 0.0181, 0.0193, 0.0150, 0.0193, 0.0148], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:32:21,002 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:33:17,378 INFO [train.py:904] (2/8) Epoch 22, batch 8450, loss[loss=0.1616, simple_loss=0.248, pruned_loss=0.03754, over 11842.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2773, pruned_loss=0.04669, over 3077158.84 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:33:24,323 INFO [optim.py:368] (2/8) 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 12:33:25,020 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4936, 3.3656, 2.7557, 2.2174, 2.1502, 2.2926, 3.4809, 3.0279], device='cuda:2'), covar=tensor([0.2939, 0.0645, 0.1774, 0.3013, 0.2960, 0.2356, 0.0449, 0.1382], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0264, 0.0300, 0.0310, 0.0293, 0.0256, 0.0291, 0.0330], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 12:33:26,150 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221608.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:34:38,831 INFO [train.py:904] (2/8) Epoch 22, batch 8500, loss[loss=0.1558, simple_loss=0.2414, pruned_loss=0.03512, over 12059.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2736, pruned_loss=0.0447, over 3043981.45 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:02,531 INFO [train.py:904] (2/8) Epoch 22, batch 8550, loss[loss=0.1908, simple_loss=0.288, pruned_loss=0.0468, over 16934.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2714, pruned_loss=0.04375, over 3044437.13 frames. ], batch size: 109, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:10,060 INFO [optim.py:368] (2/8) 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 12:37:08,422 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5218, 2.5106, 1.9944, 2.0344, 2.8302, 2.5461, 3.1046, 3.2101], device='cuda:2'), covar=tensor([0.0162, 0.0561, 0.0749, 0.0665, 0.0347, 0.0484, 0.0258, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0228, 0.0221, 0.0220, 0.0229, 0.0226, 0.0227, 0.0221], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:37:11,072 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-01 12:37:41,420 INFO [train.py:904] (2/8) Epoch 22, batch 8600, loss[loss=0.1827, simple_loss=0.2745, pruned_loss=0.0455, over 16882.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2711, pruned_loss=0.04254, over 3046539.59 frames. ], batch size: 124, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:37:53,153 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:38:49,511 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9364, 2.8206, 2.6881, 1.9891, 2.5700, 2.8169, 2.6651, 2.0032], device='cuda:2'), covar=tensor([0.0444, 0.0071, 0.0071, 0.0355, 0.0120, 0.0095, 0.0095, 0.0411], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0132, 0.0096, 0.0108, 0.0094, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 12:39:21,614 INFO [train.py:904] (2/8) Epoch 22, batch 8650, loss[loss=0.1693, simple_loss=0.2662, pruned_loss=0.03624, over 16477.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2693, pruned_loss=0.04128, over 3038405.95 frames. ], batch size: 147, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:39:30,560 INFO [optim.py:368] (2/8) 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 12:39:58,372 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:41:06,645 INFO [train.py:904] (2/8) Epoch 22, batch 8700, loss[loss=0.1753, simple_loss=0.2576, pruned_loss=0.0465, over 11850.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2671, pruned_loss=0.04028, over 3036590.09 frames. ], batch size: 250, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:41:27,670 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221863.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:42:16,263 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:42:16,391 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7603, 3.1293, 3.4076, 1.9337, 2.7904, 2.1638, 3.2710, 3.3332], device='cuda:2'), covar=tensor([0.0304, 0.0932, 0.0547, 0.2251, 0.0938, 0.1092, 0.0718, 0.1039], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0160, 0.0163, 0.0151, 0.0142, 0.0127, 0.0140, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:42:42,866 INFO [train.py:904] (2/8) Epoch 22, batch 8750, loss[loss=0.1666, simple_loss=0.2721, pruned_loss=0.03052, over 16692.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.267, pruned_loss=0.03974, over 3052785.89 frames. ], batch size: 89, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:42:53,174 INFO [optim.py:368] (2/8) 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 12:44:31,205 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:44:34,618 INFO [train.py:904] (2/8) Epoch 22, batch 8800, loss[loss=0.1568, simple_loss=0.2587, pruned_loss=0.02749, over 16925.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2655, pruned_loss=0.03878, over 3048628.85 frames. ], batch size: 96, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:46:22,303 INFO [train.py:904] (2/8) Epoch 22, batch 8850, loss[loss=0.1628, simple_loss=0.2747, pruned_loss=0.02542, over 16667.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2671, pruned_loss=0.0381, over 3027770.23 frames. ], batch size: 89, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:46:28,904 INFO [optim.py:368] (2/8) 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 12:46:35,341 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:46:40,164 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9296, 2.0797, 2.1682, 3.4461, 2.0130, 2.2900, 2.1956, 2.1504], device='cuda:2'), covar=tensor([0.1395, 0.3761, 0.3096, 0.0633, 0.4635, 0.2825, 0.3656, 0.3718], device='cuda:2'), in_proj_covar=tensor([0.0393, 0.0440, 0.0362, 0.0319, 0.0431, 0.0507, 0.0413, 0.0514], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:48:07,874 INFO [train.py:904] (2/8) Epoch 22, batch 8900, loss[loss=0.1739, simple_loss=0.2701, pruned_loss=0.0388, over 16864.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2676, pruned_loss=0.03772, over 3024808.76 frames. ], batch size: 124, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:48:21,075 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-01 12:48:42,833 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222070.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:49:47,019 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8732, 4.8207, 4.6414, 4.0830, 4.7628, 1.7825, 4.4452, 4.5897], device='cuda:2'), covar=tensor([0.0097, 0.0087, 0.0236, 0.0378, 0.0112, 0.2655, 0.0157, 0.0216], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0152, 0.0193, 0.0172, 0.0171, 0.0204, 0.0182, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:50:11,677 INFO [train.py:904] (2/8) Epoch 22, batch 8950, loss[loss=0.1636, simple_loss=0.2648, pruned_loss=0.03115, over 16690.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2669, pruned_loss=0.03803, over 3025528.25 frames. ], batch size: 134, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:50:23,608 INFO [optim.py:368] (2/8) 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 12:50:37,681 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:50:49,229 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-01 12:50:51,343 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3558, 3.4295, 2.0722, 3.8548, 2.5143, 3.7557, 2.1172, 2.7180], device='cuda:2'), covar=tensor([0.0336, 0.0399, 0.1713, 0.0178, 0.0922, 0.0528, 0.1741, 0.0821], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0171, 0.0188, 0.0155, 0.0171, 0.0208, 0.0198, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 12:51:44,914 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9687, 2.2287, 2.2836, 2.8617, 1.6383, 3.2378, 1.7759, 2.6944], device='cuda:2'), covar=tensor([0.1283, 0.0833, 0.1168, 0.0169, 0.0075, 0.0365, 0.1663, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0185, 0.0202, 0.0212, 0.0201, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:52:03,798 INFO [train.py:904] (2/8) Epoch 22, batch 9000, loss[loss=0.1614, simple_loss=0.2521, pruned_loss=0.03532, over 15297.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2639, pruned_loss=0.03709, over 3007273.39 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:52:03,798 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 12:52:11,550 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0205, 4.0532, 3.8431, 5.7626, 4.8034, 5.2235, 2.9462, 4.3699], device='cuda:2'), covar=tensor([0.1008, 0.0529, 0.0845, 0.0062, 0.0131, 0.0231, 0.1217, 0.0434], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0184, 0.0202, 0.0212, 0.0201, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 12:52:14,704 INFO [train.py:938] (2/8) Epoch 22, validation: loss=0.1453, simple_loss=0.2492, pruned_loss=0.0207, over 944034.00 frames. 2023-05-01 12:52:14,705 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 12:52:34,498 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6274, 4.5712, 4.4352, 3.8169, 4.4626, 1.6023, 4.1722, 4.3001], device='cuda:2'), covar=tensor([0.0089, 0.0086, 0.0183, 0.0342, 0.0117, 0.2809, 0.0150, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0151, 0.0193, 0.0171, 0.0171, 0.0204, 0.0182, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 12:52:36,455 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:22,155 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:58,629 INFO [train.py:904] (2/8) Epoch 22, batch 9050, loss[loss=0.176, simple_loss=0.2649, pruned_loss=0.04354, over 12794.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2645, pruned_loss=0.03749, over 3011226.20 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:54:04,013 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222205.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:54:09,039 INFO [optim.py:368] (2/8) 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 12:54:17,175 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,423 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,567 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222246.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:55:44,673 INFO [train.py:904] (2/8) Epoch 22, batch 9100, loss[loss=0.1694, simple_loss=0.2687, pruned_loss=0.03502, over 16941.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2639, pruned_loss=0.03777, over 3016099.07 frames. ], batch size: 109, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:56:11,317 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:57:42,765 INFO [train.py:904] (2/8) Epoch 22, batch 9150, loss[loss=0.1682, simple_loss=0.2591, pruned_loss=0.03865, over 16975.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2649, pruned_loss=0.03775, over 3021366.84 frames. ], batch size: 125, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:57:53,760 INFO [optim.py:368] (2/8) 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 12:57:58,309 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9605, 4.0316, 2.7828, 4.8376, 3.1593, 4.6237, 2.6760, 3.3271], device='cuda:2'), covar=tensor([0.0260, 0.0365, 0.1410, 0.0147, 0.0763, 0.0445, 0.1541, 0.0706], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0170, 0.0187, 0.0155, 0.0171, 0.0208, 0.0197, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 12:59:28,938 INFO [train.py:904] (2/8) Epoch 22, batch 9200, loss[loss=0.1625, simple_loss=0.2457, pruned_loss=0.03962, over 12267.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2607, pruned_loss=0.03702, over 3033563.67 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:59:52,291 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:01:05,892 INFO [train.py:904] (2/8) Epoch 22, batch 9250, loss[loss=0.1704, simple_loss=0.2688, pruned_loss=0.03602, over 15385.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2599, pruned_loss=0.03669, over 3013520.18 frames. ], batch size: 192, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:01:16,243 INFO [optim.py:368] (2/8) 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 13:01:28,817 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:02:56,973 INFO [train.py:904] (2/8) Epoch 22, batch 9300, loss[loss=0.1521, simple_loss=0.2426, pruned_loss=0.03083, over 16232.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2582, pruned_loss=0.0359, over 3022950.06 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:03:16,990 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:04:35,071 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5864, 3.5321, 3.5163, 2.8102, 3.4573, 2.0384, 3.2240, 2.8764], device='cuda:2'), covar=tensor([0.0142, 0.0146, 0.0175, 0.0225, 0.0114, 0.2410, 0.0142, 0.0248], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0151, 0.0192, 0.0170, 0.0170, 0.0204, 0.0181, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:04:40,780 INFO [train.py:904] (2/8) Epoch 22, batch 9350, loss[loss=0.1704, simple_loss=0.2619, pruned_loss=0.03941, over 16956.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2582, pruned_loss=0.0356, over 3037730.35 frames. ], batch size: 109, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:04:49,911 INFO [optim.py:368] (2/8) 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 13:05:23,328 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:05:32,410 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-05-01 13:05:57,779 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222541.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:06:08,604 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:20,197 INFO [train.py:904] (2/8) Epoch 22, batch 9400, loss[loss=0.162, simple_loss=0.2439, pruned_loss=0.04004, over 12110.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2576, pruned_loss=0.03556, over 3022091.54 frames. ], batch size: 246, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:06:38,149 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:55,368 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:20,121 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7043, 2.5025, 2.3253, 3.7900, 2.2093, 3.7451, 1.4037, 2.8730], device='cuda:2'), covar=tensor([0.1361, 0.0786, 0.1218, 0.0146, 0.0101, 0.0375, 0.1761, 0.0716], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0172, 0.0192, 0.0183, 0.0199, 0.0211, 0.0200, 0.0190], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:07:24,520 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222584.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:07:42,175 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0410, 2.1675, 2.5526, 2.9474, 2.7452, 3.3966, 2.4085, 3.3053], device='cuda:2'), covar=tensor([0.0206, 0.0510, 0.0348, 0.0284, 0.0353, 0.0159, 0.0461, 0.0147], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0186, 0.0173, 0.0176, 0.0190, 0.0146, 0.0189, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:07:43,446 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:58,857 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-05-01 13:08:00,587 INFO [train.py:904] (2/8) Epoch 22, batch 9450, loss[loss=0.171, simple_loss=0.261, pruned_loss=0.04043, over 15509.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2599, pruned_loss=0.03579, over 3040131.95 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:08:08,413 INFO [optim.py:368] (2/8) 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 13:08:56,846 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:09:23,938 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2021, 5.5078, 5.3086, 5.3234, 5.0257, 4.9690, 4.8681, 5.6088], device='cuda:2'), covar=tensor([0.1091, 0.0905, 0.0911, 0.0787, 0.0769, 0.0812, 0.1133, 0.0747], device='cuda:2'), in_proj_covar=tensor([0.0641, 0.0780, 0.0642, 0.0592, 0.0492, 0.0506, 0.0652, 0.0610], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:09:40,770 INFO [train.py:904] (2/8) Epoch 22, batch 9500, loss[loss=0.1533, simple_loss=0.2475, pruned_loss=0.02958, over 17052.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2589, pruned_loss=0.03516, over 3043728.32 frames. ], batch size: 55, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:10:06,949 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:11:22,395 INFO [train.py:904] (2/8) Epoch 22, batch 9550, loss[loss=0.1579, simple_loss=0.2499, pruned_loss=0.03298, over 12673.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2583, pruned_loss=0.03493, over 3045576.23 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:11:34,508 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:11:51,171 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0634, 4.1248, 3.9580, 3.6713, 3.7216, 4.0773, 3.6943, 3.8695], device='cuda:2'), covar=tensor([0.0571, 0.0680, 0.0314, 0.0266, 0.0684, 0.0516, 0.1059, 0.0547], device='cuda:2'), in_proj_covar=tensor([0.0279, 0.0406, 0.0325, 0.0320, 0.0329, 0.0375, 0.0224, 0.0386], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-05-01 13:12:22,593 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2124, 5.1549, 4.8523, 4.1366, 5.0205, 1.9508, 4.6865, 4.6857], device='cuda:2'), covar=tensor([0.0123, 0.0136, 0.0264, 0.0459, 0.0151, 0.2818, 0.0182, 0.0285], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0151, 0.0191, 0.0169, 0.0170, 0.0204, 0.0181, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:13:00,661 INFO [train.py:904] (2/8) Epoch 22, batch 9600, loss[loss=0.1778, simple_loss=0.2695, pruned_loss=0.04311, over 16472.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2596, pruned_loss=0.03599, over 3031009.19 frames. ], batch size: 75, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:14:44,326 INFO [train.py:904] (2/8) Epoch 22, batch 9650, loss[loss=0.182, simple_loss=0.2801, pruned_loss=0.04197, over 16667.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2618, pruned_loss=0.03633, over 3042930.94 frames. ], batch size: 134, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:14:58,745 INFO [optim.py:368] (2/8) 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 13:16:03,720 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222841.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:16:09,378 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4166, 2.8824, 3.1363, 1.9721, 2.8099, 2.0318, 3.0219, 3.0171], device='cuda:2'), covar=tensor([0.0289, 0.0904, 0.0529, 0.2120, 0.0814, 0.1122, 0.0681, 0.0925], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0158, 0.0163, 0.0151, 0.0141, 0.0127, 0.0140, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:16:27,395 INFO [train.py:904] (2/8) Epoch 22, batch 9700, loss[loss=0.1582, simple_loss=0.2464, pruned_loss=0.03501, over 12292.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2613, pruned_loss=0.03615, over 3057767.78 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:16:43,280 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:16:46,821 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0934, 2.5482, 2.6209, 1.9978, 2.7829, 2.8744, 2.4848, 2.4879], device='cuda:2'), covar=tensor([0.0628, 0.0240, 0.0219, 0.0917, 0.0120, 0.0224, 0.0451, 0.0416], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0104, 0.0092, 0.0134, 0.0077, 0.0119, 0.0124, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 13:17:22,932 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222879.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:17:42,989 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:08,539 INFO [train.py:904] (2/8) Epoch 22, batch 9750, loss[loss=0.1749, simple_loss=0.2717, pruned_loss=0.03901, over 15320.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2596, pruned_loss=0.03591, over 3055824.73 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:18:18,157 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:52,154 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222926.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:24,084 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:45,654 INFO [train.py:904] (2/8) Epoch 22, batch 9800, loss[loss=0.1621, simple_loss=0.2605, pruned_loss=0.03186, over 16708.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2608, pruned_loss=0.03535, over 3083812.21 frames. ], batch size: 134, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:20:15,209 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:20:40,782 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:23,046 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:26,484 INFO [train.py:904] (2/8) Epoch 22, batch 9850, loss[loss=0.1626, simple_loss=0.268, pruned_loss=0.02863, over 16706.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2615, pruned_loss=0.03484, over 3081747.27 frames. ], batch size: 83, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:21:37,441 INFO [optim.py:368] (2/8) 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 13:22:19,883 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223030.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:22:37,853 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 13:23:01,402 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:23:15,937 INFO [train.py:904] (2/8) Epoch 22, batch 9900, loss[loss=0.152, simple_loss=0.2443, pruned_loss=0.02982, over 12660.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2618, pruned_loss=0.03489, over 3079528.30 frames. ], batch size: 248, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:23:43,527 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:25:13,509 INFO [train.py:904] (2/8) Epoch 22, batch 9950, loss[loss=0.1597, simple_loss=0.2633, pruned_loss=0.02806, over 16616.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2635, pruned_loss=0.03527, over 3063711.61 frames. ], batch size: 89, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:25:27,891 INFO [optim.py:368] (2/8) 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 13:25:33,339 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-05-01 13:26:07,045 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:27:13,209 INFO [train.py:904] (2/8) Epoch 22, batch 10000, loss[loss=0.1578, simple_loss=0.2614, pruned_loss=0.02713, over 17075.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2623, pruned_loss=0.03485, over 3087141.82 frames. ], batch size: 97, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:27:51,755 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223173.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:28:04,787 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223179.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:28:12,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7454, 2.1850, 1.8083, 1.9069, 2.4869, 2.1567, 2.1637, 2.5560], device='cuda:2'), covar=tensor([0.0182, 0.0453, 0.0607, 0.0524, 0.0306, 0.0436, 0.0244, 0.0320], device='cuda:2'), in_proj_covar=tensor([0.0201, 0.0229, 0.0222, 0.0221, 0.0230, 0.0228, 0.0224, 0.0220], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:28:54,091 INFO [train.py:904] (2/8) Epoch 22, batch 10050, loss[loss=0.1714, simple_loss=0.2749, pruned_loss=0.03394, over 15362.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2634, pruned_loss=0.03504, over 3098488.63 frames. ], batch size: 191, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:29:04,328 INFO [optim.py:368] (2/8) 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 13:29:05,290 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7988, 1.3398, 1.7202, 1.6854, 1.8461, 1.9522, 1.6892, 1.8537], device='cuda:2'), covar=tensor([0.0301, 0.0447, 0.0245, 0.0347, 0.0332, 0.0215, 0.0449, 0.0156], device='cuda:2'), in_proj_covar=tensor([0.0183, 0.0187, 0.0175, 0.0178, 0.0191, 0.0147, 0.0191, 0.0145], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:29:40,780 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:42,095 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223227.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:54,207 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223234.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 13:30:27,326 INFO [train.py:904] (2/8) Epoch 22, batch 10100, loss[loss=0.1805, simple_loss=0.2721, pruned_loss=0.04447, over 15237.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2632, pruned_loss=0.03501, over 3097001.01 frames. ], batch size: 190, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:31:10,810 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:31:39,312 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223296.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:32:13,429 INFO [train.py:904] (2/8) Epoch 23, batch 0, loss[loss=0.2287, simple_loss=0.2914, pruned_loss=0.08303, over 16889.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.2914, pruned_loss=0.08303, over 16889.00 frames. ], batch size: 109, lr: 2.97e-03, grad_scale: 8.0 2023-05-01 13:32:13,430 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 13:32:20,842 INFO [train.py:938] (2/8) Epoch 23, validation: loss=0.1454, simple_loss=0.2487, pruned_loss=0.02111, over 944034.00 frames. 2023-05-01 13:32:20,843 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 13:32:28,407 INFO [optim.py:368] (2/8) 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 13:32:49,520 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:33:09,273 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:33:26,108 INFO [train.py:904] (2/8) Epoch 23, batch 50, loss[loss=0.1725, simple_loss=0.2521, pruned_loss=0.0464, over 16898.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2692, pruned_loss=0.0475, over 752168.47 frames. ], batch size: 96, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:33:46,493 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1929, 2.3153, 2.7682, 3.0896, 3.0089, 3.5991, 2.4409, 3.5240], device='cuda:2'), covar=tensor([0.0244, 0.0495, 0.0337, 0.0334, 0.0323, 0.0180, 0.0520, 0.0183], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0189, 0.0176, 0.0179, 0.0193, 0.0150, 0.0193, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:34:32,069 INFO [train.py:904] (2/8) Epoch 23, batch 100, loss[loss=0.1519, simple_loss=0.2411, pruned_loss=0.03135, over 17178.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2615, pruned_loss=0.04493, over 1327533.66 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:34:42,058 INFO [optim.py:368] (2/8) 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 13:34:55,417 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:35:08,345 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8554, 4.6376, 4.9334, 5.0783, 5.3281, 4.6841, 5.2636, 5.2998], device='cuda:2'), covar=tensor([0.2133, 0.1427, 0.1790, 0.0846, 0.0571, 0.0954, 0.0659, 0.0674], device='cuda:2'), in_proj_covar=tensor([0.0616, 0.0759, 0.0878, 0.0767, 0.0583, 0.0608, 0.0637, 0.0734], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:35:38,698 INFO [train.py:904] (2/8) Epoch 23, batch 150, loss[loss=0.181, simple_loss=0.2545, pruned_loss=0.0538, over 16861.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2617, pruned_loss=0.04419, over 1778332.59 frames. ], batch size: 96, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:36:47,608 INFO [train.py:904] (2/8) Epoch 23, batch 200, loss[loss=0.1832, simple_loss=0.2771, pruned_loss=0.04472, over 17124.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2625, pruned_loss=0.04451, over 2120792.10 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:36:51,861 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-01 13:36:57,903 INFO [optim.py:368] (2/8) 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 13:37:21,200 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223529.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:37:52,716 INFO [train.py:904] (2/8) Epoch 23, batch 250, loss[loss=0.1913, simple_loss=0.2643, pruned_loss=0.05912, over 16711.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2619, pruned_loss=0.04435, over 2384764.73 frames. ], batch size: 124, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:38:29,152 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:38:54,794 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:38:56,901 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5580, 4.5764, 4.7779, 4.5717, 4.6642, 5.2553, 4.7779, 4.4440], device='cuda:2'), covar=tensor([0.1544, 0.2082, 0.2456, 0.2204, 0.2776, 0.1095, 0.1698, 0.2502], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0579, 0.0640, 0.0477, 0.0642, 0.0666, 0.0503, 0.0639], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 13:39:04,046 INFO [train.py:904] (2/8) Epoch 23, batch 300, loss[loss=0.1452, simple_loss=0.2294, pruned_loss=0.03045, over 16812.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2587, pruned_loss=0.04278, over 2601734.91 frames. ], batch size: 39, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:39:14,755 INFO [optim.py:368] (2/8) 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 13:39:34,896 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:51,495 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 13:39:54,115 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:55,233 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:01,694 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223644.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:13,655 INFO [train.py:904] (2/8) Epoch 23, batch 350, loss[loss=0.1883, simple_loss=0.2883, pruned_loss=0.04413, over 17265.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.256, pruned_loss=0.04182, over 2764808.09 frames. ], batch size: 52, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:40:42,278 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:03,081 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:22,666 INFO [train.py:904] (2/8) Epoch 23, batch 400, loss[loss=0.1627, simple_loss=0.2577, pruned_loss=0.03385, over 16764.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2547, pruned_loss=0.04136, over 2894975.76 frames. ], batch size: 57, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:41:27,349 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0461, 5.3730, 5.0952, 5.1197, 4.8496, 4.8302, 4.7610, 5.4601], device='cuda:2'), covar=tensor([0.1325, 0.0892, 0.1209, 0.0908, 0.0910, 0.0981, 0.1351, 0.0958], device='cuda:2'), in_proj_covar=tensor([0.0671, 0.0815, 0.0671, 0.0618, 0.0514, 0.0526, 0.0685, 0.0636], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:41:34,906 INFO [optim.py:368] (2/8) 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 13:41:46,573 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223720.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:52,271 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-01 13:41:54,669 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-05-01 13:42:32,055 INFO [train.py:904] (2/8) Epoch 23, batch 450, loss[loss=0.1581, simple_loss=0.236, pruned_loss=0.04008, over 16326.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2535, pruned_loss=0.0414, over 2989645.60 frames. ], batch size: 165, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:42:52,142 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:43:03,011 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4054, 4.3625, 4.3057, 3.9927, 4.0606, 4.3970, 4.0551, 4.0890], device='cuda:2'), covar=tensor([0.0728, 0.0886, 0.0384, 0.0354, 0.0855, 0.0515, 0.0695, 0.0788], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0427, 0.0341, 0.0336, 0.0347, 0.0394, 0.0232, 0.0407], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:43:40,975 INFO [train.py:904] (2/8) Epoch 23, batch 500, loss[loss=0.1769, simple_loss=0.265, pruned_loss=0.04446, over 17027.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2527, pruned_loss=0.04063, over 3057108.80 frames. ], batch size: 53, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:43:52,983 INFO [optim.py:368] (2/8) 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 13:43:55,404 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8244, 5.1038, 5.5323, 5.4539, 5.5118, 5.1167, 4.7703, 4.9374], device='cuda:2'), covar=tensor([0.0690, 0.0887, 0.0561, 0.0734, 0.0847, 0.0730, 0.1813, 0.0594], device='cuda:2'), in_proj_covar=tensor([0.0413, 0.0459, 0.0446, 0.0413, 0.0489, 0.0468, 0.0550, 0.0374], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 13:44:16,845 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223829.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:44:25,374 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4316, 2.2358, 2.2599, 4.1694, 2.3106, 2.6883, 2.3352, 2.4513], device='cuda:2'), covar=tensor([0.1260, 0.3601, 0.3197, 0.0531, 0.4155, 0.2623, 0.3597, 0.3702], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0451, 0.0370, 0.0328, 0.0439, 0.0517, 0.0422, 0.0526], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:44:49,713 INFO [train.py:904] (2/8) Epoch 23, batch 550, loss[loss=0.1567, simple_loss=0.2478, pruned_loss=0.03286, over 17014.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2522, pruned_loss=0.04055, over 3116958.43 frames. ], batch size: 50, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:45:13,057 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-05-01 13:45:16,312 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2527, 4.2189, 4.1636, 3.8917, 3.9406, 4.2417, 3.8800, 4.0170], device='cuda:2'), covar=tensor([0.0601, 0.0704, 0.0327, 0.0278, 0.0648, 0.0424, 0.0825, 0.0579], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0431, 0.0344, 0.0340, 0.0350, 0.0397, 0.0235, 0.0410], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:45:24,265 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223877.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:45:50,307 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0997, 2.6631, 2.5846, 4.3542, 3.3289, 4.0794, 1.7596, 2.9565], device='cuda:2'), covar=tensor([0.1339, 0.0885, 0.1293, 0.0238, 0.0188, 0.0520, 0.1724, 0.0924], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0174, 0.0193, 0.0188, 0.0201, 0.0215, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:45:59,107 INFO [train.py:904] (2/8) Epoch 23, batch 600, loss[loss=0.1731, simple_loss=0.24, pruned_loss=0.05309, over 16901.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2512, pruned_loss=0.04056, over 3161840.32 frames. ], batch size: 90, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:46:10,991 INFO [optim.py:368] (2/8) 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 13:46:20,918 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5953, 3.8150, 4.2111, 2.4277, 3.3699, 2.7111, 4.0309, 3.9470], device='cuda:2'), covar=tensor([0.0290, 0.0959, 0.0438, 0.1948, 0.0836, 0.0960, 0.0584, 0.1102], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0163, 0.0166, 0.0153, 0.0145, 0.0129, 0.0142, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:46:41,951 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:47:08,240 INFO [train.py:904] (2/8) Epoch 23, batch 650, loss[loss=0.1804, simple_loss=0.2579, pruned_loss=0.05144, over 16502.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2498, pruned_loss=0.03978, over 3196095.42 frames. ], batch size: 146, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:22,013 INFO [train.py:904] (2/8) Epoch 23, batch 700, loss[loss=0.1795, simple_loss=0.255, pruned_loss=0.05197, over 12505.00 frames. ], tot_loss[loss=0.165, simple_loss=0.25, pruned_loss=0.03998, over 3213636.33 frames. ], batch size: 247, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:35,477 INFO [optim.py:368] (2/8) 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 13:48:38,627 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 13:48:40,809 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-05-01 13:49:33,339 INFO [train.py:904] (2/8) Epoch 23, batch 750, loss[loss=0.1699, simple_loss=0.2644, pruned_loss=0.03766, over 17037.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2504, pruned_loss=0.04031, over 3239009.98 frames. ], batch size: 55, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:49:55,340 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0737, 3.0433, 2.0344, 3.2363, 2.4011, 3.2948, 2.2335, 2.5946], device='cuda:2'), covar=tensor([0.0350, 0.0514, 0.1570, 0.0445, 0.0788, 0.0668, 0.1383, 0.0801], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0178, 0.0195, 0.0166, 0.0178, 0.0217, 0.0205, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:50:41,986 INFO [train.py:904] (2/8) Epoch 23, batch 800, loss[loss=0.1629, simple_loss=0.2452, pruned_loss=0.04027, over 16706.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2504, pruned_loss=0.04048, over 3258568.36 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:50:54,816 INFO [optim.py:368] (2/8) 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 13:51:45,042 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0723, 3.8489, 4.2927, 2.0662, 4.3317, 4.6020, 3.5033, 3.5769], device='cuda:2'), covar=tensor([0.0721, 0.0264, 0.0232, 0.1327, 0.0124, 0.0154, 0.0400, 0.0407], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0098, 0.0140, 0.0081, 0.0126, 0.0128, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:51:51,763 INFO [train.py:904] (2/8) Epoch 23, batch 850, loss[loss=0.1621, simple_loss=0.2549, pruned_loss=0.03463, over 17198.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2497, pruned_loss=0.03997, over 3258718.21 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:00,735 INFO [train.py:904] (2/8) Epoch 23, batch 900, loss[loss=0.1588, simple_loss=0.2328, pruned_loss=0.0424, over 16873.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.25, pruned_loss=0.03979, over 3273027.12 frames. ], batch size: 116, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:06,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6031, 3.6646, 2.2831, 3.9408, 2.8528, 3.9070, 2.2595, 2.9292], device='cuda:2'), covar=tensor([0.0297, 0.0454, 0.1598, 0.0426, 0.0826, 0.0812, 0.1537, 0.0809], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0179, 0.0195, 0.0167, 0.0179, 0.0219, 0.0206, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:53:14,895 INFO [optim.py:368] (2/8) 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 13:53:15,244 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224212.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:53:32,990 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6339, 2.6673, 2.6835, 4.7697, 3.8951, 4.1747, 1.6944, 3.1333], device='cuda:2'), covar=tensor([0.1527, 0.0939, 0.1259, 0.0252, 0.0241, 0.0455, 0.1768, 0.0836], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0190, 0.0203, 0.0215, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 13:53:46,031 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:10,198 INFO [train.py:904] (2/8) Epoch 23, batch 950, loss[loss=0.1598, simple_loss=0.2556, pruned_loss=0.03195, over 17136.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2499, pruned_loss=0.03934, over 3278051.70 frames. ], batch size: 47, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:54:38,577 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224273.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:50,479 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:55:20,381 INFO [train.py:904] (2/8) Epoch 23, batch 1000, loss[loss=0.1619, simple_loss=0.2356, pruned_loss=0.04412, over 16814.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2495, pruned_loss=0.03974, over 3287613.77 frames. ], batch size: 102, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:55:33,052 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-05-01 13:55:33,533 INFO [optim.py:368] (2/8) 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 13:56:17,332 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 13:56:31,386 INFO [train.py:904] (2/8) Epoch 23, batch 1050, loss[loss=0.1526, simple_loss=0.2295, pruned_loss=0.03782, over 16684.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2489, pruned_loss=0.04002, over 3298339.42 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:42,157 INFO [train.py:904] (2/8) Epoch 23, batch 1100, loss[loss=0.2023, simple_loss=0.2719, pruned_loss=0.06636, over 11916.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2477, pruned_loss=0.03942, over 3293472.85 frames. ], batch size: 246, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:54,067 INFO [optim.py:368] (2/8) 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 13:58:01,404 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8091, 1.9885, 2.3732, 2.6873, 2.7257, 2.7027, 2.0617, 2.9045], device='cuda:2'), covar=tensor([0.0197, 0.0510, 0.0346, 0.0318, 0.0325, 0.0336, 0.0518, 0.0180], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0194, 0.0181, 0.0186, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:58:05,468 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0868, 1.9844, 2.5496, 2.9690, 2.9445, 2.9934, 1.9624, 3.1643], device='cuda:2'), covar=tensor([0.0185, 0.0564, 0.0334, 0.0275, 0.0270, 0.0284, 0.0678, 0.0179], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0194, 0.0181, 0.0186, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 13:58:14,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0239, 4.5322, 4.4959, 3.1863, 3.7733, 4.5011, 4.0178, 2.6319], device='cuda:2'), covar=tensor([0.0471, 0.0062, 0.0044, 0.0378, 0.0141, 0.0093, 0.0081, 0.0487], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0085, 0.0085, 0.0135, 0.0100, 0.0110, 0.0096, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 13:58:51,565 INFO [train.py:904] (2/8) Epoch 23, batch 1150, loss[loss=0.1471, simple_loss=0.2292, pruned_loss=0.03249, over 16512.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2473, pruned_loss=0.03899, over 3297570.68 frames. ], batch size: 75, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 14:00:00,970 INFO [train.py:904] (2/8) Epoch 23, batch 1200, loss[loss=0.1655, simple_loss=0.2472, pruned_loss=0.04185, over 16738.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2472, pruned_loss=0.03863, over 3309675.27 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:00:14,523 INFO [optim.py:368] (2/8) 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 14:00:42,580 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-05-01 14:01:10,499 INFO [train.py:904] (2/8) Epoch 23, batch 1250, loss[loss=0.1698, simple_loss=0.2427, pruned_loss=0.04848, over 16755.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2466, pruned_loss=0.03837, over 3320290.45 frames. ], batch size: 124, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:01:22,933 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224562.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:01:31,196 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:02:20,413 INFO [train.py:904] (2/8) Epoch 23, batch 1300, loss[loss=0.1368, simple_loss=0.2221, pruned_loss=0.02579, over 16822.00 frames. ], tot_loss[loss=0.161, simple_loss=0.246, pruned_loss=0.03803, over 3330726.79 frames. ], batch size: 39, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:02:33,670 INFO [optim.py:368] (2/8) 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 14:02:43,547 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8860, 2.6592, 2.4813, 4.2073, 3.2923, 4.0618, 1.5830, 2.9090], device='cuda:2'), covar=tensor([0.1387, 0.0826, 0.1280, 0.0206, 0.0196, 0.0420, 0.1782, 0.0850], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0191, 0.0203, 0.0216, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:02:49,064 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224623.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:03:29,373 INFO [train.py:904] (2/8) Epoch 23, batch 1350, loss[loss=0.1885, simple_loss=0.2616, pruned_loss=0.05772, over 16927.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2461, pruned_loss=0.03814, over 3329507.08 frames. ], batch size: 109, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:03:48,975 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 14:04:40,082 INFO [train.py:904] (2/8) Epoch 23, batch 1400, loss[loss=0.1518, simple_loss=0.2357, pruned_loss=0.03399, over 17189.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2463, pruned_loss=0.03837, over 3321936.85 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:52,701 INFO [optim.py:368] (2/8) 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 14:05:13,957 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:33,016 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:50,556 INFO [train.py:904] (2/8) Epoch 23, batch 1450, loss[loss=0.1606, simple_loss=0.2513, pruned_loss=0.03491, over 17098.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2459, pruned_loss=0.03875, over 3306900.05 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:06:39,720 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0648, 2.1212, 2.6765, 2.9877, 2.8556, 3.4959, 2.4801, 3.4818], device='cuda:2'), covar=tensor([0.0285, 0.0528, 0.0363, 0.0377, 0.0391, 0.0198, 0.0475, 0.0172], device='cuda:2'), in_proj_covar=tensor([0.0193, 0.0193, 0.0181, 0.0186, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:06:39,729 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:06:59,977 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224801.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:07:01,746 INFO [train.py:904] (2/8) Epoch 23, batch 1500, loss[loss=0.139, simple_loss=0.2183, pruned_loss=0.02992, over 16012.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2452, pruned_loss=0.03841, over 3314702.36 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:07:02,250 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2285, 3.1805, 2.0230, 3.4294, 2.5049, 3.4622, 2.2364, 2.6617], device='cuda:2'), covar=tensor([0.0331, 0.0485, 0.1549, 0.0344, 0.0791, 0.0660, 0.1394, 0.0729], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0181, 0.0197, 0.0170, 0.0179, 0.0221, 0.0207, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:07:12,853 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4111, 5.3852, 5.1486, 4.6567, 5.1658, 2.2799, 4.9725, 5.1322], device='cuda:2'), covar=tensor([0.0129, 0.0116, 0.0225, 0.0436, 0.0144, 0.2521, 0.0149, 0.0212], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0163, 0.0205, 0.0181, 0.0183, 0.0215, 0.0195, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:07:16,529 INFO [optim.py:368] (2/8) 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 14:07:38,566 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8792, 1.4244, 1.6936, 1.7174, 1.8905, 1.9650, 1.6736, 1.8652], device='cuda:2'), covar=tensor([0.0257, 0.0386, 0.0233, 0.0325, 0.0283, 0.0199, 0.0418, 0.0153], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0193, 0.0181, 0.0185, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:08:14,250 INFO [train.py:904] (2/8) Epoch 23, batch 1550, loss[loss=0.1743, simple_loss=0.2561, pruned_loss=0.04624, over 16437.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2467, pruned_loss=0.03925, over 3309111.72 frames. ], batch size: 68, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 14:08:34,704 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224868.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:23,916 INFO [train.py:904] (2/8) Epoch 23, batch 1600, loss[loss=0.1674, simple_loss=0.2419, pruned_loss=0.0464, over 16806.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2485, pruned_loss=0.03905, over 3315944.86 frames. ], batch size: 116, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:09:36,820 INFO [optim.py:368] (2/8) 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 14:09:41,251 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224916.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:43,665 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:10:19,501 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0782, 5.6483, 5.7766, 5.4524, 5.5559, 6.1319, 5.5689, 5.2436], device='cuda:2'), covar=tensor([0.0999, 0.1975, 0.2080, 0.2203, 0.2564, 0.0933, 0.1561, 0.2528], device='cuda:2'), in_proj_covar=tensor([0.0418, 0.0613, 0.0676, 0.0501, 0.0677, 0.0702, 0.0530, 0.0674], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 14:10:32,928 INFO [train.py:904] (2/8) Epoch 23, batch 1650, loss[loss=0.1367, simple_loss=0.2233, pruned_loss=0.02509, over 16961.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.25, pruned_loss=0.03927, over 3319312.73 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:30,133 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7022, 6.0888, 5.7999, 5.9221, 5.5045, 5.4513, 5.5192, 6.2455], device='cuda:2'), covar=tensor([0.1522, 0.0981, 0.1322, 0.0900, 0.1043, 0.0648, 0.1354, 0.0951], device='cuda:2'), in_proj_covar=tensor([0.0697, 0.0851, 0.0697, 0.0643, 0.0536, 0.0543, 0.0713, 0.0663], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:11:41,679 INFO [train.py:904] (2/8) Epoch 23, batch 1700, loss[loss=0.1752, simple_loss=0.2692, pruned_loss=0.04058, over 17064.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2517, pruned_loss=0.03989, over 3327736.40 frames. ], batch size: 55, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:56,160 INFO [optim.py:368] (2/8) 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 14:12:07,186 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1243, 3.1453, 3.1110, 5.2060, 4.3128, 4.5123, 2.0112, 3.5312], device='cuda:2'), covar=tensor([0.1228, 0.0736, 0.1006, 0.0203, 0.0258, 0.0417, 0.1473, 0.0682], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0192, 0.0204, 0.0217, 0.0204, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:12:49,834 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8426, 4.9464, 5.3231, 5.2919, 5.3219, 4.9404, 4.9076, 4.7172], device='cuda:2'), covar=tensor([0.0347, 0.0547, 0.0400, 0.0448, 0.0486, 0.0461, 0.1034, 0.0518], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0471, 0.0459, 0.0422, 0.0505, 0.0481, 0.0563, 0.0384], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 14:12:52,528 INFO [train.py:904] (2/8) Epoch 23, batch 1750, loss[loss=0.1639, simple_loss=0.2443, pruned_loss=0.0417, over 16748.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2525, pruned_loss=0.03988, over 3327581.92 frames. ], batch size: 83, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:12:52,921 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6985, 3.7215, 3.9469, 2.7031, 3.5682, 4.0202, 3.6506, 2.3913], device='cuda:2'), covar=tensor([0.0465, 0.0252, 0.0056, 0.0392, 0.0115, 0.0086, 0.0112, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0086, 0.0086, 0.0136, 0.0100, 0.0111, 0.0097, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 14:13:23,597 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0120, 4.4619, 4.4559, 3.1691, 3.7727, 4.4523, 3.9725, 2.7142], device='cuda:2'), covar=tensor([0.0452, 0.0084, 0.0046, 0.0363, 0.0129, 0.0086, 0.0087, 0.0464], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0086, 0.0086, 0.0136, 0.0100, 0.0111, 0.0097, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 14:13:32,977 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:13:52,569 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:14:01,698 INFO [train.py:904] (2/8) Epoch 23, batch 1800, loss[loss=0.1585, simple_loss=0.248, pruned_loss=0.03451, over 17238.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2534, pruned_loss=0.04007, over 3322106.76 frames. ], batch size: 45, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:14:02,307 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:14:15,809 INFO [optim.py:368] (2/8) 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 14:14:32,159 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6787, 3.7795, 2.3983, 4.0943, 2.9237, 4.0110, 2.5207, 3.0629], device='cuda:2'), covar=tensor([0.0302, 0.0418, 0.1552, 0.0381, 0.0760, 0.0826, 0.1374, 0.0731], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0169, 0.0179, 0.0221, 0.0205, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:14:57,206 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:15:12,126 INFO [train.py:904] (2/8) Epoch 23, batch 1850, loss[loss=0.1875, simple_loss=0.2627, pruned_loss=0.05615, over 16901.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2543, pruned_loss=0.04034, over 3324466.96 frames. ], batch size: 109, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:15:13,098 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-05-01 14:15:27,759 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:16:19,025 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2283, 5.8263, 5.9794, 5.6352, 5.7924, 6.3123, 5.7947, 5.4986], device='cuda:2'), covar=tensor([0.0850, 0.1892, 0.2186, 0.2041, 0.2408, 0.0835, 0.1615, 0.2283], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0612, 0.0675, 0.0500, 0.0677, 0.0698, 0.0529, 0.0673], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 14:16:22,243 INFO [train.py:904] (2/8) Epoch 23, batch 1900, loss[loss=0.1637, simple_loss=0.262, pruned_loss=0.03272, over 17104.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2534, pruned_loss=0.03981, over 3321533.59 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:16:22,762 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:16:36,681 INFO [optim.py:368] (2/8) 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 14:16:43,948 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:17:31,820 INFO [train.py:904] (2/8) Epoch 23, batch 1950, loss[loss=0.1509, simple_loss=0.239, pruned_loss=0.03147, over 17200.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2537, pruned_loss=0.0399, over 3332740.88 frames. ], batch size: 44, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:17:50,055 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:18:24,639 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0427, 4.8144, 5.1117, 5.2767, 5.5210, 4.7317, 5.4643, 5.4824], device='cuda:2'), covar=tensor([0.2187, 0.1478, 0.1990, 0.0882, 0.0561, 0.1017, 0.0540, 0.0622], device='cuda:2'), in_proj_covar=tensor([0.0677, 0.0838, 0.0971, 0.0844, 0.0639, 0.0672, 0.0694, 0.0805], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:18:37,329 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8097, 2.6878, 2.3545, 2.7379, 3.0614, 2.8725, 3.4280, 3.2918], device='cuda:2'), covar=tensor([0.0146, 0.0473, 0.0545, 0.0403, 0.0319, 0.0405, 0.0281, 0.0271], device='cuda:2'), in_proj_covar=tensor([0.0221, 0.0242, 0.0232, 0.0232, 0.0243, 0.0242, 0.0243, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:18:42,383 INFO [train.py:904] (2/8) Epoch 23, batch 2000, loss[loss=0.1641, simple_loss=0.248, pruned_loss=0.04013, over 16572.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.254, pruned_loss=0.03947, over 3324981.25 frames. ], batch size: 68, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:18:56,051 INFO [optim.py:368] (2/8) 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 14:19:52,847 INFO [train.py:904] (2/8) Epoch 23, batch 2050, loss[loss=0.1745, simple_loss=0.2472, pruned_loss=0.0509, over 16724.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2539, pruned_loss=0.03948, over 3327982.17 frames. ], batch size: 89, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:20:02,748 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 14:20:35,611 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:20:54,895 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225396.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:04,272 INFO [train.py:904] (2/8) Epoch 23, batch 2100, loss[loss=0.1837, simple_loss=0.282, pruned_loss=0.04268, over 16722.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2551, pruned_loss=0.03999, over 3316742.28 frames. ], batch size: 57, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:21:09,829 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7732, 1.3956, 1.7253, 1.6865, 1.9003, 1.9752, 1.6099, 1.8544], device='cuda:2'), covar=tensor([0.0277, 0.0418, 0.0233, 0.0332, 0.0259, 0.0208, 0.0436, 0.0152], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0188, 0.0200, 0.0158, 0.0199, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:21:18,934 INFO [optim.py:368] (2/8) 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 14:21:29,425 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 14:21:44,080 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:56,549 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:03,498 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225444.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:15,764 INFO [train.py:904] (2/8) Epoch 23, batch 2150, loss[loss=0.1384, simple_loss=0.2241, pruned_loss=0.02634, over 16997.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2556, pruned_loss=0.04057, over 3317438.52 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:22:25,282 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:25,353 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3877, 5.3749, 5.2843, 4.7220, 4.8838, 5.2850, 5.2313, 4.9053], device='cuda:2'), covar=tensor([0.0615, 0.0599, 0.0305, 0.0385, 0.1126, 0.0548, 0.0295, 0.0777], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0456, 0.0363, 0.0361, 0.0367, 0.0421, 0.0246, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 14:22:29,388 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:57,584 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3617, 3.6520, 3.9422, 2.1940, 3.1417, 2.5432, 3.8771, 3.8157], device='cuda:2'), covar=tensor([0.0329, 0.0919, 0.0536, 0.2145, 0.0865, 0.1049, 0.0635, 0.1098], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:23:18,692 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225498.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:22,033 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:26,043 INFO [train.py:904] (2/8) Epoch 23, batch 2200, loss[loss=0.21, simple_loss=0.2897, pruned_loss=0.06513, over 12072.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2562, pruned_loss=0.04072, over 3317120.80 frames. ], batch size: 247, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:23:29,406 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225505.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:40,044 INFO [optim.py:368] (2/8) 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 14:23:53,567 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225522.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:54,699 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:24:36,369 INFO [train.py:904] (2/8) Epoch 23, batch 2250, loss[loss=0.1745, simple_loss=0.2502, pruned_loss=0.04939, over 16885.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2568, pruned_loss=0.04137, over 3313396.67 frames. ], batch size: 109, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:24:39,283 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8366, 1.4404, 1.6891, 1.6829, 1.8538, 2.0065, 1.6457, 1.8760], device='cuda:2'), covar=tensor([0.0278, 0.0427, 0.0236, 0.0330, 0.0304, 0.0203, 0.0429, 0.0145], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0188, 0.0200, 0.0157, 0.0199, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:24:45,099 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225559.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:24:54,934 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:18,189 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:46,973 INFO [train.py:904] (2/8) Epoch 23, batch 2300, loss[loss=0.1396, simple_loss=0.229, pruned_loss=0.02514, over 17033.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2566, pruned_loss=0.04133, over 3303626.04 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:25:56,971 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 14:26:01,554 INFO [optim.py:368] (2/8) 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 14:26:11,624 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225620.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 14:26:58,883 INFO [train.py:904] (2/8) Epoch 23, batch 2350, loss[loss=0.1489, simple_loss=0.2436, pruned_loss=0.02703, over 16859.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2572, pruned_loss=0.04202, over 3304079.38 frames. ], batch size: 42, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:27:34,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9930, 2.0065, 2.2064, 3.7146, 2.0248, 2.2442, 2.1008, 2.1575], device='cuda:2'), covar=tensor([0.1894, 0.4533, 0.3376, 0.0819, 0.4956, 0.3266, 0.4563, 0.4044], device='cuda:2'), in_proj_covar=tensor([0.0410, 0.0459, 0.0376, 0.0335, 0.0443, 0.0530, 0.0431, 0.0537], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:28:10,319 INFO [train.py:904] (2/8) Epoch 23, batch 2400, loss[loss=0.1824, simple_loss=0.2603, pruned_loss=0.05225, over 16717.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2581, pruned_loss=0.04218, over 3316630.54 frames. ], batch size: 134, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:28:23,215 INFO [optim.py:368] (2/8) 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 14:29:17,841 INFO [train.py:904] (2/8) Epoch 23, batch 2450, loss[loss=0.1526, simple_loss=0.2457, pruned_loss=0.0297, over 17189.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2586, pruned_loss=0.04212, over 3316445.41 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:29:26,303 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:10,473 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-01 14:30:17,189 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:22,202 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225798.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:29,834 INFO [train.py:904] (2/8) Epoch 23, batch 2500, loss[loss=0.1596, simple_loss=0.2506, pruned_loss=0.0343, over 16017.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2592, pruned_loss=0.04217, over 3314938.86 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:30:36,106 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 14:30:50,742 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225818.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:28,974 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:38,344 INFO [train.py:904] (2/8) Epoch 23, batch 2550, loss[loss=0.1872, simple_loss=0.2836, pruned_loss=0.0454, over 16748.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2594, pruned_loss=0.042, over 3316125.41 frames. ], batch size: 57, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:31:42,691 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2284, 5.1907, 5.1235, 4.5879, 4.7618, 5.1235, 5.0441, 4.7512], device='cuda:2'), covar=tensor([0.0620, 0.0542, 0.0310, 0.0386, 0.1083, 0.0507, 0.0331, 0.0810], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0456, 0.0362, 0.0361, 0.0367, 0.0420, 0.0245, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 14:31:49,568 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:50,851 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225862.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:14,111 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225878.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:38,128 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 23, batch 2600, loss[loss=0.1714, simple_loss=0.2663, pruned_loss=0.03825, over 17241.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2587, pruned_loss=0.04153, over 3308581.28 frames. ], batch size: 52, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:33:03,045 INFO [optim.py:368] (2/8) 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 14:33:06,103 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225915.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 14:33:16,549 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:33:58,792 INFO [train.py:904] (2/8) Epoch 23, batch 2650, loss[loss=0.1585, simple_loss=0.2499, pruned_loss=0.03353, over 17177.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2594, pruned_loss=0.04098, over 3309212.36 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:34:00,153 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3173, 3.4801, 3.9011, 2.1196, 3.0757, 2.5315, 3.8362, 3.7022], device='cuda:2'), covar=tensor([0.0300, 0.0902, 0.0547, 0.2091, 0.0884, 0.1009, 0.0620, 0.1113], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0168, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-01 14:34:04,235 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:35:08,791 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-05-01 14:35:12,272 INFO [train.py:904] (2/8) Epoch 23, batch 2700, loss[loss=0.1806, simple_loss=0.27, pruned_loss=0.04553, over 16269.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2599, pruned_loss=0.04071, over 3310968.63 frames. ], batch size: 35, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:35:24,951 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3549, 4.3940, 4.7130, 4.7090, 4.7542, 4.4426, 4.4599, 4.3007], device='cuda:2'), covar=tensor([0.0385, 0.0641, 0.0419, 0.0444, 0.0575, 0.0455, 0.0848, 0.0703], device='cuda:2'), in_proj_covar=tensor([0.0425, 0.0474, 0.0462, 0.0426, 0.0507, 0.0483, 0.0567, 0.0385], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 14:35:25,733 INFO [optim.py:368] (2/8) 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 14:36:06,286 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:36:23,069 INFO [train.py:904] (2/8) Epoch 23, batch 2750, loss[loss=0.158, simple_loss=0.2496, pruned_loss=0.03317, over 17221.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2593, pruned_loss=0.04035, over 3316213.65 frames. ], batch size: 44, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:36:42,291 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2666, 1.5848, 2.0025, 2.1237, 2.3172, 2.3591, 1.8392, 2.3522], device='cuda:2'), covar=tensor([0.0232, 0.0510, 0.0287, 0.0311, 0.0314, 0.0268, 0.0504, 0.0170], device='cuda:2'), in_proj_covar=tensor([0.0196, 0.0197, 0.0184, 0.0190, 0.0201, 0.0160, 0.0200, 0.0156], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:37:21,901 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:31,382 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 23, batch 2800, loss[loss=0.1385, simple_loss=0.2363, pruned_loss=0.02032, over 16891.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2585, pruned_loss=0.04028, over 3319458.93 frames. ], batch size: 42, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:47,345 INFO [optim.py:368] (2/8) 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 14:37:54,712 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:38:29,330 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 23, batch 2850, loss[loss=0.228, simple_loss=0.306, pruned_loss=0.07497, over 12110.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2584, pruned_loss=0.04015, over 3311804.55 frames. ], batch size: 246, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:38:47,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6752, 4.9856, 4.8387, 4.8326, 4.6155, 4.5700, 4.4766, 5.1163], device='cuda:2'), covar=tensor([0.1355, 0.0946, 0.1056, 0.0797, 0.0877, 0.1138, 0.1162, 0.0899], device='cuda:2'), in_proj_covar=tensor([0.0704, 0.0858, 0.0705, 0.0652, 0.0542, 0.0547, 0.0720, 0.0670], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:38:54,380 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:00,761 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:18,076 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:51,833 INFO [train.py:904] (2/8) Epoch 23, batch 2900, loss[loss=0.134, simple_loss=0.224, pruned_loss=0.02203, over 17234.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2582, pruned_loss=0.04085, over 3307189.43 frames. ], batch size: 45, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:40:00,214 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226209.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:05,830 INFO [optim.py:368] (2/8) 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 14:40:09,251 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226215.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 14:40:13,515 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:24,108 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:34,229 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:58,437 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:41:00,393 INFO [train.py:904] (2/8) Epoch 23, batch 2950, loss[loss=0.2033, simple_loss=0.2704, pruned_loss=0.06813, over 11407.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2569, pruned_loss=0.04127, over 3311460.66 frames. ], batch size: 246, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:41:14,089 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226263.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:41:56,357 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226294.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:42:08,778 INFO [train.py:904] (2/8) Epoch 23, batch 3000, loss[loss=0.1495, simple_loss=0.2429, pruned_loss=0.02809, over 17109.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2566, pruned_loss=0.0414, over 3314562.52 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:42:08,779 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 14:42:17,859 INFO [train.py:938] (2/8) Epoch 23, validation: loss=0.1344, simple_loss=0.2397, pruned_loss=0.01456, over 944034.00 frames. 2023-05-01 14:42:17,860 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 14:42:30,994 INFO [optim.py:368] (2/8) 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 14:43:26,884 INFO [train.py:904] (2/8) Epoch 23, batch 3050, loss[loss=0.2031, simple_loss=0.2822, pruned_loss=0.06201, over 12615.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2571, pruned_loss=0.04188, over 3321158.62 frames. ], batch size: 247, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:44:14,500 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5405, 3.5975, 4.3608, 2.3588, 3.3202, 2.5802, 4.0494, 3.8862], device='cuda:2'), covar=tensor([0.0277, 0.0899, 0.0393, 0.1944, 0.0795, 0.0988, 0.0580, 0.1055], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0168, 0.0169, 0.0155, 0.0146, 0.0131, 0.0145, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-01 14:44:29,376 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:44:37,981 INFO [train.py:904] (2/8) Epoch 23, batch 3100, loss[loss=0.148, simple_loss=0.2333, pruned_loss=0.0313, over 16762.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2552, pruned_loss=0.04129, over 3327354.84 frames. ], batch size: 39, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:44:51,600 INFO [optim.py:368] (2/8) 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 14:45:37,568 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6760, 3.9137, 2.5796, 4.4720, 3.0590, 4.4489, 2.4980, 3.1069], device='cuda:2'), covar=tensor([0.0360, 0.0418, 0.1486, 0.0304, 0.0904, 0.0517, 0.1581, 0.0832], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0171, 0.0180, 0.0223, 0.0207, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:45:47,149 INFO [train.py:904] (2/8) Epoch 23, batch 3150, loss[loss=0.1468, simple_loss=0.2464, pruned_loss=0.02361, over 17080.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2551, pruned_loss=0.04138, over 3330295.91 frames. ], batch size: 50, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:45:56,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5785, 2.6030, 2.2411, 2.4982, 2.9300, 2.6576, 3.1245, 3.1529], device='cuda:2'), covar=tensor([0.0174, 0.0459, 0.0571, 0.0473, 0.0316, 0.0436, 0.0320, 0.0302], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0245, 0.0233, 0.0235, 0.0245, 0.0243, 0.0246, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:46:20,535 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 14:46:54,853 INFO [train.py:904] (2/8) Epoch 23, batch 3200, loss[loss=0.1813, simple_loss=0.287, pruned_loss=0.03784, over 17260.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2553, pruned_loss=0.04133, over 3317017.05 frames. ], batch size: 52, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:47:09,850 INFO [optim.py:368] (2/8) 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 14:47:15,790 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:47:54,037 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7557, 6.0902, 5.8259, 5.8476, 5.4225, 5.4323, 5.4156, 6.2170], device='cuda:2'), covar=tensor([0.1411, 0.0940, 0.1122, 0.0919, 0.0950, 0.0706, 0.1263, 0.0903], device='cuda:2'), in_proj_covar=tensor([0.0704, 0.0857, 0.0706, 0.0650, 0.0542, 0.0548, 0.0720, 0.0669], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:47:58,098 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6675, 2.9859, 3.1708, 2.1618, 2.7925, 2.2098, 3.3123, 3.3126], device='cuda:2'), covar=tensor([0.0243, 0.0843, 0.0564, 0.1839, 0.0834, 0.0983, 0.0541, 0.0817], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-01 14:48:01,429 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:01,599 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8985, 4.0783, 2.7197, 4.6542, 3.2211, 4.6115, 2.8042, 3.3519], device='cuda:2'), covar=tensor([0.0308, 0.0427, 0.1462, 0.0295, 0.0815, 0.0472, 0.1441, 0.0721], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0195, 0.0170, 0.0179, 0.0223, 0.0206, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:48:04,026 INFO [train.py:904] (2/8) Epoch 23, batch 3250, loss[loss=0.1672, simple_loss=0.2653, pruned_loss=0.0345, over 16511.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.255, pruned_loss=0.04112, over 3319760.50 frames. ], batch size: 62, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:48:22,283 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:54,490 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226589.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:08,848 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226599.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:14,396 INFO [train.py:904] (2/8) Epoch 23, batch 3300, loss[loss=0.172, simple_loss=0.2615, pruned_loss=0.04125, over 16741.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2559, pruned_loss=0.04116, over 3317742.55 frames. ], batch size: 83, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:49:27,927 INFO [optim.py:368] (2/8) 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 14:50:23,120 INFO [train.py:904] (2/8) Epoch 23, batch 3350, loss[loss=0.1989, simple_loss=0.2769, pruned_loss=0.06044, over 16904.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2571, pruned_loss=0.04132, over 3317079.25 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:50:51,269 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2427, 4.0859, 4.3195, 4.4599, 4.5528, 4.1242, 4.3264, 4.5519], device='cuda:2'), covar=tensor([0.1708, 0.1254, 0.1334, 0.0699, 0.0627, 0.1362, 0.2736, 0.0669], device='cuda:2'), in_proj_covar=tensor([0.0683, 0.0849, 0.0981, 0.0855, 0.0648, 0.0680, 0.0702, 0.0817], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:51:25,612 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:51:34,003 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0073, 2.8544, 2.7677, 4.8471, 3.8903, 4.3417, 1.7100, 3.1854], device='cuda:2'), covar=tensor([0.1230, 0.0758, 0.1128, 0.0209, 0.0244, 0.0386, 0.1583, 0.0724], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0195, 0.0206, 0.0218, 0.0204, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:51:34,603 INFO [train.py:904] (2/8) Epoch 23, batch 3400, loss[loss=0.1755, simple_loss=0.2582, pruned_loss=0.04635, over 16813.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2565, pruned_loss=0.04078, over 3327648.74 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:47,765 INFO [optim.py:368] (2/8) 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 14:52:33,238 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 23, batch 3450, loss[loss=0.1848, simple_loss=0.2624, pruned_loss=0.05361, over 12031.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2559, pruned_loss=0.04085, over 3310970.07 frames. ], batch size: 247, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:53:47,250 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3934, 4.3936, 4.5453, 4.3758, 4.4211, 4.9889, 4.5209, 4.2167], device='cuda:2'), covar=tensor([0.1913, 0.2225, 0.2275, 0.2237, 0.2781, 0.1205, 0.1776, 0.2842], device='cuda:2'), in_proj_covar=tensor([0.0427, 0.0620, 0.0686, 0.0511, 0.0685, 0.0709, 0.0538, 0.0685], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 14:53:52,947 INFO [train.py:904] (2/8) Epoch 23, batch 3500, loss[loss=0.1791, simple_loss=0.2565, pruned_loss=0.05082, over 16918.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.255, pruned_loss=0.04085, over 3315666.18 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:54:07,221 INFO [optim.py:368] (2/8) 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 14:54:52,389 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9196, 2.1858, 2.4349, 3.1622, 2.2770, 2.3623, 2.3811, 2.2920], device='cuda:2'), covar=tensor([0.1419, 0.3169, 0.2601, 0.0785, 0.4076, 0.2480, 0.3149, 0.3169], device='cuda:2'), in_proj_covar=tensor([0.0409, 0.0456, 0.0375, 0.0335, 0.0441, 0.0526, 0.0428, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 14:55:03,885 INFO [train.py:904] (2/8) Epoch 23, batch 3550, loss[loss=0.188, simple_loss=0.2649, pruned_loss=0.05554, over 16452.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2541, pruned_loss=0.04052, over 3311006.64 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 16.0 2023-05-01 14:55:19,583 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-01 14:55:38,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2490, 2.7002, 2.6002, 5.0492, 3.7777, 4.1379, 1.8144, 2.9009], device='cuda:2'), covar=tensor([0.1133, 0.0887, 0.1326, 0.0169, 0.0256, 0.0562, 0.1487, 0.0984], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0194, 0.0206, 0.0218, 0.0204, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:55:43,978 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3580, 4.3185, 4.3017, 4.0125, 4.0725, 4.3566, 4.0104, 4.1564], device='cuda:2'), covar=tensor([0.0689, 0.0904, 0.0319, 0.0296, 0.0780, 0.0533, 0.0806, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0314, 0.0465, 0.0369, 0.0368, 0.0375, 0.0428, 0.0251, 0.0442], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 14:55:52,846 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:56:01,163 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 14:56:12,760 INFO [train.py:904] (2/8) Epoch 23, batch 3600, loss[loss=0.1642, simple_loss=0.2476, pruned_loss=0.04042, over 16821.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2535, pruned_loss=0.04007, over 3321070.50 frames. ], batch size: 102, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:56:28,214 INFO [optim.py:368] (2/8) 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 14:56:52,906 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1738, 2.8506, 2.7575, 4.9781, 3.7778, 4.2614, 1.8982, 3.0520], device='cuda:2'), covar=tensor([0.1213, 0.0836, 0.1170, 0.0235, 0.0232, 0.0495, 0.1516, 0.0831], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0194, 0.0206, 0.0218, 0.0204, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 14:57:02,339 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226937.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:57:24,672 INFO [train.py:904] (2/8) Epoch 23, batch 3650, loss[loss=0.1504, simple_loss=0.2233, pruned_loss=0.03877, over 16413.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2526, pruned_loss=0.04078, over 3312536.27 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:19,539 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:58:33,189 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 14:58:40,563 INFO [train.py:904] (2/8) Epoch 23, batch 3700, loss[loss=0.1813, simple_loss=0.2588, pruned_loss=0.05193, over 11448.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.251, pruned_loss=0.04189, over 3292480.77 frames. ], batch size: 248, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:56,577 INFO [optim.py:368] (2/8) 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 14:59:50,710 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227050.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:59:53,686 INFO [train.py:904] (2/8) Epoch 23, batch 3750, loss[loss=0.2028, simple_loss=0.2958, pruned_loss=0.05492, over 17038.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2515, pruned_loss=0.04303, over 3267296.16 frames. ], batch size: 53, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:59:57,457 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 15:00:03,104 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7164, 4.0364, 4.2111, 4.1771, 4.2116, 3.9763, 3.7478, 3.9517], device='cuda:2'), covar=tensor([0.0684, 0.0856, 0.0592, 0.0640, 0.0790, 0.0679, 0.1366, 0.0761], device='cuda:2'), in_proj_covar=tensor([0.0430, 0.0481, 0.0466, 0.0431, 0.0512, 0.0487, 0.0575, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 15:00:20,913 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 15:00:42,494 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0337, 5.3713, 5.1571, 5.1342, 4.8682, 4.7838, 4.8139, 5.4787], device='cuda:2'), covar=tensor([0.1260, 0.0838, 0.1042, 0.0901, 0.0824, 0.0959, 0.1123, 0.0813], device='cuda:2'), in_proj_covar=tensor([0.0699, 0.0849, 0.0700, 0.0647, 0.0537, 0.0541, 0.0713, 0.0665], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:01:07,855 INFO [train.py:904] (2/8) Epoch 23, batch 3800, loss[loss=0.1946, simple_loss=0.2755, pruned_loss=0.05681, over 12483.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2522, pruned_loss=0.04421, over 3268842.84 frames. ], batch size: 246, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:01:09,779 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2023-05-01 15:01:25,257 INFO [optim.py:368] (2/8) 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 15:02:21,618 INFO [train.py:904] (2/8) Epoch 23, batch 3850, loss[loss=0.1801, simple_loss=0.2562, pruned_loss=0.05203, over 15749.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2527, pruned_loss=0.04468, over 3264024.22 frames. ], batch size: 191, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:34,968 INFO [train.py:904] (2/8) Epoch 23, batch 3900, loss[loss=0.1684, simple_loss=0.24, pruned_loss=0.04837, over 16736.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2525, pruned_loss=0.045, over 3261765.54 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:51,651 INFO [optim.py:368] (2/8) 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 15:03:58,866 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5126, 3.3778, 3.8460, 1.8637, 3.8688, 3.8910, 3.1681, 2.8894], device='cuda:2'), covar=tensor([0.0783, 0.0249, 0.0155, 0.1222, 0.0101, 0.0208, 0.0356, 0.0444], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0140, 0.0083, 0.0129, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:04:47,721 INFO [train.py:904] (2/8) Epoch 23, batch 3950, loss[loss=0.1754, simple_loss=0.2532, pruned_loss=0.04881, over 16423.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2524, pruned_loss=0.04611, over 3267968.50 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:00,627 INFO [train.py:904] (2/8) Epoch 23, batch 4000, loss[loss=0.1816, simple_loss=0.2552, pruned_loss=0.05398, over 16822.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2525, pruned_loss=0.0466, over 3267570.84 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:17,147 INFO [optim.py:368] (2/8) 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 15:07:01,805 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:07:13,384 INFO [train.py:904] (2/8) Epoch 23, batch 4050, loss[loss=0.1729, simple_loss=0.2557, pruned_loss=0.0451, over 16618.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2533, pruned_loss=0.04602, over 3260126.96 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:00,695 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 15:08:05,230 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 15:08:27,045 INFO [train.py:904] (2/8) Epoch 23, batch 4100, loss[loss=0.2025, simple_loss=0.2933, pruned_loss=0.05584, over 15378.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2549, pruned_loss=0.04534, over 3254982.99 frames. ], batch size: 191, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:39,119 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227411.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:08:39,578 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 15:08:42,769 INFO [optim.py:368] (2/8) 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 15:09:17,471 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7519, 1.9047, 2.4703, 2.7332, 2.6406, 3.0772, 2.0738, 3.0332], device='cuda:2'), covar=tensor([0.0247, 0.0521, 0.0337, 0.0343, 0.0353, 0.0210, 0.0546, 0.0167], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0195, 0.0183, 0.0190, 0.0203, 0.0159, 0.0200, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:09:45,381 INFO [train.py:904] (2/8) Epoch 23, batch 4150, loss[loss=0.1736, simple_loss=0.272, pruned_loss=0.0376, over 16900.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2616, pruned_loss=0.04724, over 3252921.02 frames. ], batch size: 96, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:09:53,185 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0572, 2.2007, 2.2798, 3.6794, 2.1508, 2.5576, 2.3255, 2.3686], device='cuda:2'), covar=tensor([0.1459, 0.3583, 0.2832, 0.0609, 0.4049, 0.2499, 0.3559, 0.3346], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0459, 0.0374, 0.0334, 0.0440, 0.0530, 0.0429, 0.0536], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:10:03,682 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8084, 2.8724, 2.5235, 4.5339, 3.3550, 3.9453, 1.7193, 2.9864], device='cuda:2'), covar=tensor([0.1353, 0.0735, 0.1295, 0.0164, 0.0257, 0.0429, 0.1643, 0.0861], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0176, 0.0196, 0.0194, 0.0207, 0.0216, 0.0203, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:10:16,209 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227472.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:11:03,933 INFO [train.py:904] (2/8) Epoch 23, batch 4200, loss[loss=0.2148, simple_loss=0.3038, pruned_loss=0.06288, over 16492.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2684, pruned_loss=0.04876, over 3231200.63 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:11:20,463 INFO [optim.py:368] (2/8) 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 15:12:19,944 INFO [train.py:904] (2/8) Epoch 23, batch 4250, loss[loss=0.1766, simple_loss=0.2661, pruned_loss=0.04351, over 16719.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2716, pruned_loss=0.04874, over 3218321.38 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:12:49,366 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3186, 3.1686, 3.5858, 1.7668, 3.7166, 3.7216, 2.8713, 2.7576], device='cuda:2'), covar=tensor([0.0884, 0.0303, 0.0221, 0.1270, 0.0090, 0.0179, 0.0462, 0.0530], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0138, 0.0082, 0.0127, 0.0128, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:13:20,703 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 15:13:36,401 INFO [train.py:904] (2/8) Epoch 23, batch 4300, loss[loss=0.2081, simple_loss=0.2951, pruned_loss=0.06056, over 16689.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2733, pruned_loss=0.04805, over 3223162.64 frames. ], batch size: 62, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:13:55,081 INFO [optim.py:368] (2/8) 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 15:14:42,066 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=227645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:14:53,544 INFO [train.py:904] (2/8) Epoch 23, batch 4350, loss[loss=0.1699, simple_loss=0.2654, pruned_loss=0.03724, over 16903.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2758, pruned_loss=0.04882, over 3213779.09 frames. ], batch size: 96, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:15:55,219 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:15:55,413 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:16:09,816 INFO [train.py:904] (2/8) Epoch 23, batch 4400, loss[loss=0.2153, simple_loss=0.297, pruned_loss=0.06682, over 12009.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2776, pruned_loss=0.04974, over 3208921.46 frames. ], batch size: 248, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:16:27,160 INFO [optim.py:368] (2/8) 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 15:17:20,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1635, 3.4247, 3.5805, 2.0471, 3.0325, 2.1902, 3.4899, 3.5911], device='cuda:2'), covar=tensor([0.0195, 0.0692, 0.0520, 0.2102, 0.0790, 0.1051, 0.0599, 0.0860], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0167, 0.0168, 0.0153, 0.0145, 0.0129, 0.0144, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:17:22,034 INFO [train.py:904] (2/8) Epoch 23, batch 4450, loss[loss=0.1904, simple_loss=0.2774, pruned_loss=0.05169, over 17131.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2813, pruned_loss=0.05146, over 3212129.99 frames. ], batch size: 48, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:17:23,591 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:17:25,384 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8309, 3.8990, 2.4736, 4.8297, 3.1142, 4.7262, 2.7180, 3.1978], device='cuda:2'), covar=tensor([0.0301, 0.0378, 0.1609, 0.0123, 0.0822, 0.0369, 0.1324, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0176, 0.0193, 0.0165, 0.0176, 0.0217, 0.0200, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:17:41,686 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:18:35,087 INFO [train.py:904] (2/8) Epoch 23, batch 4500, loss[loss=0.2077, simple_loss=0.2854, pruned_loss=0.06504, over 11790.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.282, pruned_loss=0.05238, over 3204042.37 frames. ], batch size: 248, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:18:48,613 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9265, 3.6950, 4.3064, 2.1620, 4.5318, 4.5771, 3.1713, 3.4861], device='cuda:2'), covar=tensor([0.0760, 0.0297, 0.0202, 0.1116, 0.0067, 0.0112, 0.0468, 0.0399], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0139, 0.0082, 0.0127, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:18:52,359 INFO [optim.py:368] (2/8) 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 15:19:11,655 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:19:22,777 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0335, 5.3162, 5.1179, 5.1333, 4.8677, 4.6840, 4.7219, 5.4384], device='cuda:2'), covar=tensor([0.1266, 0.0849, 0.0969, 0.0757, 0.0786, 0.0956, 0.1113, 0.0887], device='cuda:2'), in_proj_covar=tensor([0.0685, 0.0831, 0.0685, 0.0632, 0.0525, 0.0533, 0.0697, 0.0653], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:19:48,151 INFO [train.py:904] (2/8) Epoch 23, batch 4550, loss[loss=0.2072, simple_loss=0.2983, pruned_loss=0.05803, over 16794.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2828, pruned_loss=0.05301, over 3231622.03 frames. ], batch size: 83, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:20:39,327 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227888.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:21:00,779 INFO [train.py:904] (2/8) Epoch 23, batch 4600, loss[loss=0.2214, simple_loss=0.2928, pruned_loss=0.07501, over 11242.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2836, pruned_loss=0.05353, over 3211302.70 frames. ], batch size: 246, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:21:18,796 INFO [optim.py:368] (2/8) 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 15:21:31,945 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 15:21:46,134 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7222, 3.8909, 2.8082, 2.3545, 2.6730, 2.4977, 4.2282, 3.4710], device='cuda:2'), covar=tensor([0.2755, 0.0596, 0.1884, 0.2613, 0.2525, 0.2108, 0.0469, 0.1222], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0271, 0.0308, 0.0318, 0.0302, 0.0265, 0.0300, 0.0344], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 15:22:14,193 INFO [train.py:904] (2/8) Epoch 23, batch 4650, loss[loss=0.1952, simple_loss=0.276, pruned_loss=0.05715, over 11456.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2832, pruned_loss=0.05392, over 3192322.83 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:22:16,335 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8265, 3.6337, 4.1834, 1.9885, 4.4090, 4.3438, 3.1212, 3.3356], device='cuda:2'), covar=tensor([0.0750, 0.0262, 0.0171, 0.1226, 0.0060, 0.0120, 0.0454, 0.0419], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0139, 0.0082, 0.0127, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:23:28,874 INFO [train.py:904] (2/8) Epoch 23, batch 4700, loss[loss=0.1723, simple_loss=0.2636, pruned_loss=0.04049, over 16668.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2801, pruned_loss=0.05276, over 3178987.62 frames. ], batch size: 76, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:34,899 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:23:45,412 INFO [optim.py:368] (2/8) 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 15:24:31,111 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-01 15:24:34,848 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 23, batch 4750, loss[loss=0.153, simple_loss=0.2463, pruned_loss=0.02983, over 16995.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2761, pruned_loss=0.05063, over 3191824.97 frames. ], batch size: 55, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:25:02,191 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:03,429 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:19,507 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228080.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:52,148 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6292, 4.7180, 4.5315, 4.1836, 4.1061, 4.6163, 4.3812, 4.3207], device='cuda:2'), covar=tensor([0.0648, 0.0480, 0.0322, 0.0333, 0.1094, 0.0569, 0.0549, 0.0685], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0441, 0.0349, 0.0348, 0.0354, 0.0405, 0.0239, 0.0416], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:25:54,062 INFO [train.py:904] (2/8) Epoch 23, batch 4800, loss[loss=0.1523, simple_loss=0.2516, pruned_loss=0.02649, over 16725.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2723, pruned_loss=0.04838, over 3196482.68 frames. ], batch size: 89, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:26:10,773 INFO [optim.py:368] (2/8) 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 15:26:11,946 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228115.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:26:50,648 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228141.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:27:07,819 INFO [train.py:904] (2/8) Epoch 23, batch 4850, loss[loss=0.1889, simple_loss=0.2809, pruned_loss=0.04848, over 16905.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.273, pruned_loss=0.0476, over 3186557.46 frames. ], batch size: 109, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:27:10,138 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7849, 2.7889, 2.7005, 1.8202, 2.5723, 2.7639, 2.5939, 1.8877], device='cuda:2'), covar=tensor([0.0505, 0.0085, 0.0083, 0.0420, 0.0125, 0.0132, 0.0143, 0.0446], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0099, 0.0112, 0.0096, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 15:27:29,734 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4231, 4.4076, 4.3434, 3.4884, 4.3356, 1.7377, 4.0896, 3.8640], device='cuda:2'), covar=tensor([0.0103, 0.0111, 0.0153, 0.0419, 0.0096, 0.2703, 0.0137, 0.0284], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0161, 0.0202, 0.0179, 0.0179, 0.0208, 0.0190, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:27:54,516 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228183.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:27:54,896 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 15:28:00,521 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6637, 4.5293, 4.7639, 4.8954, 5.0882, 4.5793, 5.0730, 5.0831], device='cuda:2'), covar=tensor([0.1882, 0.1330, 0.1454, 0.0692, 0.0464, 0.0957, 0.0527, 0.0561], device='cuda:2'), in_proj_covar=tensor([0.0642, 0.0796, 0.0918, 0.0803, 0.0611, 0.0639, 0.0662, 0.0766], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:28:23,601 INFO [train.py:904] (2/8) Epoch 23, batch 4900, loss[loss=0.18, simple_loss=0.2758, pruned_loss=0.04208, over 16741.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2724, pruned_loss=0.04636, over 3175625.04 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:28:42,122 INFO [optim.py:368] (2/8) 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 15:29:36,711 INFO [train.py:904] (2/8) Epoch 23, batch 4950, loss[loss=0.2197, simple_loss=0.3012, pruned_loss=0.06911, over 12356.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2715, pruned_loss=0.0454, over 3182564.65 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:29:47,066 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9597, 3.2657, 3.4016, 1.9014, 2.7654, 2.2940, 3.4863, 3.4542], device='cuda:2'), covar=tensor([0.0286, 0.0767, 0.0646, 0.2117, 0.0929, 0.0978, 0.0575, 0.0924], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0166, 0.0167, 0.0153, 0.0145, 0.0129, 0.0143, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:30:51,758 INFO [train.py:904] (2/8) Epoch 23, batch 5000, loss[loss=0.1909, simple_loss=0.2937, pruned_loss=0.04404, over 16300.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2727, pruned_loss=0.04545, over 3186363.50 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:31:09,997 INFO [optim.py:368] (2/8) 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 15:31:30,347 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:36,819 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:58,785 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:04,592 INFO [train.py:904] (2/8) Epoch 23, batch 5050, loss[loss=0.1641, simple_loss=0.2597, pruned_loss=0.03423, over 17106.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2729, pruned_loss=0.04506, over 3201633.43 frames. ], batch size: 49, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:32:06,135 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8854, 3.9242, 4.1410, 4.1059, 4.1274, 3.9555, 3.9190, 3.8944], device='cuda:2'), covar=tensor([0.0305, 0.0568, 0.0384, 0.0448, 0.0415, 0.0370, 0.0736, 0.0490], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0450, 0.0437, 0.0407, 0.0483, 0.0457, 0.0540, 0.0367], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 15:32:19,781 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228363.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:40,864 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 15:32:58,965 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228390.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:04,540 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228394.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:09,565 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:17,239 INFO [train.py:904] (2/8) Epoch 23, batch 5100, loss[loss=0.1613, simple_loss=0.2525, pruned_loss=0.03501, over 16670.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2715, pruned_loss=0.04433, over 3199852.39 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:33:34,756 INFO [optim.py:368] (2/8) 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 15:33:55,487 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228429.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:34:06,469 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 23, batch 5150, loss[loss=0.1874, simple_loss=0.2778, pruned_loss=0.04854, over 12146.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2719, pruned_loss=0.04405, over 3201518.12 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:34:54,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3081, 2.5715, 2.0958, 2.3770, 2.9295, 2.6156, 2.9116, 3.1513], device='cuda:2'), covar=tensor([0.0129, 0.0490, 0.0590, 0.0471, 0.0265, 0.0375, 0.0239, 0.0249], device='cuda:2'), in_proj_covar=tensor([0.0217, 0.0238, 0.0228, 0.0231, 0.0239, 0.0238, 0.0239, 0.0236], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:34:57,760 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9009, 3.3598, 3.2783, 2.0960, 3.1061, 3.3797, 3.1042, 1.5592], device='cuda:2'), covar=tensor([0.0715, 0.0094, 0.0106, 0.0566, 0.0146, 0.0196, 0.0224, 0.0796], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0100, 0.0112, 0.0096, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 15:35:15,460 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:35:26,002 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228490.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:35:38,105 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9191, 2.2987, 2.2668, 2.7518, 1.8991, 3.2168, 1.7305, 2.7377], device='cuda:2'), covar=tensor([0.1241, 0.0716, 0.1154, 0.0149, 0.0106, 0.0342, 0.1566, 0.0716], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0176, 0.0196, 0.0192, 0.0205, 0.0215, 0.0204, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:35:44,364 INFO [train.py:904] (2/8) Epoch 23, batch 5200, loss[loss=0.1994, simple_loss=0.2799, pruned_loss=0.05943, over 12375.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2703, pruned_loss=0.0436, over 3204275.61 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:36:01,315 INFO [optim.py:368] (2/8) 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 15:36:25,654 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228531.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:36:57,627 INFO [train.py:904] (2/8) Epoch 23, batch 5250, loss[loss=0.1894, simple_loss=0.2678, pruned_loss=0.05551, over 12482.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2678, pruned_loss=0.04324, over 3210332.40 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:37:59,484 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7889, 4.0940, 3.1075, 2.4886, 2.7746, 2.6934, 4.3542, 3.5522], device='cuda:2'), covar=tensor([0.2948, 0.0608, 0.1766, 0.2885, 0.2674, 0.1880, 0.0474, 0.1373], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0271, 0.0308, 0.0316, 0.0300, 0.0263, 0.0300, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 15:38:10,576 INFO [train.py:904] (2/8) Epoch 23, batch 5300, loss[loss=0.1844, simple_loss=0.2668, pruned_loss=0.05102, over 12779.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2645, pruned_loss=0.04231, over 3211358.81 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:38:28,436 INFO [optim.py:368] (2/8) 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 15:38:28,883 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2564, 5.2253, 5.1246, 4.3709, 5.1638, 1.8115, 4.9178, 5.0172], device='cuda:2'), covar=tensor([0.0094, 0.0091, 0.0184, 0.0508, 0.0118, 0.2705, 0.0142, 0.0198], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0160, 0.0199, 0.0177, 0.0178, 0.0206, 0.0188, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:39:23,394 INFO [train.py:904] (2/8) Epoch 23, batch 5350, loss[loss=0.1799, simple_loss=0.2609, pruned_loss=0.04949, over 12335.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2629, pruned_loss=0.04179, over 3215797.96 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:39:38,458 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:39:48,371 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7160, 4.8788, 5.0484, 4.8488, 4.8731, 5.4139, 4.8778, 4.5620], device='cuda:2'), covar=tensor([0.1167, 0.1667, 0.1834, 0.1830, 0.2601, 0.0876, 0.1482, 0.2696], device='cuda:2'), in_proj_covar=tensor([0.0410, 0.0590, 0.0647, 0.0485, 0.0649, 0.0678, 0.0508, 0.0655], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 15:40:10,171 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228685.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:16,735 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228689.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:36,803 INFO [train.py:904] (2/8) Epoch 23, batch 5400, loss[loss=0.1873, simple_loss=0.2796, pruned_loss=0.04753, over 16507.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2658, pruned_loss=0.04264, over 3211459.97 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:40:48,793 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:54,347 INFO [optim.py:368] (2/8) 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 15:41:26,140 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1962, 3.7605, 3.7459, 2.4072, 3.3498, 3.7647, 3.4247, 2.0481], device='cuda:2'), covar=tensor([0.0559, 0.0055, 0.0048, 0.0425, 0.0108, 0.0090, 0.0108, 0.0503], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0084, 0.0086, 0.0134, 0.0099, 0.0110, 0.0096, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 15:41:27,291 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:41:28,745 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1693, 2.4345, 1.9721, 2.1934, 2.7649, 2.4191, 2.7966, 2.9863], device='cuda:2'), covar=tensor([0.0142, 0.0449, 0.0595, 0.0522, 0.0290, 0.0433, 0.0206, 0.0272], device='cuda:2'), in_proj_covar=tensor([0.0217, 0.0239, 0.0229, 0.0231, 0.0239, 0.0239, 0.0238, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:41:54,046 INFO [train.py:904] (2/8) Epoch 23, batch 5450, loss[loss=0.2178, simple_loss=0.299, pruned_loss=0.06833, over 12007.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2686, pruned_loss=0.04414, over 3199147.97 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:41:57,281 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 15:42:42,358 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228784.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:42:44,354 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228785.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:43:12,591 INFO [train.py:904] (2/8) Epoch 23, batch 5500, loss[loss=0.1925, simple_loss=0.2807, pruned_loss=0.05215, over 17100.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2758, pruned_loss=0.04833, over 3175358.81 frames. ], batch size: 49, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:43:32,425 INFO [optim.py:368] (2/8) 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 15:44:13,404 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4929, 3.4434, 3.4698, 2.7921, 3.3468, 2.1688, 3.1884, 2.8564], device='cuda:2'), covar=tensor([0.0167, 0.0145, 0.0174, 0.0235, 0.0109, 0.2156, 0.0138, 0.0237], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0177, 0.0206, 0.0189, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:44:31,569 INFO [train.py:904] (2/8) Epoch 23, batch 5550, loss[loss=0.3012, simple_loss=0.3531, pruned_loss=0.1246, over 10671.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2837, pruned_loss=0.05427, over 3126166.78 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:45:24,004 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3087, 4.3586, 4.2226, 3.9106, 3.9040, 4.2844, 4.0485, 4.0944], device='cuda:2'), covar=tensor([0.0640, 0.0621, 0.0310, 0.0334, 0.0838, 0.0514, 0.0736, 0.0603], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0440, 0.0347, 0.0345, 0.0353, 0.0404, 0.0236, 0.0413], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:45:52,970 INFO [train.py:904] (2/8) Epoch 23, batch 5600, loss[loss=0.1919, simple_loss=0.2764, pruned_loss=0.05373, over 16565.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2881, pruned_loss=0.05768, over 3105718.89 frames. ], batch size: 62, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:46:10,854 INFO [optim.py:368] (2/8) 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 15:46:31,345 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-01 15:46:45,604 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:47:13,516 INFO [train.py:904] (2/8) Epoch 23, batch 5650, loss[loss=0.2074, simple_loss=0.2927, pruned_loss=0.06107, over 16364.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2933, pruned_loss=0.06213, over 3058707.00 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:47:58,454 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 15:48:04,712 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228985.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:10,644 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:11,052 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-01 15:48:21,855 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228996.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:32,719 INFO [train.py:904] (2/8) Epoch 23, batch 5700, loss[loss=0.2207, simple_loss=0.3102, pruned_loss=0.06555, over 16547.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2948, pruned_loss=0.06375, over 3040801.36 frames. ], batch size: 62, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:48:42,058 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3372, 3.0227, 3.3801, 1.6939, 3.5460, 3.5872, 2.8136, 2.6900], device='cuda:2'), covar=tensor([0.0812, 0.0312, 0.0229, 0.1331, 0.0098, 0.0186, 0.0465, 0.0513], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0139, 0.0082, 0.0127, 0.0128, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:48:50,701 INFO [optim.py:368] (2/8) 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 15:48:52,581 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8718, 2.7710, 2.8337, 2.1786, 2.6646, 2.1424, 2.7274, 2.8968], device='cuda:2'), covar=tensor([0.0296, 0.0744, 0.0500, 0.1645, 0.0788, 0.0930, 0.0565, 0.0709], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0153, 0.0146, 0.0130, 0.0143, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:49:18,828 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229033.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:49:24,806 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:49:50,897 INFO [train.py:904] (2/8) Epoch 23, batch 5750, loss[loss=0.221, simple_loss=0.3064, pruned_loss=0.06775, over 15240.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2978, pruned_loss=0.06562, over 3003508.48 frames. ], batch size: 191, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:49:59,825 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1131, 2.4011, 2.6046, 1.9960, 2.7078, 2.7785, 2.4671, 2.4142], device='cuda:2'), covar=tensor([0.0637, 0.0232, 0.0233, 0.0833, 0.0123, 0.0276, 0.0403, 0.0440], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0098, 0.0139, 0.0082, 0.0126, 0.0128, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 15:50:44,094 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229085.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:50:44,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9347, 2.1717, 2.4622, 3.1422, 2.2995, 2.4251, 2.3826, 2.3306], device='cuda:2'), covar=tensor([0.1425, 0.3583, 0.2373, 0.0731, 0.3872, 0.2366, 0.3353, 0.3141], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0448, 0.0367, 0.0325, 0.0432, 0.0516, 0.0419, 0.0524], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:50:58,595 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3374, 3.8514, 3.8589, 2.5597, 3.5651, 3.9179, 3.6014, 2.1576], device='cuda:2'), covar=tensor([0.0557, 0.0067, 0.0066, 0.0431, 0.0103, 0.0107, 0.0091, 0.0478], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0084, 0.0085, 0.0134, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 15:51:13,969 INFO [train.py:904] (2/8) Epoch 23, batch 5800, loss[loss=0.1892, simple_loss=0.2843, pruned_loss=0.04705, over 16269.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.297, pruned_loss=0.06469, over 2989291.74 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:51:32,213 INFO [optim.py:368] (2/8) 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 15:52:01,459 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229133.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:52:01,551 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5361, 3.5893, 3.3663, 3.0237, 3.1783, 3.5076, 3.3088, 3.3138], device='cuda:2'), covar=tensor([0.0578, 0.0621, 0.0297, 0.0264, 0.0530, 0.0422, 0.1520, 0.0457], device='cuda:2'), in_proj_covar=tensor([0.0295, 0.0438, 0.0345, 0.0343, 0.0352, 0.0401, 0.0235, 0.0411], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:52:31,388 INFO [train.py:904] (2/8) Epoch 23, batch 5850, loss[loss=0.2038, simple_loss=0.2771, pruned_loss=0.06522, over 11385.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.295, pruned_loss=0.06261, over 3018729.03 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:53:53,138 INFO [train.py:904] (2/8) Epoch 23, batch 5900, loss[loss=0.2027, simple_loss=0.2986, pruned_loss=0.0534, over 17053.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2945, pruned_loss=0.0622, over 3034799.36 frames. ], batch size: 41, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:54:16,064 INFO [optim.py:368] (2/8) 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 15:55:17,140 INFO [train.py:904] (2/8) Epoch 23, batch 5950, loss[loss=0.1877, simple_loss=0.2798, pruned_loss=0.04783, over 16698.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2948, pruned_loss=0.06049, over 3049979.88 frames. ], batch size: 57, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:55:20,357 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5567, 2.6652, 2.1694, 2.5130, 3.0506, 2.6644, 3.1925, 3.2261], device='cuda:2'), covar=tensor([0.0124, 0.0433, 0.0573, 0.0455, 0.0263, 0.0391, 0.0222, 0.0258], device='cuda:2'), in_proj_covar=tensor([0.0215, 0.0237, 0.0227, 0.0229, 0.0237, 0.0236, 0.0236, 0.0234], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 15:56:18,150 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229291.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:33,809 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:37,756 INFO [train.py:904] (2/8) Epoch 23, batch 6000, loss[loss=0.1976, simple_loss=0.2819, pruned_loss=0.05664, over 16723.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2936, pruned_loss=0.05941, over 3096002.86 frames. ], batch size: 124, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:56:37,756 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 15:56:49,487 INFO [train.py:938] (2/8) Epoch 23, validation: loss=0.1497, simple_loss=0.2623, pruned_loss=0.01859, over 944034.00 frames. 2023-05-01 15:56:49,488 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 15:57:07,746 INFO [optim.py:368] (2/8) 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 15:57:26,306 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-01 15:57:40,012 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-05-01 15:58:03,285 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 15:58:06,099 INFO [train.py:904] (2/8) Epoch 23, batch 6050, loss[loss=0.2057, simple_loss=0.2766, pruned_loss=0.06738, over 11559.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2921, pruned_loss=0.05856, over 3118264.57 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:58:20,415 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:58:30,369 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 15:59:05,204 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 15:59:19,813 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 15:59:21,989 INFO [train.py:904] (2/8) Epoch 23, batch 6100, loss[loss=0.1896, simple_loss=0.2806, pruned_loss=0.04927, over 16686.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2918, pruned_loss=0.0574, over 3137346.75 frames. ], batch size: 83, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:59:40,506 INFO [optim.py:368] (2/8) 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 16:00:22,341 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229443.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:00:36,341 INFO [train.py:904] (2/8) Epoch 23, batch 6150, loss[loss=0.1955, simple_loss=0.2805, pruned_loss=0.05529, over 16776.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2896, pruned_loss=0.05652, over 3130001.34 frames. ], batch size: 62, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 16:01:15,189 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229477.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:01:53,538 INFO [train.py:904] (2/8) Epoch 23, batch 6200, loss[loss=0.1706, simple_loss=0.2649, pruned_loss=0.03812, over 16765.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2877, pruned_loss=0.05599, over 3137274.18 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:01:55,793 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229504.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:02:14,506 INFO [optim.py:368] (2/8) 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 16:02:49,411 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:03:11,715 INFO [train.py:904] (2/8) Epoch 23, batch 6250, loss[loss=0.1854, simple_loss=0.287, pruned_loss=0.04187, over 16761.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2876, pruned_loss=0.05657, over 3109260.03 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:03:17,475 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5871, 3.6113, 2.1679, 3.9909, 2.7171, 3.9218, 2.3720, 2.8856], device='cuda:2'), covar=tensor([0.0233, 0.0349, 0.1593, 0.0245, 0.0796, 0.0700, 0.1417, 0.0747], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0165, 0.0177, 0.0217, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:03:22,967 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8291, 3.2153, 3.3890, 1.9604, 2.9325, 2.3090, 3.4024, 3.4485], device='cuda:2'), covar=tensor([0.0252, 0.0765, 0.0587, 0.2129, 0.0818, 0.0974, 0.0572, 0.0872], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0154, 0.0147, 0.0131, 0.0144, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:03:48,903 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:09,692 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:28,580 INFO [train.py:904] (2/8) Epoch 23, batch 6300, loss[loss=0.1973, simple_loss=0.2888, pruned_loss=0.05286, over 16813.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2867, pruned_loss=0.0557, over 3116097.21 frames. ], batch size: 102, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:04:29,033 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9471, 4.9394, 4.7546, 4.0516, 4.8430, 1.8914, 4.6085, 4.4842], device='cuda:2'), covar=tensor([0.0089, 0.0077, 0.0183, 0.0369, 0.0089, 0.2607, 0.0133, 0.0219], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0160, 0.0201, 0.0179, 0.0178, 0.0208, 0.0189, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:04:50,641 INFO [optim.py:368] (2/8) 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 16:05:27,311 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:42,104 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7306, 3.9216, 2.9968, 2.3726, 2.6683, 2.4916, 4.1898, 3.5185], device='cuda:2'), covar=tensor([0.2922, 0.0623, 0.1832, 0.2757, 0.2655, 0.2073, 0.0453, 0.1309], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0270, 0.0306, 0.0316, 0.0298, 0.0262, 0.0299, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 16:05:48,007 INFO [train.py:904] (2/8) Epoch 23, batch 6350, loss[loss=0.2261, simple_loss=0.296, pruned_loss=0.07809, over 11588.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2866, pruned_loss=0.05664, over 3092812.84 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:05:48,963 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 16:05:49,832 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0504, 3.9716, 4.1407, 4.2815, 4.3780, 3.9470, 4.3161, 4.3968], device='cuda:2'), covar=tensor([0.1918, 0.1223, 0.1474, 0.0720, 0.0668, 0.1449, 0.0869, 0.0789], device='cuda:2'), in_proj_covar=tensor([0.0635, 0.0787, 0.0906, 0.0792, 0.0604, 0.0631, 0.0653, 0.0758], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:05:53,056 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229656.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:57,427 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5385, 4.2509, 4.1968, 2.8480, 3.7901, 4.2341, 3.7479, 2.5077], device='cuda:2'), covar=tensor([0.0537, 0.0043, 0.0045, 0.0375, 0.0092, 0.0096, 0.0093, 0.0417], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0084, 0.0085, 0.0134, 0.0098, 0.0111, 0.0095, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 16:05:57,443 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229659.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:06:52,287 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9143, 1.9948, 2.5762, 2.8651, 2.7772, 3.3081, 2.1817, 3.1969], device='cuda:2'), covar=tensor([0.0226, 0.0551, 0.0314, 0.0310, 0.0334, 0.0168, 0.0541, 0.0169], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0194, 0.0180, 0.0185, 0.0200, 0.0156, 0.0199, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:07:04,876 INFO [train.py:904] (2/8) Epoch 23, batch 6400, loss[loss=0.1886, simple_loss=0.2665, pruned_loss=0.0554, over 17234.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2875, pruned_loss=0.05804, over 3090268.22 frames. ], batch size: 44, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:07:11,465 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7435, 3.7589, 3.8865, 3.7000, 3.8362, 4.2165, 3.8470, 3.5774], device='cuda:2'), covar=tensor([0.2269, 0.2209, 0.2511, 0.2266, 0.2784, 0.1985, 0.1805, 0.2655], device='cuda:2'), in_proj_covar=tensor([0.0414, 0.0598, 0.0659, 0.0492, 0.0655, 0.0690, 0.0515, 0.0662], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 16:07:14,782 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1102, 5.0856, 4.9629, 4.5723, 4.6082, 5.0130, 4.8896, 4.7176], device='cuda:2'), covar=tensor([0.0559, 0.0625, 0.0288, 0.0344, 0.1017, 0.0480, 0.0337, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0437, 0.0343, 0.0339, 0.0346, 0.0398, 0.0233, 0.0409], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:07:24,870 INFO [optim.py:368] (2/8) 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 16:07:32,376 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229720.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:07:48,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7408, 4.7362, 4.5908, 3.8020, 4.6573, 1.6348, 4.4242, 4.2295], device='cuda:2'), covar=tensor([0.0141, 0.0128, 0.0208, 0.0446, 0.0124, 0.3032, 0.0244, 0.0289], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0160, 0.0201, 0.0179, 0.0178, 0.0208, 0.0190, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:08:21,140 INFO [train.py:904] (2/8) Epoch 23, batch 6450, loss[loss=0.1911, simple_loss=0.2797, pruned_loss=0.0512, over 16765.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2877, pruned_loss=0.05747, over 3079214.50 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:08:37,751 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6842, 1.7884, 1.5766, 1.5204, 1.9702, 1.6321, 1.5538, 1.8842], device='cuda:2'), covar=tensor([0.0277, 0.0326, 0.0444, 0.0375, 0.0245, 0.0282, 0.0241, 0.0235], device='cuda:2'), in_proj_covar=tensor([0.0213, 0.0235, 0.0226, 0.0227, 0.0236, 0.0235, 0.0235, 0.0232], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:09:34,332 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229799.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:09:40,112 INFO [train.py:904] (2/8) Epoch 23, batch 6500, loss[loss=0.2256, simple_loss=0.3089, pruned_loss=0.07119, over 16722.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2863, pruned_loss=0.05693, over 3096147.97 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:09:59,296 INFO [optim.py:368] (2/8) 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 16:10:25,442 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:10:57,906 INFO [train.py:904] (2/8) Epoch 23, batch 6550, loss[loss=0.1838, simple_loss=0.2792, pruned_loss=0.04416, over 17193.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2881, pruned_loss=0.05699, over 3125430.88 frames. ], batch size: 44, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:10:58,919 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8155, 3.9046, 4.1323, 4.1199, 4.1125, 3.9077, 3.9033, 3.9298], device='cuda:2'), covar=tensor([0.0393, 0.0668, 0.0518, 0.0480, 0.0496, 0.0506, 0.0828, 0.0492], device='cuda:2'), in_proj_covar=tensor([0.0413, 0.0459, 0.0447, 0.0415, 0.0491, 0.0467, 0.0552, 0.0373], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 16:12:02,433 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229894.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:12:15,578 INFO [train.py:904] (2/8) Epoch 23, batch 6600, loss[loss=0.1854, simple_loss=0.2749, pruned_loss=0.04795, over 16963.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2903, pruned_loss=0.05717, over 3112533.48 frames. ], batch size: 55, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:32,101 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3189, 4.4095, 4.2337, 3.9920, 3.9555, 4.3416, 4.0002, 4.1001], device='cuda:2'), covar=tensor([0.0647, 0.0655, 0.0282, 0.0284, 0.0799, 0.0461, 0.0777, 0.0568], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0436, 0.0341, 0.0339, 0.0347, 0.0397, 0.0233, 0.0408], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:12:35,467 INFO [optim.py:368] (2/8) 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 16:12:36,350 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 16:12:40,549 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-01 16:13:01,780 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:33,860 INFO [train.py:904] (2/8) Epoch 23, batch 6650, loss[loss=0.2382, simple_loss=0.3048, pruned_loss=0.08581, over 11621.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2907, pruned_loss=0.05838, over 3100289.71 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:13:36,352 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9273, 4.9017, 4.7281, 3.9977, 4.8093, 1.7353, 4.5593, 4.4198], device='cuda:2'), covar=tensor([0.0082, 0.0082, 0.0193, 0.0412, 0.0098, 0.2901, 0.0124, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0177, 0.0207, 0.0189, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:13:37,565 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:38,786 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:26,027 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229987.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:52,371 INFO [train.py:904] (2/8) Epoch 23, batch 6700, loss[loss=0.2316, simple_loss=0.3011, pruned_loss=0.08106, over 11468.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2894, pruned_loss=0.05857, over 3105463.95 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:14:54,790 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230004.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:15:11,262 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230015.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:15:12,080 INFO [optim.py:368] (2/8) 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 16:15:41,053 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4986, 3.5573, 1.8753, 4.0485, 2.7170, 3.9693, 1.9969, 2.6812], device='cuda:2'), covar=tensor([0.0310, 0.0442, 0.2083, 0.0243, 0.0870, 0.0633, 0.1923, 0.0925], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0178, 0.0196, 0.0166, 0.0177, 0.0218, 0.0204, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:16:02,800 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:16:09,694 INFO [train.py:904] (2/8) Epoch 23, batch 6750, loss[loss=0.1945, simple_loss=0.285, pruned_loss=0.05196, over 15417.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2888, pruned_loss=0.05909, over 3083058.14 frames. ], batch size: 190, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:17:19,742 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230099.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:17:24,163 INFO [train.py:904] (2/8) Epoch 23, batch 6800, loss[loss=0.2293, simple_loss=0.3029, pruned_loss=0.07786, over 11632.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2888, pruned_loss=0.05892, over 3084220.91 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:17:43,642 INFO [optim.py:368] (2/8) 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 16:18:11,167 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:22,468 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4491, 3.4189, 3.4501, 2.4734, 3.3107, 2.0026, 3.0659, 2.7113], device='cuda:2'), covar=tensor([0.0239, 0.0225, 0.0267, 0.0462, 0.0168, 0.2920, 0.0214, 0.0358], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0176, 0.0207, 0.0188, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:18:31,927 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:36,501 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230150.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:40,518 INFO [train.py:904] (2/8) Epoch 23, batch 6850, loss[loss=0.2015, simple_loss=0.3047, pruned_loss=0.04917, over 16677.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2903, pruned_loss=0.05913, over 3089096.17 frames. ], batch size: 62, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:19:23,352 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:19:33,213 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5016, 2.2057, 2.9825, 3.3877, 3.0940, 3.8571, 2.3985, 3.8231], device='cuda:2'), covar=tensor([0.0167, 0.0526, 0.0328, 0.0256, 0.0310, 0.0133, 0.0572, 0.0153], device='cuda:2'), in_proj_covar=tensor([0.0190, 0.0193, 0.0179, 0.0185, 0.0199, 0.0155, 0.0198, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:19:56,853 INFO [train.py:904] (2/8) Epoch 23, batch 6900, loss[loss=0.2156, simple_loss=0.3068, pruned_loss=0.06218, over 17254.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2923, pruned_loss=0.05815, over 3105313.79 frames. ], batch size: 52, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:20:10,815 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:20,076 INFO [optim.py:368] (2/8) 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 16:20:45,514 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:58,977 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230241.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:13,098 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2104, 4.3689, 4.4638, 4.3224, 4.3035, 4.8338, 4.3987, 4.1362], device='cuda:2'), covar=tensor([0.1702, 0.1942, 0.2543, 0.1948, 0.2545, 0.1090, 0.1650, 0.2427], device='cuda:2'), in_proj_covar=tensor([0.0414, 0.0596, 0.0661, 0.0494, 0.0657, 0.0685, 0.0515, 0.0662], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 16:21:13,108 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:18,009 INFO [train.py:904] (2/8) Epoch 23, batch 6950, loss[loss=0.2046, simple_loss=0.2863, pruned_loss=0.06149, over 15410.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2936, pruned_loss=0.05953, over 3101250.73 frames. ], batch size: 191, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:21:24,023 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8235, 3.6073, 4.0767, 2.0967, 4.1866, 4.3073, 3.1208, 3.2025], device='cuda:2'), covar=tensor([0.0753, 0.0252, 0.0219, 0.1151, 0.0082, 0.0152, 0.0419, 0.0448], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0098, 0.0138, 0.0082, 0.0126, 0.0127, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:21:29,571 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 16:21:38,301 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 16:22:01,066 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230281.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:28,213 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 16:22:34,104 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230302.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,886 INFO [train.py:904] (2/8) Epoch 23, batch 7000, loss[loss=0.207, simple_loss=0.3006, pruned_loss=0.05673, over 16672.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2938, pruned_loss=0.05899, over 3096366.05 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:52,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4002, 3.2629, 2.6269, 2.1570, 2.2132, 2.2666, 3.3560, 2.9759], device='cuda:2'), covar=tensor([0.3117, 0.0780, 0.1867, 0.2819, 0.2600, 0.2247, 0.0561, 0.1413], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0268, 0.0305, 0.0314, 0.0296, 0.0260, 0.0298, 0.0338], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 16:22:53,484 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230315.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:22:56,093 INFO [optim.py:368] (2/8) 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 16:23:35,213 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:23:50,811 INFO [train.py:904] (2/8) Epoch 23, batch 7050, loss[loss=0.1957, simple_loss=0.2868, pruned_loss=0.05228, over 16586.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2943, pruned_loss=0.05831, over 3111899.26 frames. ], batch size: 62, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:24:05,842 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230363.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:24:20,090 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6558, 4.6430, 5.0096, 4.9649, 5.0294, 4.7026, 4.6912, 4.4916], device='cuda:2'), covar=tensor([0.0338, 0.0555, 0.0366, 0.0418, 0.0444, 0.0390, 0.0897, 0.0522], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0463, 0.0449, 0.0416, 0.0492, 0.0471, 0.0557, 0.0376], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 16:25:07,473 INFO [train.py:904] (2/8) Epoch 23, batch 7100, loss[loss=0.2208, simple_loss=0.3046, pruned_loss=0.06853, over 16873.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2925, pruned_loss=0.05841, over 3107737.83 frames. ], batch size: 109, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:25:30,796 INFO [optim.py:368] (2/8) 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 16:25:47,918 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0261, 2.2762, 2.3650, 2.6645, 1.8905, 3.1666, 1.8696, 2.7113], device='cuda:2'), covar=tensor([0.1238, 0.0670, 0.1130, 0.0196, 0.0125, 0.0379, 0.1543, 0.0735], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0178, 0.0200, 0.0195, 0.0209, 0.0219, 0.0207, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:26:17,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8344, 2.1358, 2.4506, 3.0850, 2.2008, 2.3028, 2.3497, 2.2343], device='cuda:2'), covar=tensor([0.1422, 0.3278, 0.2466, 0.0775, 0.4096, 0.2433, 0.2911, 0.3298], device='cuda:2'), in_proj_covar=tensor([0.0402, 0.0450, 0.0367, 0.0325, 0.0435, 0.0517, 0.0421, 0.0526], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:26:24,958 INFO [train.py:904] (2/8) Epoch 23, batch 7150, loss[loss=0.2286, simple_loss=0.3068, pruned_loss=0.07518, over 16397.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2908, pruned_loss=0.0585, over 3104908.89 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:26:49,313 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:27:42,211 INFO [train.py:904] (2/8) Epoch 23, batch 7200, loss[loss=0.1666, simple_loss=0.2611, pruned_loss=0.03604, over 16139.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2883, pruned_loss=0.05655, over 3121885.75 frames. ], batch size: 165, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:27:47,860 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230506.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:03,476 INFO [optim.py:368] (2/8) 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 16:28:16,806 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:16,934 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2822, 3.4591, 3.6452, 2.1136, 3.1802, 2.3860, 3.6330, 3.8139], device='cuda:2'), covar=tensor([0.0222, 0.0743, 0.0537, 0.1998, 0.0760, 0.0921, 0.0570, 0.0845], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:28:22,505 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230529.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:28:30,056 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:40,484 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1884, 2.8690, 3.0712, 1.7197, 3.1868, 3.3023, 2.6250, 2.5444], device='cuda:2'), covar=tensor([0.0864, 0.0323, 0.0225, 0.1299, 0.0110, 0.0220, 0.0523, 0.0513], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0099, 0.0138, 0.0082, 0.0125, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:28:40,503 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9706, 3.0153, 1.8865, 3.2482, 2.3004, 3.3173, 2.0734, 2.4822], device='cuda:2'), covar=tensor([0.0349, 0.0429, 0.1748, 0.0241, 0.0948, 0.0595, 0.1629, 0.0868], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0163, 0.0176, 0.0217, 0.0203, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:28:58,899 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2442, 4.2362, 4.1519, 3.3748, 4.1811, 1.6859, 3.9911, 3.7735], device='cuda:2'), covar=tensor([0.0113, 0.0101, 0.0184, 0.0322, 0.0093, 0.2927, 0.0127, 0.0260], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0159, 0.0199, 0.0177, 0.0175, 0.0206, 0.0188, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:28:58,903 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230550.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:02,529 INFO [train.py:904] (2/8) Epoch 23, batch 7250, loss[loss=0.2385, simple_loss=0.3039, pruned_loss=0.0865, over 11622.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2861, pruned_loss=0.0554, over 3122424.91 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:29:51,419 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:03,547 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:08,112 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230597.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:10,079 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230598.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:18,078 INFO [train.py:904] (2/8) Epoch 23, batch 7300, loss[loss=0.171, simple_loss=0.2726, pruned_loss=0.03473, over 16904.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2856, pruned_loss=0.05505, over 3118107.92 frames. ], batch size: 96, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:30:39,629 INFO [optim.py:368] (2/8) 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 16:30:57,526 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3480, 2.8831, 2.6293, 2.2365, 2.2374, 2.2995, 2.9202, 2.7862], device='cuda:2'), covar=tensor([0.2408, 0.0693, 0.1563, 0.2449, 0.2360, 0.2010, 0.0569, 0.1229], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0271, 0.0308, 0.0318, 0.0300, 0.0263, 0.0300, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 16:30:59,101 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 16:31:19,788 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230643.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:31:33,668 INFO [train.py:904] (2/8) Epoch 23, batch 7350, loss[loss=0.23, simple_loss=0.2982, pruned_loss=0.08091, over 11381.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.287, pruned_loss=0.05668, over 3063947.93 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:31:34,927 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1229, 2.1312, 2.7495, 3.0395, 2.8957, 3.6700, 2.2936, 3.5393], device='cuda:2'), covar=tensor([0.0220, 0.0541, 0.0291, 0.0312, 0.0312, 0.0138, 0.0532, 0.0135], device='cuda:2'), in_proj_covar=tensor([0.0190, 0.0193, 0.0179, 0.0185, 0.0199, 0.0155, 0.0197, 0.0154], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:31:51,344 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9237, 4.1711, 3.9719, 4.0250, 3.7308, 3.8142, 3.8472, 4.1600], device='cuda:2'), covar=tensor([0.1106, 0.0853, 0.1104, 0.0882, 0.0828, 0.1562, 0.1007, 0.0976], device='cuda:2'), in_proj_covar=tensor([0.0674, 0.0814, 0.0677, 0.0626, 0.0518, 0.0529, 0.0688, 0.0645], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:32:07,350 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 16:32:33,457 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 23, batch 7400, loss[loss=0.1968, simple_loss=0.2899, pruned_loss=0.05184, over 16703.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2879, pruned_loss=0.05706, over 3065946.31 frames. ], batch size: 134, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:33:03,507 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5131, 3.4756, 3.4771, 2.7489, 3.3135, 2.1510, 3.0665, 2.7877], device='cuda:2'), covar=tensor([0.0153, 0.0142, 0.0181, 0.0212, 0.0104, 0.2224, 0.0141, 0.0235], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0158, 0.0198, 0.0175, 0.0174, 0.0205, 0.0186, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:33:13,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4109, 3.4213, 1.9923, 3.8494, 2.5980, 3.8403, 2.1183, 2.6923], device='cuda:2'), covar=tensor([0.0298, 0.0407, 0.1815, 0.0251, 0.0864, 0.0635, 0.1696, 0.0900], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0176, 0.0195, 0.0164, 0.0176, 0.0217, 0.0203, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:33:16,062 INFO [optim.py:368] (2/8) 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 16:34:13,423 INFO [train.py:904] (2/8) Epoch 23, batch 7450, loss[loss=0.1754, simple_loss=0.2767, pruned_loss=0.03701, over 16840.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.289, pruned_loss=0.05825, over 3058965.73 frames. ], batch size: 102, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:34:48,165 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6479, 3.7116, 2.3603, 4.2366, 2.9162, 4.1427, 2.5341, 3.0013], device='cuda:2'), covar=tensor([0.0287, 0.0358, 0.1544, 0.0263, 0.0769, 0.0624, 0.1388, 0.0827], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0177, 0.0195, 0.0164, 0.0176, 0.0217, 0.0203, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:34:59,330 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8550, 2.0880, 2.3828, 3.1079, 2.1565, 2.2972, 2.2534, 2.1888], device='cuda:2'), covar=tensor([0.1430, 0.3496, 0.2695, 0.0783, 0.4216, 0.2601, 0.3534, 0.3708], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0450, 0.0367, 0.0325, 0.0435, 0.0517, 0.0421, 0.0526], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:35:18,628 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:35,336 INFO [train.py:904] (2/8) Epoch 23, batch 7500, loss[loss=0.2208, simple_loss=0.2979, pruned_loss=0.07178, over 16228.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2891, pruned_loss=0.05745, over 3083680.10 frames. ], batch size: 165, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:35:38,010 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2174, 2.2688, 2.3118, 3.8485, 2.2863, 2.6381, 2.3570, 2.4214], device='cuda:2'), covar=tensor([0.1379, 0.3403, 0.2919, 0.0556, 0.4067, 0.2524, 0.3437, 0.3296], device='cuda:2'), in_proj_covar=tensor([0.0401, 0.0450, 0.0367, 0.0325, 0.0435, 0.0517, 0.0422, 0.0526], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:35:40,989 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:58,126 INFO [optim.py:368] (2/8) 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 16:36:09,655 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230824.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:36:52,030 INFO [train.py:904] (2/8) Epoch 23, batch 7550, loss[loss=0.21, simple_loss=0.2915, pruned_loss=0.0643, over 15375.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2885, pruned_loss=0.05813, over 3060934.66 frames. ], batch size: 190, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:36:52,478 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:36:53,595 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:33,343 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:42,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3128, 3.9338, 4.5652, 2.4125, 4.6972, 4.8636, 3.4782, 3.6982], device='cuda:2'), covar=tensor([0.0704, 0.0300, 0.0209, 0.1141, 0.0067, 0.0148, 0.0395, 0.0413], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0108, 0.0099, 0.0139, 0.0082, 0.0126, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:37:45,735 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:47,014 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:58,268 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230897.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:38:06,491 INFO [train.py:904] (2/8) Epoch 23, batch 7600, loss[loss=0.2136, simple_loss=0.2831, pruned_loss=0.07205, over 11694.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2884, pruned_loss=0.05911, over 3047111.54 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:38:16,825 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 16:38:27,912 INFO [optim.py:368] (2/8) 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 16:39:10,331 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:10,575 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0702, 3.2454, 3.2624, 1.9883, 2.8683, 2.1655, 3.4545, 3.5256], device='cuda:2'), covar=tensor([0.0256, 0.0829, 0.0675, 0.2313, 0.0964, 0.1119, 0.0710, 0.1005], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0167, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:39:15,333 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-05-01 16:39:19,847 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:21,933 INFO [train.py:904] (2/8) Epoch 23, batch 7650, loss[loss=0.1869, simple_loss=0.281, pruned_loss=0.04636, over 16511.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2891, pruned_loss=0.05958, over 3059679.55 frames. ], batch size: 75, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:39:43,687 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 16:40:01,376 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 16:40:35,935 INFO [train.py:904] (2/8) Epoch 23, batch 7700, loss[loss=0.2061, simple_loss=0.2912, pruned_loss=0.06048, over 16764.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2891, pruned_loss=0.05961, over 3063856.33 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:40:57,683 INFO [optim.py:368] (2/8) 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 16:41:05,311 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 16:41:49,930 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-01 16:41:53,916 INFO [train.py:904] (2/8) Epoch 23, batch 7750, loss[loss=0.1791, simple_loss=0.2772, pruned_loss=0.04054, over 16828.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2895, pruned_loss=0.05966, over 3070253.43 frames. ], batch size: 102, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:43:09,649 INFO [train.py:904] (2/8) Epoch 23, batch 7800, loss[loss=0.1753, simple_loss=0.2711, pruned_loss=0.03979, over 16817.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2905, pruned_loss=0.06053, over 3056398.16 frames. ], batch size: 102, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:43:30,321 INFO [optim.py:368] (2/8) 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 16:43:34,027 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6509, 1.7474, 1.5000, 1.3963, 1.8552, 1.5103, 1.5682, 1.8585], device='cuda:2'), covar=tensor([0.0269, 0.0400, 0.0547, 0.0483, 0.0292, 0.0374, 0.0227, 0.0294], device='cuda:2'), in_proj_covar=tensor([0.0212, 0.0234, 0.0225, 0.0228, 0.0236, 0.0233, 0.0234, 0.0232], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:43:41,758 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231124.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:44:17,269 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:24,895 INFO [train.py:904] (2/8) Epoch 23, batch 7850, loss[loss=0.2369, simple_loss=0.3027, pruned_loss=0.08559, over 11562.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2909, pruned_loss=0.05984, over 3062582.45 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:44:39,614 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-05-01 16:44:52,236 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:05,960 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:16,800 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231189.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:37,199 INFO [train.py:904] (2/8) Epoch 23, batch 7900, loss[loss=0.1997, simple_loss=0.282, pruned_loss=0.05864, over 16733.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2901, pruned_loss=0.05941, over 3051856.10 frames. ], batch size: 62, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:45:57,448 INFO [optim.py:368] (2/8) 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 16:46:01,326 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-01 16:46:16,872 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:31,294 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:44,778 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:54,843 INFO [train.py:904] (2/8) Epoch 23, batch 7950, loss[loss=0.1818, simple_loss=0.2775, pruned_loss=0.04303, over 17126.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2902, pruned_loss=0.05995, over 3050353.15 frames. ], batch size: 47, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:47:59,478 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6313, 2.5441, 1.9080, 2.6934, 2.1485, 2.7625, 2.1874, 2.3786], device='cuda:2'), covar=tensor([0.0301, 0.0309, 0.1234, 0.0226, 0.0629, 0.0441, 0.1064, 0.0565], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0175, 0.0193, 0.0163, 0.0175, 0.0216, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:48:01,058 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 16:48:11,526 INFO [train.py:904] (2/8) Epoch 23, batch 8000, loss[loss=0.2109, simple_loss=0.297, pruned_loss=0.06243, over 16858.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2912, pruned_loss=0.06073, over 3054286.59 frames. ], batch size: 83, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:22,725 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7955, 3.8981, 4.1292, 4.0899, 4.1244, 3.8980, 3.8799, 3.8983], device='cuda:2'), covar=tensor([0.0394, 0.0648, 0.0462, 0.0487, 0.0479, 0.0506, 0.0933, 0.0573], device='cuda:2'), in_proj_covar=tensor([0.0412, 0.0457, 0.0444, 0.0413, 0.0489, 0.0466, 0.0550, 0.0372], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 16:48:26,410 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231312.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:48:33,449 INFO [optim.py:368] (2/8) 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 16:49:26,911 INFO [train.py:904] (2/8) Epoch 23, batch 8050, loss[loss=0.189, simple_loss=0.2759, pruned_loss=0.05107, over 16930.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2905, pruned_loss=0.0598, over 3068380.82 frames. ], batch size: 109, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:50:38,352 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-01 16:50:42,759 INFO [train.py:904] (2/8) Epoch 23, batch 8100, loss[loss=0.1862, simple_loss=0.2773, pruned_loss=0.04751, over 16453.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.29, pruned_loss=0.05923, over 3080268.48 frames. ], batch size: 75, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:51:00,328 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 16:51:04,319 INFO [optim.py:368] (2/8) 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 16:51:51,102 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:51:56,212 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5119, 3.4802, 3.4665, 2.7115, 3.3513, 2.0966, 3.1500, 2.8995], device='cuda:2'), covar=tensor([0.0161, 0.0137, 0.0193, 0.0223, 0.0106, 0.2317, 0.0141, 0.0252], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0159, 0.0200, 0.0178, 0.0176, 0.0208, 0.0188, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:51:59,308 INFO [train.py:904] (2/8) Epoch 23, batch 8150, loss[loss=0.2123, simple_loss=0.2867, pruned_loss=0.06891, over 11563.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2875, pruned_loss=0.05785, over 3097359.17 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:05,171 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:53:09,589 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6849, 4.7193, 5.0747, 5.0084, 5.0387, 4.7541, 4.7018, 4.5630], device='cuda:2'), covar=tensor([0.0322, 0.0536, 0.0367, 0.0412, 0.0506, 0.0395, 0.0968, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0412, 0.0457, 0.0444, 0.0413, 0.0489, 0.0466, 0.0550, 0.0371], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 16:53:14,922 INFO [train.py:904] (2/8) Epoch 23, batch 8200, loss[loss=0.2107, simple_loss=0.2856, pruned_loss=0.06791, over 11499.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2853, pruned_loss=0.05722, over 3100367.45 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:19,237 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0088, 1.9983, 2.5422, 2.9372, 2.8316, 3.2159, 2.2059, 3.2017], device='cuda:2'), covar=tensor([0.0209, 0.0502, 0.0333, 0.0281, 0.0284, 0.0200, 0.0538, 0.0189], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0191, 0.0177, 0.0182, 0.0196, 0.0154, 0.0196, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:53:38,107 INFO [optim.py:368] (2/8) 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 16:54:26,263 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:54:37,493 INFO [train.py:904] (2/8) Epoch 23, batch 8250, loss[loss=0.1706, simple_loss=0.2608, pruned_loss=0.04017, over 12171.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2843, pruned_loss=0.05492, over 3072828.23 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:54:53,668 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8021, 3.8543, 4.0047, 3.7520, 3.9023, 4.3138, 3.9657, 3.6464], device='cuda:2'), covar=tensor([0.2274, 0.2268, 0.2239, 0.2476, 0.2766, 0.1498, 0.1612, 0.2618], device='cuda:2'), in_proj_covar=tensor([0.0413, 0.0597, 0.0661, 0.0495, 0.0658, 0.0687, 0.0515, 0.0662], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 16:55:42,941 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:55:57,538 INFO [train.py:904] (2/8) Epoch 23, batch 8300, loss[loss=0.1761, simple_loss=0.2787, pruned_loss=0.03681, over 16694.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2817, pruned_loss=0.05212, over 3066152.10 frames. ], batch size: 89, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:56:04,248 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231607.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:56:22,156 INFO [optim.py:368] (2/8) 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 16:57:11,252 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231647.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:57:19,275 INFO [train.py:904] (2/8) Epoch 23, batch 8350, loss[loss=0.1807, simple_loss=0.2757, pruned_loss=0.04286, over 16204.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2813, pruned_loss=0.05003, over 3081552.11 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:57:54,065 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0384, 3.9902, 3.8786, 2.9642, 3.9466, 1.6914, 3.7083, 3.5066], device='cuda:2'), covar=tensor([0.0145, 0.0145, 0.0237, 0.0431, 0.0129, 0.3374, 0.0171, 0.0341], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0158, 0.0200, 0.0177, 0.0175, 0.0208, 0.0188, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 16:57:54,114 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0626, 3.0979, 1.9676, 3.3183, 2.4066, 3.3207, 2.1696, 2.6794], device='cuda:2'), covar=tensor([0.0326, 0.0330, 0.1495, 0.0268, 0.0797, 0.0493, 0.1403, 0.0654], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0174, 0.0192, 0.0162, 0.0175, 0.0215, 0.0202, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:57:55,445 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0417, 2.3628, 2.3417, 2.8534, 1.8027, 3.1759, 1.7537, 2.8206], device='cuda:2'), covar=tensor([0.1222, 0.0615, 0.0999, 0.0187, 0.0074, 0.0354, 0.1532, 0.0630], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0174, 0.0196, 0.0190, 0.0206, 0.0214, 0.0203, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 16:58:41,956 INFO [train.py:904] (2/8) Epoch 23, batch 8400, loss[loss=0.1522, simple_loss=0.2557, pruned_loss=0.02428, over 16868.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2781, pruned_loss=0.04785, over 3064462.75 frames. ], batch size: 102, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:58:50,852 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231708.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 16:58:57,642 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 16:59:06,587 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 23, batch 8450, loss[loss=0.1856, simple_loss=0.2744, pruned_loss=0.04839, over 12391.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2762, pruned_loss=0.04601, over 3067542.17 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:15,503 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-01 17:01:24,902 INFO [train.py:904] (2/8) Epoch 23, batch 8500, loss[loss=0.147, simple_loss=0.2342, pruned_loss=0.02995, over 11868.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2727, pruned_loss=0.04425, over 3049738.66 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:35,402 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 17:01:48,516 INFO [optim.py:368] (2/8) 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 17:02:48,223 INFO [train.py:904] (2/8) Epoch 23, batch 8550, loss[loss=0.1845, simple_loss=0.2746, pruned_loss=0.04718, over 16419.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2704, pruned_loss=0.04342, over 3044307.18 frames. ], batch size: 68, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:02:58,120 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0378, 3.0550, 1.8670, 3.3029, 2.2533, 3.2922, 2.1481, 2.6134], device='cuda:2'), covar=tensor([0.0317, 0.0370, 0.1625, 0.0265, 0.0879, 0.0549, 0.1468, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0173, 0.0192, 0.0161, 0.0175, 0.0213, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 17:04:28,270 INFO [train.py:904] (2/8) Epoch 23, batch 8600, loss[loss=0.1698, simple_loss=0.2566, pruned_loss=0.04149, over 12069.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2699, pruned_loss=0.04221, over 3035536.82 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:04:36,642 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:05:00,379 INFO [optim.py:368] (2/8) 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 17:05:07,177 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-01 17:06:03,592 INFO [train.py:904] (2/8) Epoch 23, batch 8650, loss[loss=0.1756, simple_loss=0.2667, pruned_loss=0.04226, over 16627.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2682, pruned_loss=0.04107, over 3020304.38 frames. ], batch size: 62, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:06:10,095 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:07:40,225 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7791, 3.0698, 3.4618, 1.9977, 2.9394, 2.1600, 3.2767, 3.3406], device='cuda:2'), covar=tensor([0.0264, 0.0837, 0.0452, 0.2052, 0.0767, 0.1008, 0.0637, 0.0876], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0160, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 17:07:53,209 INFO [train.py:904] (2/8) Epoch 23, batch 8700, loss[loss=0.161, simple_loss=0.2573, pruned_loss=0.03235, over 16851.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2654, pruned_loss=0.03972, over 3045231.58 frames. ], batch size: 124, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:07:54,258 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232003.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:08:23,060 INFO [optim.py:368] (2/8) 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 17:09:27,783 INFO [train.py:904] (2/8) Epoch 23, batch 8750, loss[loss=0.1501, simple_loss=0.238, pruned_loss=0.03112, over 11922.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.265, pruned_loss=0.03924, over 3034152.28 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:09:42,376 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4453, 3.3269, 2.7234, 2.1617, 2.2030, 2.3022, 3.4667, 2.9337], device='cuda:2'), covar=tensor([0.3040, 0.0694, 0.1836, 0.2938, 0.2820, 0.2328, 0.0490, 0.1661], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0263, 0.0302, 0.0311, 0.0292, 0.0259, 0.0293, 0.0333], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:11:15,140 INFO [train.py:904] (2/8) Epoch 23, batch 8800, loss[loss=0.1667, simple_loss=0.2631, pruned_loss=0.03522, over 16784.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2639, pruned_loss=0.03829, over 3040956.88 frames. ], batch size: 124, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:11:29,643 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9679, 4.2552, 4.1056, 4.1191, 3.7976, 3.8400, 3.8897, 4.2608], device='cuda:2'), covar=tensor([0.1096, 0.0922, 0.0892, 0.0776, 0.0813, 0.1677, 0.1017, 0.0897], device='cuda:2'), in_proj_covar=tensor([0.0666, 0.0810, 0.0670, 0.0620, 0.0513, 0.0527, 0.0681, 0.0636], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:11:46,653 INFO [optim.py:368] (2/8) 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 17:12:57,913 INFO [train.py:904] (2/8) Epoch 23, batch 8850, loss[loss=0.1887, simple_loss=0.3009, pruned_loss=0.0383, over 16910.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2669, pruned_loss=0.03779, over 3047473.64 frames. ], batch size: 116, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:13:46,067 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1789, 3.9948, 4.2362, 4.3230, 4.4887, 4.0627, 4.4449, 4.5253], device='cuda:2'), covar=tensor([0.1654, 0.1249, 0.1399, 0.0759, 0.0506, 0.1201, 0.0676, 0.0582], device='cuda:2'), in_proj_covar=tensor([0.0614, 0.0762, 0.0876, 0.0771, 0.0586, 0.0614, 0.0639, 0.0745], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:14:42,081 INFO [train.py:904] (2/8) Epoch 23, batch 8900, loss[loss=0.1642, simple_loss=0.2533, pruned_loss=0.03751, over 12489.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2674, pruned_loss=0.03734, over 3045059.14 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:15:11,424 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1098, 3.4006, 3.4094, 2.3183, 3.0861, 3.4470, 3.2607, 1.9889], device='cuda:2'), covar=tensor([0.0537, 0.0048, 0.0054, 0.0395, 0.0110, 0.0082, 0.0087, 0.0506], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0083, 0.0083, 0.0132, 0.0097, 0.0108, 0.0093, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:15:12,055 INFO [optim.py:368] (2/8) 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 17:16:30,655 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-05-01 17:16:33,791 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:16:45,331 INFO [train.py:904] (2/8) Epoch 23, batch 8950, loss[loss=0.1831, simple_loss=0.2727, pruned_loss=0.04675, over 12365.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2671, pruned_loss=0.03747, over 3066548.90 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:18:21,538 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0284, 2.1251, 2.2229, 3.5398, 2.0800, 2.3855, 2.2382, 2.1971], device='cuda:2'), covar=tensor([0.1417, 0.3846, 0.3250, 0.0636, 0.4473, 0.2696, 0.3800, 0.3872], device='cuda:2'), in_proj_covar=tensor([0.0397, 0.0446, 0.0367, 0.0321, 0.0431, 0.0510, 0.0418, 0.0519], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:18:29,513 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1345, 5.4326, 5.2228, 5.2273, 4.9696, 4.9456, 4.8258, 5.5349], device='cuda:2'), covar=tensor([0.1186, 0.0855, 0.1006, 0.0828, 0.0806, 0.0837, 0.1216, 0.0801], device='cuda:2'), in_proj_covar=tensor([0.0667, 0.0812, 0.0670, 0.0620, 0.0515, 0.0528, 0.0682, 0.0639], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:18:31,849 INFO [train.py:904] (2/8) Epoch 23, batch 9000, loss[loss=0.1522, simple_loss=0.2469, pruned_loss=0.02876, over 15359.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2631, pruned_loss=0.03596, over 3069580.86 frames. ], batch size: 192, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:18:31,849 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 17:18:42,667 INFO [train.py:938] (2/8) Epoch 23, validation: loss=0.1452, simple_loss=0.249, pruned_loss=0.02066, over 944034.00 frames. 2023-05-01 17:18:42,668 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 17:18:43,696 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232303.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:18:54,263 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232308.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:19:01,180 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8283, 4.6031, 4.8532, 4.9911, 5.1857, 4.6427, 5.1969, 5.1838], device='cuda:2'), covar=tensor([0.1912, 0.1366, 0.1709, 0.0803, 0.0567, 0.0862, 0.0556, 0.0755], device='cuda:2'), in_proj_covar=tensor([0.0617, 0.0764, 0.0876, 0.0773, 0.0588, 0.0615, 0.0641, 0.0747], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:19:18,067 INFO [optim.py:368] (2/8) 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 17:20:01,865 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6240, 3.9054, 2.8176, 2.2239, 2.3816, 2.3945, 4.1627, 3.3465], device='cuda:2'), covar=tensor([0.3006, 0.0525, 0.1793, 0.2974, 0.2897, 0.2166, 0.0367, 0.1293], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0263, 0.0301, 0.0311, 0.0291, 0.0259, 0.0292, 0.0332], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:20:24,509 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232351.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:20:28,118 INFO [train.py:904] (2/8) Epoch 23, batch 9050, loss[loss=0.1679, simple_loss=0.2583, pruned_loss=0.03876, over 16181.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2639, pruned_loss=0.03621, over 3078173.35 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:20:37,212 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6114, 3.5590, 3.5588, 2.8308, 3.4929, 2.0047, 3.3458, 3.0244], device='cuda:2'), covar=tensor([0.0186, 0.0168, 0.0198, 0.0272, 0.0133, 0.2582, 0.0165, 0.0293], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0157, 0.0198, 0.0174, 0.0174, 0.0206, 0.0186, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:21:23,462 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:22:12,819 INFO [train.py:904] (2/8) Epoch 23, batch 9100, loss[loss=0.1905, simple_loss=0.299, pruned_loss=0.041, over 16249.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2638, pruned_loss=0.03688, over 3080241.69 frames. ], batch size: 166, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:22:46,300 INFO [optim.py:368] (2/8) 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 17:23:46,430 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232442.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:24:09,377 INFO [train.py:904] (2/8) Epoch 23, batch 9150, loss[loss=0.16, simple_loss=0.2567, pruned_loss=0.03164, over 15341.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2644, pruned_loss=0.03686, over 3076295.89 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:25:00,538 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:25:45,335 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6915, 3.7240, 2.7303, 2.2085, 2.3353, 2.4275, 3.9368, 3.1992], device='cuda:2'), covar=tensor([0.2849, 0.0571, 0.1919, 0.2964, 0.2937, 0.2231, 0.0405, 0.1466], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0263, 0.0302, 0.0311, 0.0290, 0.0259, 0.0292, 0.0333], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:25:52,924 INFO [train.py:904] (2/8) Epoch 23, batch 9200, loss[loss=0.1564, simple_loss=0.237, pruned_loss=0.0379, over 11977.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2605, pruned_loss=0.03607, over 3080419.07 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:26:19,917 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5217, 4.8469, 4.6669, 4.6471, 4.3675, 4.3573, 4.2734, 4.9169], device='cuda:2'), covar=tensor([0.1185, 0.0872, 0.0934, 0.0821, 0.0842, 0.1344, 0.1160, 0.0971], device='cuda:2'), in_proj_covar=tensor([0.0664, 0.0806, 0.0666, 0.0617, 0.0512, 0.0526, 0.0678, 0.0636], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:26:21,320 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:26:24,112 INFO [optim.py:368] (2/8) 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 17:26:57,037 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:27:09,416 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 17:27:21,250 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4122, 3.0903, 3.2954, 1.7394, 3.4538, 3.5215, 2.8615, 2.7738], device='cuda:2'), covar=tensor([0.0766, 0.0296, 0.0212, 0.1298, 0.0089, 0.0168, 0.0424, 0.0453], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0104, 0.0094, 0.0134, 0.0078, 0.0120, 0.0122, 0.0125], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 17:27:27,732 INFO [train.py:904] (2/8) Epoch 23, batch 9250, loss[loss=0.1756, simple_loss=0.2708, pruned_loss=0.0402, over 16679.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.26, pruned_loss=0.0357, over 3073382.80 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:27:29,750 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 17:28:23,215 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232579.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:00,887 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9445, 2.3133, 2.3698, 2.9201, 1.7477, 3.0857, 1.7312, 2.8038], device='cuda:2'), covar=tensor([0.1421, 0.0799, 0.1204, 0.0189, 0.0100, 0.0438, 0.1741, 0.0731], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0173, 0.0194, 0.0186, 0.0200, 0.0211, 0.0201, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 17:29:17,460 INFO [train.py:904] (2/8) Epoch 23, batch 9300, loss[loss=0.1679, simple_loss=0.2657, pruned_loss=0.03502, over 15225.00 frames. ], tot_loss[loss=0.165, simple_loss=0.259, pruned_loss=0.03553, over 3076037.95 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:29:18,417 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232603.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:31,440 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6650, 3.8298, 2.2881, 4.3520, 2.8710, 4.2186, 2.5950, 3.1285], device='cuda:2'), covar=tensor([0.0315, 0.0354, 0.1742, 0.0243, 0.0836, 0.0531, 0.1464, 0.0754], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0171, 0.0188, 0.0157, 0.0171, 0.0207, 0.0198, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 17:29:58,634 INFO [optim.py:368] (2/8) 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 17:30:02,476 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5839, 2.5416, 2.4068, 4.2679, 2.6814, 4.0352, 1.4228, 2.9683], device='cuda:2'), covar=tensor([0.1485, 0.0871, 0.1268, 0.0176, 0.0142, 0.0386, 0.1784, 0.0726], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0173, 0.0194, 0.0186, 0.0199, 0.0211, 0.0201, 0.0191], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 17:30:41,697 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 17:31:04,807 INFO [train.py:904] (2/8) Epoch 23, batch 9350, loss[loss=0.1748, simple_loss=0.262, pruned_loss=0.04386, over 16586.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2584, pruned_loss=0.03524, over 3080038.79 frames. ], batch size: 62, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:31:59,216 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9922, 2.7091, 2.8938, 2.1647, 2.7250, 2.1819, 2.7343, 2.9540], device='cuda:2'), covar=tensor([0.0297, 0.0896, 0.0479, 0.1795, 0.0770, 0.0992, 0.0626, 0.0762], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0158, 0.0163, 0.0150, 0.0142, 0.0126, 0.0139, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 17:32:47,939 INFO [train.py:904] (2/8) Epoch 23, batch 9400, loss[loss=0.1787, simple_loss=0.275, pruned_loss=0.04123, over 15394.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2584, pruned_loss=0.03491, over 3082149.60 frames. ], batch size: 190, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:33:21,842 INFO [optim.py:368] (2/8) 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 17:33:57,945 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232737.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:34:29,649 INFO [train.py:904] (2/8) Epoch 23, batch 9450, loss[loss=0.1717, simple_loss=0.2616, pruned_loss=0.04092, over 16949.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2603, pruned_loss=0.03521, over 3081415.66 frames. ], batch size: 109, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:00,034 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7949, 4.1309, 3.0222, 2.3270, 2.5565, 2.5807, 4.4276, 3.4602], device='cuda:2'), covar=tensor([0.2891, 0.0515, 0.1777, 0.3072, 0.2965, 0.2113, 0.0319, 0.1309], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0262, 0.0299, 0.0309, 0.0286, 0.0257, 0.0289, 0.0329], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:36:04,897 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5626, 3.4989, 3.5247, 2.6044, 3.4541, 1.8767, 3.2136, 2.8241], device='cuda:2'), covar=tensor([0.0199, 0.0204, 0.0211, 0.0398, 0.0156, 0.3303, 0.0201, 0.0427], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0157, 0.0196, 0.0172, 0.0173, 0.0205, 0.0185, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:36:09,220 INFO [train.py:904] (2/8) Epoch 23, batch 9500, loss[loss=0.1719, simple_loss=0.2628, pruned_loss=0.04056, over 16936.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2594, pruned_loss=0.03494, over 3067094.37 frames. ], batch size: 109, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:44,545 INFO [optim.py:368] (2/8) 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 17:37:08,359 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:37:54,042 INFO [train.py:904] (2/8) Epoch 23, batch 9550, loss[loss=0.1842, simple_loss=0.2842, pruned_loss=0.04203, over 15333.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2585, pruned_loss=0.03484, over 3069886.75 frames. ], batch size: 192, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:38:37,447 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:38:37,656 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4317, 2.9595, 3.0822, 2.0024, 2.7905, 2.1814, 2.9538, 3.1841], device='cuda:2'), covar=tensor([0.0368, 0.0831, 0.0635, 0.2139, 0.0889, 0.1041, 0.0756, 0.0949], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0157, 0.0163, 0.0150, 0.0141, 0.0126, 0.0139, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 17:39:33,898 INFO [train.py:904] (2/8) Epoch 23, batch 9600, loss[loss=0.1756, simple_loss=0.2738, pruned_loss=0.03877, over 15479.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2606, pruned_loss=0.03578, over 3060318.16 frames. ], batch size: 191, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:39:34,575 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232903.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:40:05,774 INFO [optim.py:368] (2/8) 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 17:41:17,234 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:41:20,712 INFO [train.py:904] (2/8) Epoch 23, batch 9650, loss[loss=0.1555, simple_loss=0.253, pruned_loss=0.02899, over 16575.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2622, pruned_loss=0.03604, over 3036294.75 frames. ], batch size: 68, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:41:22,168 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3724, 3.4378, 3.6602, 3.6580, 3.6488, 3.4799, 3.5120, 3.5388], device='cuda:2'), covar=tensor([0.0416, 0.0695, 0.0486, 0.0433, 0.0607, 0.0530, 0.0771, 0.0478], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0440, 0.0432, 0.0396, 0.0473, 0.0448, 0.0528, 0.0361], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 17:42:41,576 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:00,267 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 17:43:09,186 INFO [train.py:904] (2/8) Epoch 23, batch 9700, loss[loss=0.1686, simple_loss=0.2639, pruned_loss=0.03663, over 16803.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2616, pruned_loss=0.03615, over 3038973.47 frames. ], batch size: 124, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:43:25,043 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6588, 1.8641, 2.1968, 2.5735, 2.5334, 2.8875, 1.9031, 2.8857], device='cuda:2'), covar=tensor([0.0232, 0.0555, 0.0391, 0.0340, 0.0349, 0.0222, 0.0573, 0.0176], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0188, 0.0174, 0.0177, 0.0193, 0.0150, 0.0192, 0.0147], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:43:40,110 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:43,780 INFO [optim.py:368] (2/8) 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 17:44:22,049 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:44:50,793 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233051.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:44:53,542 INFO [train.py:904] (2/8) Epoch 23, batch 9750, loss[loss=0.1629, simple_loss=0.2609, pruned_loss=0.03247, over 16943.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2603, pruned_loss=0.03611, over 3047894.55 frames. ], batch size: 116, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:45:31,743 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 17:45:43,952 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:45:59,985 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233085.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:46:31,579 INFO [train.py:904] (2/8) Epoch 23, batch 9800, loss[loss=0.1545, simple_loss=0.2618, pruned_loss=0.02356, over 16502.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2607, pruned_loss=0.03522, over 3070253.63 frames. ], batch size: 68, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:47:03,692 INFO [optim.py:368] (2/8) 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 17:47:07,715 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5144, 4.5296, 4.2866, 3.7767, 4.4322, 1.7347, 4.2164, 4.0738], device='cuda:2'), covar=tensor([0.0084, 0.0084, 0.0212, 0.0279, 0.0095, 0.2690, 0.0124, 0.0257], device='cuda:2'), in_proj_covar=tensor([0.0165, 0.0157, 0.0196, 0.0171, 0.0173, 0.0206, 0.0185, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:47:26,790 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:48:16,806 INFO [train.py:904] (2/8) Epoch 23, batch 9850, loss[loss=0.1609, simple_loss=0.2588, pruned_loss=0.03147, over 16921.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2616, pruned_loss=0.0351, over 3070693.02 frames. ], batch size: 109, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:48:20,837 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9091, 3.9011, 4.0477, 3.8691, 3.9882, 4.3661, 3.9891, 3.6868], device='cuda:2'), covar=tensor([0.2043, 0.2430, 0.2457, 0.2493, 0.2826, 0.1507, 0.1613, 0.2703], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0570, 0.0631, 0.0471, 0.0628, 0.0657, 0.0491, 0.0627], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:49:01,110 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233174.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:49:12,407 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-05-01 17:49:13,852 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233180.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:50:08,889 INFO [train.py:904] (2/8) Epoch 23, batch 9900, loss[loss=0.1914, simple_loss=0.2919, pruned_loss=0.04551, over 16946.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2619, pruned_loss=0.03504, over 3069777.80 frames. ], batch size: 109, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:50:46,347 INFO [optim.py:368] (2/8) 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 17:50:52,787 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:51:06,658 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6988, 2.6253, 1.7415, 2.8166, 2.1205, 2.7982, 1.9822, 2.3591], device='cuda:2'), covar=tensor([0.0266, 0.0348, 0.1374, 0.0255, 0.0654, 0.0474, 0.1319, 0.0532], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0168, 0.0186, 0.0155, 0.0169, 0.0205, 0.0197, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-01 17:52:06,031 INFO [train.py:904] (2/8) Epoch 23, batch 9950, loss[loss=0.1662, simple_loss=0.267, pruned_loss=0.03268, over 15377.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2638, pruned_loss=0.0352, over 3060688.63 frames. ], batch size: 191, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:52:20,198 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:52:55,433 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5514, 4.5913, 4.4150, 4.0220, 4.1161, 4.4993, 4.2664, 4.2013], device='cuda:2'), covar=tensor([0.0631, 0.0851, 0.0388, 0.0391, 0.0935, 0.0680, 0.0545, 0.0715], device='cuda:2'), in_proj_covar=tensor([0.0281, 0.0414, 0.0327, 0.0325, 0.0328, 0.0379, 0.0223, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-05-01 17:52:57,504 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8498, 2.8113, 2.6711, 1.9042, 2.5797, 2.8003, 2.6480, 1.8958], device='cuda:2'), covar=tensor([0.0480, 0.0071, 0.0073, 0.0410, 0.0130, 0.0100, 0.0099, 0.0449], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0083, 0.0084, 0.0132, 0.0097, 0.0107, 0.0093, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:53:57,553 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8662, 2.8356, 2.6883, 1.9175, 2.6367, 2.8297, 2.6794, 1.9147], device='cuda:2'), covar=tensor([0.0497, 0.0072, 0.0067, 0.0403, 0.0115, 0.0096, 0.0094, 0.0453], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0083, 0.0084, 0.0133, 0.0097, 0.0108, 0.0093, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:54:07,452 INFO [train.py:904] (2/8) Epoch 23, batch 10000, loss[loss=0.1473, simple_loss=0.2528, pruned_loss=0.02087, over 16926.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2625, pruned_loss=0.03488, over 3075113.96 frames. ], batch size: 102, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:54:40,344 INFO [optim.py:368] (2/8) 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 17:54:41,025 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:55:36,795 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233346.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:55:47,980 INFO [train.py:904] (2/8) Epoch 23, batch 10050, loss[loss=0.181, simple_loss=0.2808, pruned_loss=0.04062, over 16367.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2626, pruned_loss=0.03479, over 3086236.65 frames. ], batch size: 146, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:56:12,167 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2561, 4.3175, 4.4717, 4.3013, 4.3570, 4.8471, 4.4276, 4.0673], device='cuda:2'), covar=tensor([0.1504, 0.1996, 0.2120, 0.1993, 0.2517, 0.1038, 0.1670, 0.2728], device='cuda:2'), in_proj_covar=tensor([0.0388, 0.0567, 0.0630, 0.0471, 0.0627, 0.0658, 0.0491, 0.0627], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 17:56:31,051 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:57:20,830 INFO [train.py:904] (2/8) Epoch 23, batch 10100, loss[loss=0.1561, simple_loss=0.25, pruned_loss=0.03108, over 16459.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2629, pruned_loss=0.03503, over 3081467.23 frames. ], batch size: 68, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:57:41,373 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3406, 4.3024, 4.1507, 3.5968, 4.2494, 1.8338, 4.0549, 3.9398], device='cuda:2'), covar=tensor([0.0117, 0.0139, 0.0214, 0.0295, 0.0118, 0.2586, 0.0148, 0.0251], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0156, 0.0194, 0.0170, 0.0172, 0.0204, 0.0184, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 17:57:53,979 INFO [optim.py:368] (2/8) 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 17:59:06,334 INFO [train.py:904] (2/8) Epoch 24, batch 0, loss[loss=0.2038, simple_loss=0.2805, pruned_loss=0.06357, over 16820.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2805, pruned_loss=0.06357, over 16820.00 frames. ], batch size: 96, lr: 2.84e-03, grad_scale: 8.0 2023-05-01 17:59:06,335 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 17:59:14,236 INFO [train.py:938] (2/8) Epoch 24, validation: loss=0.145, simple_loss=0.2483, pruned_loss=0.02085, over 944034.00 frames. 2023-05-01 17:59:14,237 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 17:59:59,963 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0049, 1.9910, 2.5573, 2.9081, 2.7501, 3.4433, 1.8836, 3.5288], device='cuda:2'), covar=tensor([0.0244, 0.0683, 0.0376, 0.0363, 0.0374, 0.0231, 0.0779, 0.0182], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0190, 0.0176, 0.0179, 0.0195, 0.0151, 0.0194, 0.0149], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:00:06,376 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 18:00:23,695 INFO [train.py:904] (2/8) Epoch 24, batch 50, loss[loss=0.1551, simple_loss=0.2429, pruned_loss=0.03363, over 17219.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2675, pruned_loss=0.0467, over 743186.08 frames. ], batch size: 44, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:00:44,711 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9404, 2.9863, 2.7285, 2.8822, 3.3239, 3.0113, 3.6021, 3.4490], device='cuda:2'), covar=tensor([0.0156, 0.0449, 0.0493, 0.0446, 0.0313, 0.0416, 0.0281, 0.0305], device='cuda:2'), in_proj_covar=tensor([0.0208, 0.0232, 0.0223, 0.0224, 0.0231, 0.0232, 0.0226, 0.0225], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:00:52,607 INFO [optim.py:368] (2/8) 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 18:01:32,986 INFO [train.py:904] (2/8) Epoch 24, batch 100, loss[loss=0.1717, simple_loss=0.2481, pruned_loss=0.04763, over 16702.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2629, pruned_loss=0.04457, over 1312516.03 frames. ], batch size: 83, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:01:53,723 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0564, 3.0687, 1.9509, 3.2340, 2.4260, 3.2446, 2.1466, 2.6229], device='cuda:2'), covar=tensor([0.0366, 0.0435, 0.1615, 0.0310, 0.0826, 0.0736, 0.1416, 0.0771], device='cuda:2'), in_proj_covar=tensor([0.0166, 0.0171, 0.0189, 0.0159, 0.0172, 0.0210, 0.0199, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:02:40,666 INFO [train.py:904] (2/8) Epoch 24, batch 150, loss[loss=0.1679, simple_loss=0.2645, pruned_loss=0.0357, over 17145.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2623, pruned_loss=0.04318, over 1750691.65 frames. ], batch size: 48, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:02:56,017 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:08,349 INFO [optim.py:368] (2/8) 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 18:03:30,577 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:39,354 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233646.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:03:48,631 INFO [train.py:904] (2/8) Epoch 24, batch 200, loss[loss=0.1937, simple_loss=0.2718, pruned_loss=0.0578, over 16880.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2628, pruned_loss=0.04419, over 2095865.81 frames. ], batch size: 116, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:04:17,982 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:34,600 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0557, 2.2057, 2.5637, 2.9785, 2.8420, 3.4406, 2.2712, 3.4362], device='cuda:2'), covar=tensor([0.0288, 0.0535, 0.0378, 0.0359, 0.0388, 0.0214, 0.0551, 0.0197], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0192, 0.0178, 0.0181, 0.0197, 0.0153, 0.0196, 0.0151], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:04:45,849 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:52,541 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2079, 3.2300, 3.4949, 2.3778, 3.2421, 3.5952, 3.2895, 2.0103], device='cuda:2'), covar=tensor([0.0574, 0.0166, 0.0075, 0.0443, 0.0140, 0.0132, 0.0126, 0.0542], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0086, 0.0085, 0.0135, 0.0098, 0.0109, 0.0095, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 18:04:54,301 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233700.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:04:58,753 INFO [train.py:904] (2/8) Epoch 24, batch 250, loss[loss=0.1666, simple_loss=0.2456, pruned_loss=0.04383, over 16890.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2605, pruned_loss=0.04371, over 2366906.96 frames. ], batch size: 90, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:05:07,754 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233710.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:05:25,966 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:05:26,867 INFO [optim.py:368] (2/8) 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 18:06:08,094 INFO [train.py:904] (2/8) Epoch 24, batch 300, loss[loss=0.1509, simple_loss=0.2427, pruned_loss=0.02957, over 16822.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.258, pruned_loss=0.04258, over 2583127.48 frames. ], batch size: 42, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:06:33,427 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233771.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:06:45,286 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1317, 2.1927, 2.3449, 3.8397, 2.2122, 2.5132, 2.2843, 2.3216], device='cuda:2'), covar=tensor([0.1622, 0.3869, 0.3272, 0.0741, 0.4099, 0.2773, 0.4097, 0.3413], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0450, 0.0372, 0.0324, 0.0435, 0.0513, 0.0422, 0.0524], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:07:04,683 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4732, 4.2619, 4.5466, 4.6830, 4.7867, 4.3693, 4.6482, 4.7610], device='cuda:2'), covar=tensor([0.1830, 0.1210, 0.1318, 0.0680, 0.0576, 0.1074, 0.2703, 0.0670], device='cuda:2'), in_proj_covar=tensor([0.0638, 0.0786, 0.0905, 0.0796, 0.0605, 0.0628, 0.0662, 0.0766], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:07:16,410 INFO [train.py:904] (2/8) Epoch 24, batch 350, loss[loss=0.1643, simple_loss=0.2478, pruned_loss=0.04039, over 15480.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2554, pruned_loss=0.04158, over 2745690.11 frames. ], batch size: 190, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:07:43,747 INFO [optim.py:368] (2/8) 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 18:08:25,441 INFO [train.py:904] (2/8) Epoch 24, batch 400, loss[loss=0.1617, simple_loss=0.2488, pruned_loss=0.03727, over 15576.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2544, pruned_loss=0.04128, over 2869567.40 frames. ], batch size: 191, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:32,976 INFO [train.py:904] (2/8) Epoch 24, batch 450, loss[loss=0.1619, simple_loss=0.2531, pruned_loss=0.03539, over 16637.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2539, pruned_loss=0.04144, over 2968079.46 frames. ], batch size: 62, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:35,716 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7233, 2.6385, 2.3117, 2.5623, 3.0044, 2.7659, 3.2545, 3.2167], device='cuda:2'), covar=tensor([0.0160, 0.0456, 0.0555, 0.0466, 0.0312, 0.0438, 0.0273, 0.0301], device='cuda:2'), in_proj_covar=tensor([0.0216, 0.0238, 0.0228, 0.0230, 0.0238, 0.0238, 0.0234, 0.0232], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:09:50,168 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:09:59,639 INFO [optim.py:368] (2/8) 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 18:10:38,994 INFO [train.py:904] (2/8) Epoch 24, batch 500, loss[loss=0.1413, simple_loss=0.2273, pruned_loss=0.02764, over 16858.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2521, pruned_loss=0.04065, over 3046836.44 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:10:53,521 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233963.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:11:01,576 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8679, 3.9026, 3.9605, 3.7902, 3.8464, 4.3180, 3.8969, 3.5368], device='cuda:2'), covar=tensor([0.1924, 0.2101, 0.2322, 0.2339, 0.2945, 0.1722, 0.1707, 0.2832], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0599, 0.0662, 0.0496, 0.0661, 0.0690, 0.0515, 0.0659], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 18:11:35,684 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4217, 4.4589, 4.6003, 4.4128, 4.4412, 5.0197, 4.5368, 4.2077], device='cuda:2'), covar=tensor([0.1744, 0.2308, 0.2397, 0.2441, 0.2958, 0.1268, 0.1949, 0.2902], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0599, 0.0662, 0.0496, 0.0662, 0.0691, 0.0516, 0.0660], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 18:11:37,607 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233995.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:11:50,840 INFO [train.py:904] (2/8) Epoch 24, batch 550, loss[loss=0.1889, simple_loss=0.2646, pruned_loss=0.0566, over 16282.00 frames. ], tot_loss[loss=0.166, simple_loss=0.251, pruned_loss=0.04047, over 3097811.21 frames. ], batch size: 165, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:12:17,095 INFO [optim.py:368] (2/8) 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 18:12:20,400 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:12:57,925 INFO [train.py:904] (2/8) Epoch 24, batch 600, loss[loss=0.1479, simple_loss=0.2335, pruned_loss=0.0312, over 17024.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2503, pruned_loss=0.04063, over 3142448.65 frames. ], batch size: 41, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:13:08,461 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 18:13:17,487 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234066.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:13:45,356 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 24, batch 650, loss[loss=0.159, simple_loss=0.2591, pruned_loss=0.02945, over 16800.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2498, pruned_loss=0.04028, over 3179850.94 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:14:36,190 INFO [optim.py:368] (2/8) 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 18:14:43,093 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 18:14:51,510 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9107, 4.3694, 3.0773, 2.4121, 2.6776, 2.5851, 4.6978, 3.5895], device='cuda:2'), covar=tensor([0.2901, 0.0594, 0.1876, 0.2947, 0.3064, 0.2183, 0.0376, 0.1450], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0270, 0.0308, 0.0318, 0.0297, 0.0265, 0.0298, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 18:15:16,421 INFO [train.py:904] (2/8) Epoch 24, batch 700, loss[loss=0.1692, simple_loss=0.2494, pruned_loss=0.04453, over 12748.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.249, pruned_loss=0.03976, over 3203500.08 frames. ], batch size: 246, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:16:23,770 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:16:24,492 INFO [train.py:904] (2/8) Epoch 24, batch 750, loss[loss=0.1887, simple_loss=0.259, pruned_loss=0.05922, over 16722.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2497, pruned_loss=0.03987, over 3235906.02 frames. ], batch size: 134, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:16:45,416 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 18:16:52,424 INFO [optim.py:368] (2/8) 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 18:17:04,963 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234231.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:17:33,976 INFO [train.py:904] (2/8) Epoch 24, batch 800, loss[loss=0.1603, simple_loss=0.2405, pruned_loss=0.04007, over 16223.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2495, pruned_loss=0.03937, over 3255570.96 frames. ], batch size: 165, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:17:48,322 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234263.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:18:29,128 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234292.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:18:33,106 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234295.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:18:43,327 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 18:18:43,642 INFO [train.py:904] (2/8) Epoch 24, batch 850, loss[loss=0.1625, simple_loss=0.2558, pruned_loss=0.03458, over 16713.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2487, pruned_loss=0.03903, over 3255372.81 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:18:54,136 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0251, 2.2919, 2.6153, 2.9781, 2.7758, 3.4186, 2.4178, 3.4532], device='cuda:2'), covar=tensor([0.0270, 0.0485, 0.0363, 0.0316, 0.0369, 0.0236, 0.0489, 0.0192], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0185, 0.0200, 0.0157, 0.0198, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:19:11,879 INFO [optim.py:368] (2/8) 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 18:19:19,300 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9092, 5.2112, 5.3823, 5.1602, 5.1869, 5.8162, 5.3279, 4.9044], device='cuda:2'), covar=tensor([0.1222, 0.2209, 0.2748, 0.2046, 0.2780, 0.1179, 0.1912, 0.2639], device='cuda:2'), in_proj_covar=tensor([0.0414, 0.0606, 0.0672, 0.0502, 0.0670, 0.0704, 0.0523, 0.0669], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 18:19:38,794 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6397, 4.5411, 4.5533, 4.2539, 4.2609, 4.5709, 4.3844, 4.3587], device='cuda:2'), covar=tensor([0.0689, 0.0944, 0.0339, 0.0332, 0.0886, 0.0538, 0.0501, 0.0660], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0444, 0.0351, 0.0350, 0.0353, 0.0406, 0.0238, 0.0420], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:19:40,584 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:19:50,928 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 18:19:52,508 INFO [train.py:904] (2/8) Epoch 24, batch 900, loss[loss=0.179, simple_loss=0.2662, pruned_loss=0.04594, over 17003.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2475, pruned_loss=0.03837, over 3275980.18 frames. ], batch size: 55, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:20:11,531 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234366.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:20:33,659 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:20:42,288 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 18:21:00,705 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-05-01 18:21:02,818 INFO [train.py:904] (2/8) Epoch 24, batch 950, loss[loss=0.1424, simple_loss=0.2266, pruned_loss=0.02904, over 16663.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2479, pruned_loss=0.03871, over 3281141.25 frames. ], batch size: 37, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:21:17,243 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:27,631 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-05-01 18:21:30,309 INFO [optim.py:368] (2/8) 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 18:22:10,609 INFO [train.py:904] (2/8) Epoch 24, batch 1000, loss[loss=0.142, simple_loss=0.2317, pruned_loss=0.02616, over 17218.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2466, pruned_loss=0.03859, over 3287503.51 frames. ], batch size: 45, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:22:48,989 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0872, 2.1571, 2.3617, 3.6986, 2.2191, 2.4825, 2.2722, 2.3202], device='cuda:2'), covar=tensor([0.1551, 0.3719, 0.3039, 0.0764, 0.4019, 0.2572, 0.3904, 0.3103], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0456, 0.0376, 0.0329, 0.0440, 0.0522, 0.0428, 0.0533], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:22:53,421 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-01 18:23:22,529 INFO [train.py:904] (2/8) Epoch 24, batch 1050, loss[loss=0.1568, simple_loss=0.2367, pruned_loss=0.03841, over 16209.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2463, pruned_loss=0.03827, over 3300269.90 frames. ], batch size: 165, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:23:50,907 INFO [optim.py:368] (2/8) 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 18:24:23,510 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4925, 5.8694, 5.6388, 5.6334, 5.2632, 5.2961, 5.2716, 6.0119], device='cuda:2'), covar=tensor([0.1467, 0.1022, 0.1025, 0.0962, 0.0973, 0.0761, 0.1231, 0.1003], device='cuda:2'), in_proj_covar=tensor([0.0701, 0.0853, 0.0701, 0.0655, 0.0540, 0.0546, 0.0720, 0.0670], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:24:30,872 INFO [train.py:904] (2/8) Epoch 24, batch 1100, loss[loss=0.163, simple_loss=0.2428, pruned_loss=0.04162, over 16318.00 frames. ], tot_loss[loss=0.161, simple_loss=0.246, pruned_loss=0.03798, over 3306889.52 frames. ], batch size: 165, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:24:37,557 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234558.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:25:18,091 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:25:34,397 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 18:25:38,666 INFO [train.py:904] (2/8) Epoch 24, batch 1150, loss[loss=0.1762, simple_loss=0.2542, pruned_loss=0.04906, over 16475.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2456, pruned_loss=0.03778, over 3306534.99 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:25:52,427 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-01 18:25:55,662 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3646, 2.3125, 2.3398, 4.1196, 2.3180, 2.6953, 2.3915, 2.4104], device='cuda:2'), covar=tensor([0.1407, 0.3795, 0.3272, 0.0642, 0.4350, 0.2708, 0.3965, 0.3670], device='cuda:2'), in_proj_covar=tensor([0.0409, 0.0458, 0.0377, 0.0331, 0.0441, 0.0524, 0.0429, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:26:06,112 INFO [optim.py:368] (2/8) 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 18:26:33,542 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 18:26:46,785 INFO [train.py:904] (2/8) Epoch 24, batch 1200, loss[loss=0.1568, simple_loss=0.2424, pruned_loss=0.03558, over 16763.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2449, pruned_loss=0.03722, over 3312641.95 frames. ], batch size: 83, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:27:09,079 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:21,684 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8659, 4.6828, 4.9329, 5.0815, 5.2749, 4.7217, 5.2716, 5.2964], device='cuda:2'), covar=tensor([0.2154, 0.1436, 0.1727, 0.0834, 0.0625, 0.0934, 0.0725, 0.0665], device='cuda:2'), in_proj_covar=tensor([0.0667, 0.0822, 0.0950, 0.0832, 0.0635, 0.0656, 0.0689, 0.0801], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:27:24,993 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:26,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6179, 2.3911, 1.9877, 2.2106, 2.7483, 2.5073, 2.6650, 2.7954], device='cuda:2'), covar=tensor([0.0250, 0.0411, 0.0525, 0.0473, 0.0239, 0.0340, 0.0233, 0.0310], device='cuda:2'), in_proj_covar=tensor([0.0224, 0.0244, 0.0233, 0.0234, 0.0244, 0.0243, 0.0242, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:27:38,814 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4995, 3.5798, 3.8144, 2.7018, 3.4637, 3.9482, 3.5726, 2.4066], device='cuda:2'), covar=tensor([0.0531, 0.0240, 0.0071, 0.0422, 0.0134, 0.0122, 0.0127, 0.0451], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0087, 0.0088, 0.0136, 0.0100, 0.0112, 0.0097, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 18:27:53,727 INFO [train.py:904] (2/8) Epoch 24, batch 1250, loss[loss=0.153, simple_loss=0.237, pruned_loss=0.03451, over 16826.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2447, pruned_loss=0.03727, over 3309947.32 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:28:20,469 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9781, 2.1530, 2.4546, 2.8504, 2.8052, 2.9330, 2.0909, 3.0427], device='cuda:2'), covar=tensor([0.0201, 0.0481, 0.0344, 0.0312, 0.0336, 0.0306, 0.0548, 0.0208], device='cuda:2'), in_proj_covar=tensor([0.0196, 0.0198, 0.0184, 0.0189, 0.0204, 0.0160, 0.0201, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:28:21,121 INFO [optim.py:368] (2/8) 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 18:28:25,262 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:30,422 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234729.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:31,681 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234730.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:28:48,902 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9768, 2.9575, 2.6966, 4.6187, 3.7328, 4.2879, 1.7498, 3.1739], device='cuda:2'), covar=tensor([0.1367, 0.0751, 0.1228, 0.0241, 0.0251, 0.0427, 0.1664, 0.0821], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0175, 0.0195, 0.0193, 0.0203, 0.0216, 0.0205, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:29:02,400 INFO [train.py:904] (2/8) Epoch 24, batch 1300, loss[loss=0.1753, simple_loss=0.255, pruned_loss=0.04775, over 16567.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2459, pruned_loss=0.03814, over 3301672.03 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:29:41,588 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:29:47,320 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:30:10,003 INFO [train.py:904] (2/8) Epoch 24, batch 1350, loss[loss=0.1589, simple_loss=0.2363, pruned_loss=0.04074, over 16454.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2455, pruned_loss=0.03745, over 3312236.96 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:30:38,933 INFO [optim.py:368] (2/8) 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 18:31:07,048 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234843.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:19,624 INFO [train.py:904] (2/8) Epoch 24, batch 1400, loss[loss=0.1698, simple_loss=0.2463, pruned_loss=0.04666, over 16927.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2456, pruned_loss=0.03718, over 3315235.08 frames. ], batch size: 96, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:31:28,012 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:37,820 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7860, 2.7916, 2.7886, 4.8917, 3.9306, 4.3502, 1.6382, 3.0489], device='cuda:2'), covar=tensor([0.1477, 0.0871, 0.1234, 0.0245, 0.0248, 0.0425, 0.1781, 0.0900], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0176, 0.0196, 0.0193, 0.0204, 0.0217, 0.0206, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:32:00,214 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:07,283 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:29,519 INFO [train.py:904] (2/8) Epoch 24, batch 1450, loss[loss=0.1532, simple_loss=0.2438, pruned_loss=0.03135, over 17171.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2454, pruned_loss=0.03729, over 3317390.73 frames. ], batch size: 46, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:32:34,692 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:41,825 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6541, 3.7193, 2.2034, 4.1755, 2.8764, 4.1097, 2.4356, 3.0230], device='cuda:2'), covar=tensor([0.0357, 0.0504, 0.1817, 0.0380, 0.0932, 0.0647, 0.1561, 0.0849], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0170, 0.0179, 0.0221, 0.0206, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:32:58,961 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:26,761 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234943.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:38,961 INFO [train.py:904] (2/8) Epoch 24, batch 1500, loss[loss=0.1603, simple_loss=0.2502, pruned_loss=0.03526, over 17135.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2451, pruned_loss=0.03748, over 3320508.97 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:33:45,611 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:49,610 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234960.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:53,059 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 18:34:43,295 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 18:34:49,721 INFO [train.py:904] (2/8) Epoch 24, batch 1550, loss[loss=0.2034, simple_loss=0.2827, pruned_loss=0.06205, over 11585.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2473, pruned_loss=0.03872, over 3303408.96 frames. ], batch size: 248, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:35:02,675 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2588, 3.4850, 3.4740, 2.2215, 2.9956, 2.4789, 3.6573, 3.7209], device='cuda:2'), covar=tensor([0.0249, 0.0815, 0.0636, 0.1973, 0.0916, 0.0999, 0.0516, 0.0865], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:35:12,883 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235019.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:15,799 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:18,854 INFO [optim.py:368] (2/8) 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 18:35:20,352 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235025.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:35:47,855 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0329, 5.0349, 4.8132, 3.6017, 4.9103, 1.6005, 4.4678, 4.4966], device='cuda:2'), covar=tensor([0.0195, 0.0170, 0.0306, 0.0854, 0.0176, 0.4023, 0.0281, 0.0450], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0166, 0.0207, 0.0182, 0.0183, 0.0215, 0.0195, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:35:58,113 INFO [train.py:904] (2/8) Epoch 24, batch 1600, loss[loss=0.1371, simple_loss=0.2297, pruned_loss=0.02229, over 17219.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2497, pruned_loss=0.04006, over 3292176.39 frames. ], batch size: 45, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:36:37,993 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:36:58,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2859, 3.6788, 3.9653, 2.1203, 3.2213, 2.6483, 3.7079, 3.8320], device='cuda:2'), covar=tensor([0.0338, 0.0877, 0.0435, 0.2069, 0.0801, 0.0889, 0.0691, 0.1107], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0165, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:37:06,937 INFO [train.py:904] (2/8) Epoch 24, batch 1650, loss[loss=0.1426, simple_loss=0.2287, pruned_loss=0.02823, over 16950.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2513, pruned_loss=0.04036, over 3302874.33 frames. ], batch size: 41, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:37:35,279 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235138.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:38:11,779 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5284, 4.3880, 4.4457, 4.1456, 4.1992, 4.4751, 4.2032, 4.2157], device='cuda:2'), covar=tensor([0.0595, 0.0819, 0.0279, 0.0307, 0.0737, 0.0521, 0.0608, 0.0617], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0456, 0.0359, 0.0359, 0.0362, 0.0417, 0.0244, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:38:16,050 INFO [train.py:904] (2/8) Epoch 24, batch 1700, loss[loss=0.1793, simple_loss=0.2642, pruned_loss=0.04723, over 16505.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2525, pruned_loss=0.04007, over 3309030.01 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:38:50,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3181, 4.3606, 4.7012, 4.6712, 4.7187, 4.4078, 4.4329, 4.2822], device='cuda:2'), covar=tensor([0.0402, 0.0726, 0.0389, 0.0403, 0.0553, 0.0508, 0.0853, 0.0738], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0470, 0.0458, 0.0422, 0.0502, 0.0479, 0.0562, 0.0383], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 18:39:24,668 INFO [train.py:904] (2/8) Epoch 24, batch 1750, loss[loss=0.1569, simple_loss=0.2415, pruned_loss=0.0362, over 17218.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2545, pruned_loss=0.04076, over 3299144.67 frames. ], batch size: 44, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:52,420 INFO [optim.py:368] (2/8) 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 18:40:12,949 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235238.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:40:33,176 INFO [train.py:904] (2/8) Epoch 24, batch 1800, loss[loss=0.1943, simple_loss=0.2711, pruned_loss=0.05879, over 16902.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2557, pruned_loss=0.04097, over 3309140.33 frames. ], batch size: 116, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:40:36,487 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:40:57,921 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4539, 3.4774, 3.4883, 2.7880, 3.3188, 2.0929, 3.0997, 2.7043], device='cuda:2'), covar=tensor([0.0181, 0.0161, 0.0182, 0.0235, 0.0109, 0.2601, 0.0152, 0.0302], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0167, 0.0209, 0.0184, 0.0184, 0.0216, 0.0198, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:41:40,803 INFO [train.py:904] (2/8) Epoch 24, batch 1850, loss[loss=0.1919, simple_loss=0.2694, pruned_loss=0.05722, over 16843.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2567, pruned_loss=0.04086, over 3318796.71 frames. ], batch size: 116, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:41:48,074 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 18:41:56,544 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,454 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,701 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:11,305 INFO [optim.py:368] (2/8) 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 18:42:11,768 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:35,516 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1901, 4.2230, 4.5902, 4.5531, 4.5980, 4.2777, 4.2973, 4.2258], device='cuda:2'), covar=tensor([0.0402, 0.0692, 0.0400, 0.0424, 0.0513, 0.0470, 0.0887, 0.0662], device='cuda:2'), in_proj_covar=tensor([0.0427, 0.0474, 0.0463, 0.0426, 0.0506, 0.0483, 0.0567, 0.0387], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 18:42:49,548 INFO [train.py:904] (2/8) Epoch 24, batch 1900, loss[loss=0.1751, simple_loss=0.2646, pruned_loss=0.04279, over 16691.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2563, pruned_loss=0.04048, over 3323133.36 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:43:01,082 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235360.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:43:01,478 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 18:43:17,587 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235373.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:43:21,610 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 18:43:31,565 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:00,956 INFO [train.py:904] (2/8) Epoch 24, batch 1950, loss[loss=0.1744, simple_loss=0.2504, pruned_loss=0.04924, over 16739.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2557, pruned_loss=0.03986, over 3327241.74 frames. ], batch size: 89, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:44:10,996 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2050, 3.3193, 3.5357, 2.1077, 2.9641, 2.4700, 3.6494, 3.6044], device='cuda:2'), covar=tensor([0.0228, 0.0892, 0.0594, 0.2010, 0.0894, 0.0984, 0.0524, 0.0871], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:44:26,806 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235421.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 18:44:31,127 INFO [optim.py:368] (2/8) 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 18:44:34,021 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8361, 2.0799, 2.5277, 2.8404, 2.7307, 3.3322, 2.3037, 3.2808], device='cuda:2'), covar=tensor([0.0299, 0.0514, 0.0367, 0.0369, 0.0391, 0.0199, 0.0541, 0.0173], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0187, 0.0202, 0.0160, 0.0199, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:44:37,916 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:44,082 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5206, 4.5263, 4.4851, 4.0078, 4.5432, 1.7034, 4.2474, 4.0972], device='cuda:2'), covar=tensor([0.0145, 0.0133, 0.0193, 0.0307, 0.0104, 0.3105, 0.0165, 0.0230], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0167, 0.0208, 0.0184, 0.0183, 0.0215, 0.0197, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:44:48,933 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235438.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:45:08,804 INFO [train.py:904] (2/8) Epoch 24, batch 2000, loss[loss=0.1514, simple_loss=0.253, pruned_loss=0.02487, over 17299.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2549, pruned_loss=0.0393, over 3325031.17 frames. ], batch size: 52, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:45:16,440 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9540, 3.8137, 4.1500, 2.4434, 4.3470, 4.3569, 3.2130, 3.4063], device='cuda:2'), covar=tensor([0.0719, 0.0227, 0.0217, 0.1050, 0.0089, 0.0232, 0.0468, 0.0416], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0141, 0.0083, 0.0129, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:45:20,729 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 18:45:36,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7609, 1.9501, 2.3977, 2.6436, 2.7191, 2.6776, 2.0174, 2.8986], device='cuda:2'), covar=tensor([0.0178, 0.0499, 0.0312, 0.0289, 0.0288, 0.0327, 0.0502, 0.0180], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0196, 0.0183, 0.0188, 0.0203, 0.0161, 0.0200, 0.0158], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:45:46,948 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8158, 2.4903, 2.0256, 2.2564, 2.8478, 2.5841, 2.8444, 2.9545], device='cuda:2'), covar=tensor([0.0275, 0.0429, 0.0563, 0.0462, 0.0268, 0.0369, 0.0251, 0.0299], device='cuda:2'), in_proj_covar=tensor([0.0227, 0.0245, 0.0234, 0.0235, 0.0246, 0.0245, 0.0245, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:45:53,864 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235486.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:46:02,484 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0267, 5.0289, 4.8349, 4.3797, 4.9834, 2.0180, 4.6679, 4.6792], device='cuda:2'), covar=tensor([0.0109, 0.0099, 0.0222, 0.0392, 0.0093, 0.2875, 0.0139, 0.0238], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0166, 0.0208, 0.0183, 0.0183, 0.0214, 0.0196, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:46:16,407 INFO [train.py:904] (2/8) Epoch 24, batch 2050, loss[loss=0.18, simple_loss=0.264, pruned_loss=0.04802, over 11681.00 frames. ], tot_loss[loss=0.167, simple_loss=0.254, pruned_loss=0.03998, over 3321462.77 frames. ], batch size: 246, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:46:46,281 INFO [optim.py:368] (2/8) 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 18:46:58,337 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8620, 4.0590, 2.5539, 4.6671, 3.2391, 4.5813, 2.9178, 3.3370], device='cuda:2'), covar=tensor([0.0317, 0.0379, 0.1637, 0.0277, 0.0818, 0.0571, 0.1308, 0.0724], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0173, 0.0180, 0.0224, 0.0207, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:47:03,587 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:47:24,546 INFO [train.py:904] (2/8) Epoch 24, batch 2100, loss[loss=0.1933, simple_loss=0.2907, pruned_loss=0.04797, over 16718.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.255, pruned_loss=0.04095, over 3316415.65 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:47:38,728 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3743, 3.8632, 4.0072, 2.1635, 3.1376, 2.7129, 3.8518, 3.9414], device='cuda:2'), covar=tensor([0.0303, 0.0851, 0.0533, 0.2157, 0.0919, 0.0927, 0.0647, 0.1078], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:48:10,073 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:31,076 INFO [train.py:904] (2/8) Epoch 24, batch 2150, loss[loss=0.1629, simple_loss=0.2415, pruned_loss=0.04212, over 16891.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2563, pruned_loss=0.04174, over 3313346.90 frames. ], batch size: 96, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:48:43,103 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:46,870 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:49,848 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 18:49:31,466 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7033, 2.4381, 1.9908, 2.2030, 2.7641, 2.5096, 2.7821, 2.8632], device='cuda:2'), covar=tensor([0.0207, 0.0394, 0.0511, 0.0450, 0.0242, 0.0348, 0.0197, 0.0300], device='cuda:2'), in_proj_covar=tensor([0.0227, 0.0245, 0.0234, 0.0235, 0.0245, 0.0245, 0.0245, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:49:41,310 INFO [train.py:904] (2/8) Epoch 24, batch 2200, loss[loss=0.1536, simple_loss=0.2397, pruned_loss=0.03373, over 16806.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2566, pruned_loss=0.04259, over 3314373.31 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:49:54,417 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235662.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:49:57,658 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:18,224 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235679.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:50,799 INFO [train.py:904] (2/8) Epoch 24, batch 2250, loss[loss=0.1787, simple_loss=0.2633, pruned_loss=0.04706, over 16777.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2571, pruned_loss=0.04288, over 3302174.41 frames. ], batch size: 83, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:51:08,550 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235716.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:51:20,155 INFO [optim.py:368] (2/8) 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 18:51:41,287 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:51:57,980 INFO [train.py:904] (2/8) Epoch 24, batch 2300, loss[loss=0.155, simple_loss=0.253, pruned_loss=0.02855, over 17109.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2568, pruned_loss=0.0426, over 3307555.66 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:52:50,321 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:53:08,147 INFO [train.py:904] (2/8) Epoch 24, batch 2350, loss[loss=0.1394, simple_loss=0.2349, pruned_loss=0.02193, over 17197.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2573, pruned_loss=0.04276, over 3304551.48 frames. ], batch size: 44, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:53:13,764 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 18:53:37,789 INFO [optim.py:368] (2/8) 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 18:54:06,798 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 18:54:14,695 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:54:15,315 INFO [train.py:904] (2/8) Epoch 24, batch 2400, loss[loss=0.1585, simple_loss=0.2504, pruned_loss=0.03328, over 17097.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2581, pruned_loss=0.04245, over 3284387.88 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:22,142 INFO [train.py:904] (2/8) Epoch 24, batch 2450, loss[loss=0.1579, simple_loss=0.2428, pruned_loss=0.03645, over 16881.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2587, pruned_loss=0.04195, over 3296835.56 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:32,480 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 18:55:33,240 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:55:51,879 INFO [optim.py:368] (2/8) 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] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 18:56:28,796 INFO [train.py:904] (2/8) Epoch 24, batch 2500, loss[loss=0.1362, simple_loss=0.2258, pruned_loss=0.02327, over 16738.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2571, pruned_loss=0.04066, over 3308587.71 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:56:37,230 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:57:05,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7523, 3.7001, 3.8506, 3.5806, 3.7628, 4.2264, 3.8804, 3.5217], device='cuda:2'), covar=tensor([0.2397, 0.2849, 0.2974, 0.2899, 0.3275, 0.2470, 0.1857, 0.3089], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0622, 0.0686, 0.0514, 0.0683, 0.0716, 0.0536, 0.0683], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 18:57:28,646 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5934, 2.6290, 2.6844, 4.4959, 2.5352, 2.9962, 2.6936, 2.7748], device='cuda:2'), covar=tensor([0.1305, 0.3421, 0.2931, 0.0539, 0.3962, 0.2540, 0.3497, 0.3452], device='cuda:2'), in_proj_covar=tensor([0.0411, 0.0461, 0.0378, 0.0334, 0.0442, 0.0529, 0.0432, 0.0539], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:57:38,842 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3061, 4.1127, 4.3949, 4.4972, 4.5956, 4.1942, 4.4246, 4.5942], device='cuda:2'), covar=tensor([0.1731, 0.1302, 0.1319, 0.0729, 0.0627, 0.1273, 0.2500, 0.0842], device='cuda:2'), in_proj_covar=tensor([0.0682, 0.0845, 0.0973, 0.0853, 0.0652, 0.0673, 0.0708, 0.0820], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 18:57:41,344 INFO [train.py:904] (2/8) Epoch 24, batch 2550, loss[loss=0.1465, simple_loss=0.2453, pruned_loss=0.02387, over 17279.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2579, pruned_loss=0.04104, over 3308911.91 frames. ], batch size: 52, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:57:47,259 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-01 18:57:58,590 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236016.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:58:01,871 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:10,533 INFO [optim.py:368] (2/8) 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 18:58:24,269 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236035.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:49,127 INFO [train.py:904] (2/8) Epoch 24, batch 2600, loss[loss=0.1837, simple_loss=0.2828, pruned_loss=0.04234, over 16726.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2574, pruned_loss=0.04086, over 3313425.69 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:59:03,926 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236064.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:59:25,467 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:59:34,847 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7579, 4.0015, 2.5025, 4.6542, 3.1637, 4.5798, 2.7857, 3.3752], device='cuda:2'), covar=tensor([0.0359, 0.0434, 0.1662, 0.0350, 0.0856, 0.0646, 0.1434, 0.0792], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0182, 0.0197, 0.0173, 0.0179, 0.0223, 0.0206, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 18:59:58,188 INFO [train.py:904] (2/8) Epoch 24, batch 2650, loss[loss=0.1656, simple_loss=0.2486, pruned_loss=0.04129, over 16278.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2582, pruned_loss=0.04095, over 3319896.94 frames. ], batch size: 165, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:00:01,930 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 19:00:12,839 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 19:00:27,606 INFO [optim.py:368] (2/8) 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 19:00:58,561 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:01:05,908 INFO [train.py:904] (2/8) Epoch 24, batch 2700, loss[loss=0.2056, simple_loss=0.292, pruned_loss=0.05965, over 16688.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2576, pruned_loss=0.04048, over 3321530.83 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:01:57,231 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7150, 3.8687, 2.5656, 4.5577, 3.0608, 4.4991, 2.7879, 3.3044], device='cuda:2'), covar=tensor([0.0330, 0.0409, 0.1473, 0.0310, 0.0822, 0.0475, 0.1308, 0.0687], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0174, 0.0180, 0.0224, 0.0206, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:02:15,469 INFO [train.py:904] (2/8) Epoch 24, batch 2750, loss[loss=0.1421, simple_loss=0.2327, pruned_loss=0.02577, over 16776.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2576, pruned_loss=0.03986, over 3330955.44 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:02:44,667 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 24, batch 2800, loss[loss=0.1814, simple_loss=0.2614, pruned_loss=0.05072, over 16389.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.0393, over 3334882.69 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:04:32,914 INFO [train.py:904] (2/8) Epoch 24, batch 2850, loss[loss=0.1891, simple_loss=0.281, pruned_loss=0.04857, over 16775.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2561, pruned_loss=0.03883, over 3332239.36 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:04:45,989 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-05-01 19:05:04,031 INFO [optim.py:368] (2/8) 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 19:05:17,504 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:05:42,362 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2304, 5.1131, 5.0701, 4.5723, 4.6911, 5.0995, 5.0453, 4.6839], device='cuda:2'), covar=tensor([0.0598, 0.0539, 0.0357, 0.0383, 0.1220, 0.0510, 0.0383, 0.0860], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0467, 0.0367, 0.0367, 0.0372, 0.0425, 0.0250, 0.0443], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 19:05:43,130 INFO [train.py:904] (2/8) Epoch 24, batch 2900, loss[loss=0.1675, simple_loss=0.2669, pruned_loss=0.0341, over 17265.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2556, pruned_loss=0.03938, over 3330595.90 frames. ], batch size: 52, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:06:13,304 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:25,482 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236383.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:53,578 INFO [train.py:904] (2/8) Epoch 24, batch 2950, loss[loss=0.1728, simple_loss=0.2527, pruned_loss=0.04646, over 16802.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2561, pruned_loss=0.04044, over 3336209.64 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:07:24,081 INFO [optim.py:368] (2/8) 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 19:07:54,680 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:08:02,709 INFO [train.py:904] (2/8) Epoch 24, batch 3000, loss[loss=0.1448, simple_loss=0.2326, pruned_loss=0.02846, over 17192.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2561, pruned_loss=0.04078, over 3331192.73 frames. ], batch size: 44, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:08:02,710 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 19:08:12,048 INFO [train.py:938] (2/8) Epoch 24, validation: loss=0.1342, simple_loss=0.2393, pruned_loss=0.0145, over 944034.00 frames. 2023-05-01 19:08:12,048 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 19:08:36,949 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3997, 3.4649, 3.9409, 2.2711, 3.1998, 2.3824, 3.8772, 3.7176], device='cuda:2'), covar=tensor([0.0266, 0.0991, 0.0515, 0.1975, 0.0829, 0.1055, 0.0555, 0.0999], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:09:12,138 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 24, batch 3050, loss[loss=0.1797, simple_loss=0.2614, pruned_loss=0.04896, over 16355.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2563, pruned_loss=0.04098, over 3322911.62 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:09:32,872 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9141, 5.0005, 5.2154, 4.9949, 5.0289, 5.6703, 5.2096, 4.8504], device='cuda:2'), covar=tensor([0.1360, 0.2283, 0.2608, 0.2228, 0.2916, 0.1115, 0.1829, 0.2645], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0620, 0.0683, 0.0513, 0.0683, 0.0715, 0.0534, 0.0680], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 19:09:43,673 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6705, 6.0795, 5.7976, 5.8597, 5.4688, 5.4354, 5.4772, 6.2256], device='cuda:2'), covar=tensor([0.1312, 0.0986, 0.1151, 0.0916, 0.0943, 0.0731, 0.1307, 0.0905], device='cuda:2'), in_proj_covar=tensor([0.0713, 0.0871, 0.0715, 0.0670, 0.0553, 0.0558, 0.0730, 0.0682], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:09:53,423 INFO [optim.py:368] (2/8) 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 19:10:32,486 INFO [train.py:904] (2/8) Epoch 24, batch 3100, loss[loss=0.1759, simple_loss=0.249, pruned_loss=0.05142, over 16728.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2562, pruned_loss=0.04082, over 3329379.74 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:11:43,887 INFO [train.py:904] (2/8) Epoch 24, batch 3150, loss[loss=0.1664, simple_loss=0.2503, pruned_loss=0.0412, over 16862.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2542, pruned_loss=0.04012, over 3336982.10 frames. ], batch size: 96, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:11:45,987 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 19:12:13,744 INFO [optim.py:368] (2/8) 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 19:12:52,317 INFO [train.py:904] (2/8) Epoch 24, batch 3200, loss[loss=0.169, simple_loss=0.2612, pruned_loss=0.03841, over 16700.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2538, pruned_loss=0.03988, over 3346162.82 frames. ], batch size: 62, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:13:15,013 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5764, 5.9667, 5.7166, 5.7620, 5.3608, 5.3405, 5.3310, 6.0837], device='cuda:2'), covar=tensor([0.1310, 0.0872, 0.0921, 0.0769, 0.0930, 0.0741, 0.1182, 0.0864], device='cuda:2'), in_proj_covar=tensor([0.0711, 0.0868, 0.0712, 0.0666, 0.0552, 0.0554, 0.0727, 0.0677], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:13:15,169 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5619, 2.9850, 3.0494, 5.0075, 4.3072, 4.4921, 1.5658, 3.4892], device='cuda:2'), covar=tensor([0.1493, 0.0750, 0.1075, 0.0169, 0.0205, 0.0319, 0.1700, 0.0667], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0177, 0.0197, 0.0197, 0.0207, 0.0218, 0.0205, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:13:21,971 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:14:01,576 INFO [train.py:904] (2/8) Epoch 24, batch 3250, loss[loss=0.1811, simple_loss=0.2661, pruned_loss=0.04804, over 16534.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2535, pruned_loss=0.03938, over 3348918.74 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:14:26,977 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:14:31,669 INFO [optim.py:368] (2/8) 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 19:15:07,272 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 19:15:11,539 INFO [train.py:904] (2/8) Epoch 24, batch 3300, loss[loss=0.1796, simple_loss=0.261, pruned_loss=0.04904, over 16892.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2541, pruned_loss=0.03924, over 3347869.97 frames. ], batch size: 109, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:21,167 INFO [train.py:904] (2/8) Epoch 24, batch 3350, loss[loss=0.1663, simple_loss=0.2507, pruned_loss=0.041, over 16813.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2549, pruned_loss=0.03971, over 3344535.20 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:25,291 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3434, 2.6263, 2.0883, 2.2979, 2.9199, 2.6420, 3.0511, 3.1179], device='cuda:2'), covar=tensor([0.0253, 0.0515, 0.0628, 0.0548, 0.0300, 0.0463, 0.0284, 0.0286], device='cuda:2'), in_proj_covar=tensor([0.0229, 0.0245, 0.0234, 0.0236, 0.0245, 0.0246, 0.0247, 0.0243], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:16:51,747 INFO [optim.py:368] (2/8) 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 19:17:26,782 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-01 19:17:33,208 INFO [train.py:904] (2/8) Epoch 24, batch 3400, loss[loss=0.1395, simple_loss=0.2263, pruned_loss=0.02629, over 17196.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2553, pruned_loss=0.03968, over 3330772.38 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:18:31,041 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236892.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:18:37,303 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 19:18:45,125 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-05-01 19:18:45,598 INFO [train.py:904] (2/8) Epoch 24, batch 3450, loss[loss=0.1711, simple_loss=0.2488, pruned_loss=0.04677, over 16663.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2539, pruned_loss=0.03935, over 3319856.56 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:05,250 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9981, 5.3846, 5.1245, 5.1412, 4.8619, 4.8033, 4.8172, 5.4470], device='cuda:2'), covar=tensor([0.1391, 0.0915, 0.1084, 0.0894, 0.0903, 0.1126, 0.1196, 0.0999], device='cuda:2'), in_proj_covar=tensor([0.0718, 0.0873, 0.0720, 0.0673, 0.0557, 0.0559, 0.0733, 0.0684], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:19:15,853 INFO [optim.py:368] (2/8) 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 19:19:51,372 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 19:19:55,267 INFO [train.py:904] (2/8) Epoch 24, batch 3500, loss[loss=0.1877, simple_loss=0.2712, pruned_loss=0.05211, over 16486.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2537, pruned_loss=0.0392, over 3311712.41 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:56,896 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236953.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 19:21:06,715 INFO [train.py:904] (2/8) Epoch 24, batch 3550, loss[loss=0.1437, simple_loss=0.2331, pruned_loss=0.02714, over 17235.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2521, pruned_loss=0.03865, over 3303616.37 frames. ], batch size: 45, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:21:35,731 INFO [optim.py:368] (2/8) 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 19:21:47,360 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0767, 2.5592, 1.9008, 2.3217, 2.8449, 2.6694, 2.9750, 3.0391], device='cuda:2'), covar=tensor([0.0228, 0.0450, 0.0675, 0.0500, 0.0299, 0.0381, 0.0287, 0.0295], device='cuda:2'), in_proj_covar=tensor([0.0228, 0.0243, 0.0233, 0.0235, 0.0244, 0.0244, 0.0246, 0.0242], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:22:15,045 INFO [train.py:904] (2/8) Epoch 24, batch 3600, loss[loss=0.181, simple_loss=0.2564, pruned_loss=0.05279, over 11772.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2517, pruned_loss=0.03884, over 3282175.89 frames. ], batch size: 246, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:23:26,120 INFO [train.py:904] (2/8) Epoch 24, batch 3650, loss[loss=0.1716, simple_loss=0.2412, pruned_loss=0.05104, over 16538.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2504, pruned_loss=0.03905, over 3286951.35 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:23:58,725 INFO [optim.py:368] (2/8) 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 19:24:25,231 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.44 vs. limit=2.0 2023-05-01 19:24:39,858 INFO [train.py:904] (2/8) Epoch 24, batch 3700, loss[loss=0.1757, simple_loss=0.2501, pruned_loss=0.05068, over 16717.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2489, pruned_loss=0.0406, over 3284321.81 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:25:53,325 INFO [train.py:904] (2/8) Epoch 24, batch 3750, loss[loss=0.1722, simple_loss=0.2425, pruned_loss=0.05094, over 16893.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2496, pruned_loss=0.04222, over 3284384.77 frames. ], batch size: 116, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:26:10,810 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8450, 4.9023, 5.1095, 4.9618, 4.9759, 5.5734, 5.0688, 4.7445], device='cuda:2'), covar=tensor([0.1274, 0.2080, 0.2287, 0.2134, 0.2671, 0.1012, 0.1802, 0.2612], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0624, 0.0686, 0.0516, 0.0684, 0.0716, 0.0539, 0.0679], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 19:26:25,690 INFO [optim.py:368] (2/8) 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 19:26:57,489 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237248.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:27:03,606 INFO [train.py:904] (2/8) Epoch 24, batch 3800, loss[loss=0.1747, simple_loss=0.2582, pruned_loss=0.04562, over 16649.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2508, pruned_loss=0.04334, over 3278743.95 frames. ], batch size: 134, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:28:05,924 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 19:28:20,060 INFO [train.py:904] (2/8) Epoch 24, batch 3850, loss[loss=0.187, simple_loss=0.2553, pruned_loss=0.05938, over 16763.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2515, pruned_loss=0.04427, over 3281422.38 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:28:52,885 INFO [optim.py:368] (2/8) 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 19:29:30,846 INFO [train.py:904] (2/8) Epoch 24, batch 3900, loss[loss=0.1734, simple_loss=0.2482, pruned_loss=0.04932, over 16770.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2511, pruned_loss=0.04531, over 3291770.21 frames. ], batch size: 102, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:29:59,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1070, 4.1221, 4.4258, 4.3969, 4.4304, 4.1730, 4.2109, 4.1314], device='cuda:2'), covar=tensor([0.0415, 0.0707, 0.0431, 0.0436, 0.0561, 0.0490, 0.0795, 0.0638], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0483, 0.0467, 0.0430, 0.0512, 0.0487, 0.0573, 0.0391], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 19:30:43,384 INFO [train.py:904] (2/8) Epoch 24, batch 3950, loss[loss=0.1589, simple_loss=0.2263, pruned_loss=0.04574, over 16881.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2516, pruned_loss=0.04603, over 3277775.24 frames. ], batch size: 109, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:30:52,182 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 19:31:17,989 INFO [optim.py:368] (2/8) 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 19:31:41,558 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 19:31:53,148 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 19:31:57,060 INFO [train.py:904] (2/8) Epoch 24, batch 4000, loss[loss=0.1697, simple_loss=0.2506, pruned_loss=0.04444, over 16827.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2516, pruned_loss=0.04637, over 3276998.69 frames. ], batch size: 90, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:32:44,773 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6269, 5.6774, 5.4274, 4.7167, 5.6427, 2.3154, 5.3548, 5.0477], device='cuda:2'), covar=tensor([0.0053, 0.0040, 0.0148, 0.0359, 0.0057, 0.2651, 0.0086, 0.0217], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0169, 0.0211, 0.0187, 0.0187, 0.0216, 0.0199, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:32:53,198 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237491.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:33:10,917 INFO [train.py:904] (2/8) Epoch 24, batch 4050, loss[loss=0.1556, simple_loss=0.2484, pruned_loss=0.03136, over 16857.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2524, pruned_loss=0.04563, over 3270599.14 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:33:43,736 INFO [optim.py:368] (2/8) 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 19:33:58,578 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-05-01 19:34:17,620 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=237548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:34:23,852 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237552.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:34:24,476 INFO [train.py:904] (2/8) Epoch 24, batch 4100, loss[loss=0.1808, simple_loss=0.266, pruned_loss=0.0478, over 16661.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2539, pruned_loss=0.0452, over 3266568.49 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:34:34,586 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7224, 5.0300, 4.8266, 4.8154, 4.5476, 4.4831, 4.4532, 5.1403], device='cuda:2'), covar=tensor([0.1339, 0.0805, 0.1031, 0.0834, 0.0818, 0.1205, 0.1051, 0.0780], device='cuda:2'), in_proj_covar=tensor([0.0710, 0.0861, 0.0711, 0.0667, 0.0550, 0.0553, 0.0727, 0.0674], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:34:38,869 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237563.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:35:01,012 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0521, 2.2856, 2.6826, 3.1374, 3.0575, 3.5372, 2.2853, 3.4897], device='cuda:2'), covar=tensor([0.0248, 0.0495, 0.0321, 0.0278, 0.0286, 0.0181, 0.0550, 0.0137], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0195, 0.0184, 0.0189, 0.0204, 0.0163, 0.0201, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:35:30,060 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=237596.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:35:40,173 INFO [train.py:904] (2/8) Epoch 24, batch 4150, loss[loss=0.2192, simple_loss=0.2938, pruned_loss=0.07228, over 11400.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2613, pruned_loss=0.04794, over 3218738.92 frames. ], batch size: 248, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:36:14,090 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 19:36:56,393 INFO [train.py:904] (2/8) Epoch 24, batch 4200, loss[loss=0.1941, simple_loss=0.2927, pruned_loss=0.04782, over 16111.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2681, pruned_loss=0.04919, over 3196810.77 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:38:10,756 INFO [train.py:904] (2/8) Epoch 24, batch 4250, loss[loss=0.1812, simple_loss=0.2745, pruned_loss=0.04392, over 16271.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.272, pruned_loss=0.04967, over 3154102.66 frames. ], batch size: 35, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:38:16,747 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-05-01 19:38:45,324 INFO [optim.py:368] (2/8) 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 19:38:48,881 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4951, 3.5094, 2.6517, 2.2063, 2.3286, 2.3743, 3.6378, 3.1186], device='cuda:2'), covar=tensor([0.3119, 0.0721, 0.2011, 0.2971, 0.2727, 0.2160, 0.0582, 0.1404], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0272, 0.0309, 0.0320, 0.0304, 0.0267, 0.0299, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 19:39:16,372 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 19:39:26,152 INFO [train.py:904] (2/8) Epoch 24, batch 4300, loss[loss=0.189, simple_loss=0.2862, pruned_loss=0.0459, over 16870.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2725, pruned_loss=0.04805, over 3164974.16 frames. ], batch size: 109, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:40:01,108 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 19:40:23,424 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237790.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:40:41,735 INFO [train.py:904] (2/8) Epoch 24, batch 4350, loss[loss=0.1839, simple_loss=0.2775, pruned_loss=0.04514, over 15359.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2757, pruned_loss=0.04909, over 3165951.01 frames. ], batch size: 191, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:41:14,400 INFO [optim.py:368] (2/8) 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 19:41:33,071 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7009, 4.7668, 5.0460, 5.0119, 5.0492, 4.7404, 4.7013, 4.4670], device='cuda:2'), covar=tensor([0.0279, 0.0448, 0.0286, 0.0356, 0.0392, 0.0355, 0.0782, 0.0495], device='cuda:2'), in_proj_covar=tensor([0.0418, 0.0468, 0.0453, 0.0417, 0.0498, 0.0474, 0.0555, 0.0381], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 19:41:45,863 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:52,205 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237851.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:54,782 INFO [train.py:904] (2/8) Epoch 24, batch 4400, loss[loss=0.1926, simple_loss=0.2818, pruned_loss=0.05167, over 16151.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2774, pruned_loss=0.05033, over 3164273.29 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:42:57,206 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 19:43:05,639 INFO [train.py:904] (2/8) Epoch 24, batch 4450, loss[loss=0.2037, simple_loss=0.2983, pruned_loss=0.05459, over 16803.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2807, pruned_loss=0.05137, over 3176194.30 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:18,862 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237912.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:43:28,554 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237919.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 19:43:38,899 INFO [optim.py:368] (2/8) 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 19:43:54,583 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0561, 5.0918, 5.4115, 5.4007, 5.4662, 5.0851, 5.0476, 4.7124], device='cuda:2'), covar=tensor([0.0275, 0.0424, 0.0329, 0.0304, 0.0420, 0.0328, 0.0962, 0.0487], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0464, 0.0450, 0.0413, 0.0496, 0.0470, 0.0551, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 19:44:16,772 INFO [train.py:904] (2/8) Epoch 24, batch 4500, loss[loss=0.1901, simple_loss=0.2833, pruned_loss=0.04845, over 16759.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2815, pruned_loss=0.0522, over 3192341.49 frames. ], batch size: 89, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:44:37,783 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2553, 2.4776, 2.3539, 3.9747, 2.2006, 2.7539, 2.4967, 2.5200], device='cuda:2'), covar=tensor([0.1353, 0.2974, 0.2857, 0.0570, 0.4017, 0.2175, 0.2930, 0.3228], device='cuda:2'), in_proj_covar=tensor([0.0411, 0.0461, 0.0376, 0.0332, 0.0442, 0.0530, 0.0431, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:44:46,712 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237973.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:45:02,464 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0561, 3.0775, 2.6194, 2.9517, 3.4571, 3.0441, 3.6154, 3.6550], device='cuda:2'), covar=tensor([0.0074, 0.0365, 0.0481, 0.0361, 0.0208, 0.0338, 0.0200, 0.0211], device='cuda:2'), in_proj_covar=tensor([0.0222, 0.0239, 0.0229, 0.0231, 0.0240, 0.0240, 0.0241, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:45:03,600 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2299, 5.2013, 4.9463, 3.6907, 5.1939, 1.7522, 4.8088, 4.4724], device='cuda:2'), covar=tensor([0.0088, 0.0079, 0.0213, 0.0636, 0.0086, 0.3696, 0.0135, 0.0384], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0183, 0.0212, 0.0195, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:45:32,218 INFO [train.py:904] (2/8) Epoch 24, batch 4550, loss[loss=0.1858, simple_loss=0.2725, pruned_loss=0.04958, over 16856.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2814, pruned_loss=0.05269, over 3199310.45 frames. ], batch size: 42, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:46:04,570 INFO [optim.py:368] (2/8) 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 19:46:35,267 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4477, 3.5309, 3.8311, 2.0164, 3.1761, 2.3434, 3.7319, 3.8277], device='cuda:2'), covar=tensor([0.0197, 0.0798, 0.0494, 0.2255, 0.0819, 0.1036, 0.0545, 0.0921], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0167, 0.0168, 0.0154, 0.0146, 0.0131, 0.0143, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:46:44,383 INFO [train.py:904] (2/8) Epoch 24, batch 4600, loss[loss=0.177, simple_loss=0.274, pruned_loss=0.04007, over 16915.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2823, pruned_loss=0.0528, over 3207342.67 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:47:00,807 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:15,208 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238074.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:28,189 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0758, 4.0667, 2.7364, 5.0219, 3.2562, 4.8429, 3.0261, 3.2991], device='cuda:2'), covar=tensor([0.0263, 0.0343, 0.1497, 0.0117, 0.0823, 0.0372, 0.1229, 0.0737], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0179, 0.0195, 0.0169, 0.0177, 0.0220, 0.0203, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:47:35,629 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9567, 5.4186, 5.6151, 5.2714, 5.3179, 5.9365, 5.4024, 5.1693], device='cuda:2'), covar=tensor([0.1011, 0.1730, 0.1945, 0.2182, 0.2751, 0.0932, 0.1473, 0.2335], device='cuda:2'), in_proj_covar=tensor([0.0415, 0.0608, 0.0665, 0.0503, 0.0665, 0.0697, 0.0523, 0.0663], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 19:47:56,651 INFO [train.py:904] (2/8) Epoch 24, batch 4650, loss[loss=0.1885, simple_loss=0.2734, pruned_loss=0.05184, over 17150.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2814, pruned_loss=0.05298, over 3215649.11 frames. ], batch size: 46, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:48:25,425 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-05-01 19:48:29,649 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 19:48:30,771 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238126.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:43,735 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:59,275 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:00,460 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:09,298 INFO [train.py:904] (2/8) Epoch 24, batch 4700, loss[loss=0.1679, simple_loss=0.2565, pruned_loss=0.03968, over 16567.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2789, pruned_loss=0.05179, over 3225299.58 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:49:58,158 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:09,158 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238195.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:12,806 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8989, 5.2084, 4.9961, 5.0381, 4.7490, 4.6850, 4.6069, 5.2954], device='cuda:2'), covar=tensor([0.1240, 0.0838, 0.0904, 0.0734, 0.0812, 0.0944, 0.1055, 0.0930], device='cuda:2'), in_proj_covar=tensor([0.0691, 0.0837, 0.0694, 0.0649, 0.0536, 0.0539, 0.0707, 0.0658], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 19:50:20,662 INFO [train.py:904] (2/8) Epoch 24, batch 4750, loss[loss=0.1727, simple_loss=0.2564, pruned_loss=0.04455, over 16693.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2746, pruned_loss=0.0499, over 3228259.89 frames. ], batch size: 57, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:50:43,665 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238219.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:50:53,741 INFO [optim.py:368] (2/8) 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 19:51:19,564 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6425, 2.7132, 2.6176, 4.7516, 3.4878, 4.1702, 1.7489, 2.9487], device='cuda:2'), covar=tensor([0.1447, 0.0837, 0.1238, 0.0137, 0.0261, 0.0371, 0.1601, 0.0890], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0194, 0.0205, 0.0216, 0.0204, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:51:26,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8628, 3.7936, 3.9407, 3.6990, 3.8553, 4.2687, 3.9276, 3.6020], device='cuda:2'), covar=tensor([0.2004, 0.2311, 0.2245, 0.2709, 0.2826, 0.1863, 0.1570, 0.2801], device='cuda:2'), in_proj_covar=tensor([0.0412, 0.0604, 0.0660, 0.0499, 0.0662, 0.0693, 0.0521, 0.0659], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 19:51:31,474 INFO [train.py:904] (2/8) Epoch 24, batch 4800, loss[loss=0.177, simple_loss=0.2733, pruned_loss=0.04039, over 16444.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2719, pruned_loss=0.04817, over 3207629.67 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:51:52,764 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238267.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:51:54,667 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238268.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:52:27,243 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 19:52:46,482 INFO [train.py:904] (2/8) Epoch 24, batch 4850, loss[loss=0.1719, simple_loss=0.2683, pruned_loss=0.03773, over 16611.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2726, pruned_loss=0.04718, over 3207023.22 frames. ], batch size: 76, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:53:22,733 INFO [optim.py:368] (2/8) 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 19:54:04,040 INFO [train.py:904] (2/8) Epoch 24, batch 4900, loss[loss=0.1799, simple_loss=0.2666, pruned_loss=0.0466, over 16638.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2723, pruned_loss=0.04623, over 3191516.86 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:54:08,245 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2758, 3.3623, 2.0399, 3.7105, 2.5362, 3.6964, 2.2092, 2.6908], device='cuda:2'), covar=tensor([0.0336, 0.0394, 0.1671, 0.0201, 0.0892, 0.0487, 0.1653, 0.0831], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0178, 0.0194, 0.0168, 0.0176, 0.0219, 0.0202, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:54:16,954 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 19:55:17,228 INFO [train.py:904] (2/8) Epoch 24, batch 4950, loss[loss=0.2092, simple_loss=0.3025, pruned_loss=0.05797, over 16775.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.272, pruned_loss=0.04566, over 3187277.83 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:41,591 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:41,744 INFO [zipformer.py:625] (2/8) 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] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:56:20,282 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238446.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:56:29,585 INFO [train.py:904] (2/8) Epoch 24, batch 5000, loss[loss=0.1749, simple_loss=0.2745, pruned_loss=0.03762, over 16463.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2738, pruned_loss=0.04548, over 3192638.83 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:57:11,123 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238481.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:57:11,981 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238482.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:30,534 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:42,732 INFO [train.py:904] (2/8) Epoch 24, batch 5050, loss[loss=0.1661, simple_loss=0.2565, pruned_loss=0.03785, over 16693.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2738, pruned_loss=0.04523, over 3206154.36 frames. ], batch size: 76, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:58:15,455 INFO [optim.py:368] (2/8) 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 19:58:15,914 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7120, 3.4777, 4.0345, 1.8207, 4.2134, 4.2181, 3.1641, 3.1005], device='cuda:2'), covar=tensor([0.0809, 0.0308, 0.0200, 0.1451, 0.0078, 0.0116, 0.0433, 0.0496], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0110, 0.0100, 0.0140, 0.0082, 0.0127, 0.0130, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 19:58:53,423 INFO [train.py:904] (2/8) Epoch 24, batch 5100, loss[loss=0.1794, simple_loss=0.2733, pruned_loss=0.04274, over 16512.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2722, pruned_loss=0.04497, over 3194945.23 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:59:16,694 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:02,284 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8232, 1.3919, 1.6652, 1.7389, 1.8492, 1.9117, 1.6243, 1.8115], device='cuda:2'), covar=tensor([0.0249, 0.0405, 0.0206, 0.0305, 0.0280, 0.0183, 0.0445, 0.0125], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0194, 0.0182, 0.0186, 0.0202, 0.0160, 0.0199, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:00:07,249 INFO [train.py:904] (2/8) Epoch 24, batch 5150, loss[loss=0.1581, simple_loss=0.2499, pruned_loss=0.03311, over 12120.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2723, pruned_loss=0.04424, over 3191969.12 frames. ], batch size: 246, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:00:26,118 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:38,785 INFO [optim.py:368] (2/8) 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 20:01:17,675 INFO [train.py:904] (2/8) Epoch 24, batch 5200, loss[loss=0.186, simple_loss=0.2747, pruned_loss=0.04869, over 15433.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2705, pruned_loss=0.04358, over 3192183.92 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:01:27,346 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3213, 4.4070, 4.2118, 3.9327, 3.9271, 4.3026, 3.9852, 4.0577], device='cuda:2'), covar=tensor([0.0585, 0.0581, 0.0288, 0.0318, 0.0824, 0.0582, 0.0805, 0.0666], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0446, 0.0349, 0.0350, 0.0352, 0.0405, 0.0238, 0.0421], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:01:32,630 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 20:01:57,530 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:02:13,351 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 20:02:28,268 INFO [train.py:904] (2/8) Epoch 24, batch 5250, loss[loss=0.1836, simple_loss=0.2783, pruned_loss=0.04449, over 15289.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2681, pruned_loss=0.04334, over 3190715.87 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:02:51,374 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238720.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:01,284 INFO [optim.py:368] (2/8) 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 20:03:07,691 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238730.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:25,215 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:39,613 INFO [train.py:904] (2/8) Epoch 24, batch 5300, loss[loss=0.1443, simple_loss=0.2371, pruned_loss=0.02577, over 16944.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2648, pruned_loss=0.04199, over 3192607.83 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:04:00,434 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:12,298 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238776.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:04:15,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238778.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:20,293 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:50,566 INFO [train.py:904] (2/8) Epoch 24, batch 5350, loss[loss=0.19, simple_loss=0.288, pruned_loss=0.04602, over 15397.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.263, pruned_loss=0.04135, over 3207570.58 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:05:05,447 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3887, 4.4002, 4.2611, 3.2559, 4.3523, 1.5046, 4.0391, 3.9108], device='cuda:2'), covar=tensor([0.0135, 0.0137, 0.0227, 0.0640, 0.0140, 0.3614, 0.0182, 0.0353], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0164, 0.0205, 0.0182, 0.0180, 0.0211, 0.0193, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:05:21,928 INFO [optim.py:368] (2/8) 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 20:05:27,686 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238830.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:05:29,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0170, 3.7244, 4.2926, 2.0923, 4.4755, 4.4462, 3.3138, 3.4030], device='cuda:2'), covar=tensor([0.0652, 0.0253, 0.0143, 0.1180, 0.0051, 0.0109, 0.0367, 0.0404], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0109, 0.0100, 0.0139, 0.0082, 0.0126, 0.0128, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:05:59,848 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 20:06:01,798 INFO [train.py:904] (2/8) Epoch 24, batch 5400, loss[loss=0.2013, simple_loss=0.287, pruned_loss=0.05783, over 12245.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2656, pruned_loss=0.04184, over 3216867.18 frames. ], batch size: 247, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:18,015 INFO [train.py:904] (2/8) Epoch 24, batch 5450, loss[loss=0.1897, simple_loss=0.2794, pruned_loss=0.05002, over 16211.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2689, pruned_loss=0.0437, over 3209436.23 frames. ], batch size: 35, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:54,284 INFO [optim.py:368] (2/8) 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 20:08:10,990 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238936.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:08:35,281 INFO [train.py:904] (2/8) Epoch 24, batch 5500, loss[loss=0.2148, simple_loss=0.302, pruned_loss=0.06376, over 16816.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.276, pruned_loss=0.04823, over 3178422.03 frames. ], batch size: 39, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:09:43,452 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238997.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:09:53,111 INFO [train.py:904] (2/8) Epoch 24, batch 5550, loss[loss=0.1645, simple_loss=0.2575, pruned_loss=0.03579, over 16746.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2827, pruned_loss=0.05265, over 3152001.24 frames. ], batch size: 57, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:10:30,506 INFO [optim.py:368] (2/8) 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 20:10:48,340 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:11:00,065 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1215, 2.5029, 2.6254, 1.9790, 2.7218, 2.7911, 2.4283, 2.4249], device='cuda:2'), covar=tensor([0.0651, 0.0224, 0.0232, 0.0855, 0.0124, 0.0272, 0.0415, 0.0405], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0140, 0.0083, 0.0128, 0.0129, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:11:12,507 INFO [train.py:904] (2/8) Epoch 24, batch 5600, loss[loss=0.1762, simple_loss=0.2699, pruned_loss=0.04125, over 16468.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2875, pruned_loss=0.05688, over 3119619.55 frames. ], batch size: 75, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:11:23,831 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9690, 5.2434, 4.9917, 4.9807, 4.7753, 4.7202, 4.6859, 5.3425], device='cuda:2'), covar=tensor([0.1161, 0.0844, 0.1054, 0.0965, 0.0840, 0.0879, 0.1203, 0.0802], device='cuda:2'), in_proj_covar=tensor([0.0687, 0.0829, 0.0689, 0.0641, 0.0531, 0.0531, 0.0699, 0.0650], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:11:52,267 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239076.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:12:36,908 INFO [train.py:904] (2/8) Epoch 24, batch 5650, loss[loss=0.1963, simple_loss=0.2906, pruned_loss=0.05098, over 16700.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2923, pruned_loss=0.06063, over 3082640.27 frames. ], batch size: 89, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:13:06,268 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 20:13:11,109 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 20:13:55,697 INFO [train.py:904] (2/8) Epoch 24, batch 5700, loss[loss=0.2242, simple_loss=0.3082, pruned_loss=0.07011, over 15352.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2948, pruned_loss=0.06314, over 3053847.14 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:13:59,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2751, 2.8527, 3.0082, 1.9831, 2.7143, 2.0360, 3.0344, 3.0876], device='cuda:2'), covar=tensor([0.0301, 0.0855, 0.0592, 0.2041, 0.0907, 0.1102, 0.0642, 0.0862], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0165, 0.0167, 0.0153, 0.0145, 0.0130, 0.0143, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:14:28,099 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1769, 4.2549, 4.0752, 3.8094, 3.8139, 4.2022, 3.8699, 3.9742], device='cuda:2'), covar=tensor([0.0623, 0.0533, 0.0310, 0.0307, 0.0780, 0.0471, 0.0902, 0.0625], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0444, 0.0348, 0.0348, 0.0351, 0.0403, 0.0237, 0.0420], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:15:14,247 INFO [train.py:904] (2/8) Epoch 24, batch 5750, loss[loss=0.2218, simple_loss=0.2905, pruned_loss=0.07658, over 11233.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2983, pruned_loss=0.06597, over 2993074.51 frames. ], batch size: 250, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:53,491 INFO [optim.py:368] (2/8) 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 20:16:37,341 INFO [train.py:904] (2/8) Epoch 24, batch 5800, loss[loss=0.1785, simple_loss=0.2761, pruned_loss=0.04042, over 16714.00 frames. ], tot_loss[loss=0.212, simple_loss=0.297, pruned_loss=0.06349, over 3032593.49 frames. ], batch size: 89, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:17:20,321 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0379, 3.1147, 2.7610, 2.9846, 3.4621, 3.0766, 3.5651, 3.6419], device='cuda:2'), covar=tensor([0.0103, 0.0360, 0.0458, 0.0372, 0.0240, 0.0341, 0.0295, 0.0233], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0238, 0.0230, 0.0230, 0.0240, 0.0238, 0.0239, 0.0236], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:17:33,832 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1501, 3.1978, 1.9114, 3.4119, 2.4385, 3.4949, 2.1776, 2.6667], device='cuda:2'), covar=tensor([0.0320, 0.0422, 0.1754, 0.0290, 0.0858, 0.0546, 0.1568, 0.0796], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0169, 0.0179, 0.0219, 0.0205, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:17:40,219 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239292.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:17:56,784 INFO [train.py:904] (2/8) Epoch 24, batch 5850, loss[loss=0.2004, simple_loss=0.29, pruned_loss=0.05543, over 16673.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2942, pruned_loss=0.0617, over 3039105.40 frames. ], batch size: 62, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:18:33,902 INFO [optim.py:368] (2/8) 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 20:18:52,926 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239337.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:19:01,615 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0653, 3.3799, 3.6763, 1.8137, 3.8713, 3.9228, 3.0005, 2.6988], device='cuda:2'), covar=tensor([0.1258, 0.0234, 0.0207, 0.1377, 0.0087, 0.0160, 0.0444, 0.0643], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0109, 0.0100, 0.0138, 0.0082, 0.0127, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:19:19,104 INFO [train.py:904] (2/8) Epoch 24, batch 5900, loss[loss=0.1787, simple_loss=0.2805, pruned_loss=0.03845, over 16405.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2935, pruned_loss=0.06096, over 3052997.82 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:20:02,559 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 20:20:14,707 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239385.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:20:42,777 INFO [train.py:904] (2/8) Epoch 24, batch 5950, loss[loss=0.1878, simple_loss=0.2826, pruned_loss=0.04648, over 16721.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2943, pruned_loss=0.05978, over 3070784.56 frames. ], batch size: 83, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:21:13,187 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 20:21:21,253 INFO [optim.py:368] (2/8) 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 20:21:31,912 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6118, 2.5096, 2.2524, 3.4869, 2.3584, 3.6367, 1.4188, 2.6852], device='cuda:2'), covar=tensor([0.1461, 0.0800, 0.1359, 0.0195, 0.0183, 0.0431, 0.1851, 0.0878], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0176, 0.0195, 0.0193, 0.0204, 0.0215, 0.0203, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:22:03,333 INFO [train.py:904] (2/8) Epoch 24, batch 6000, loss[loss=0.1733, simple_loss=0.2666, pruned_loss=0.04004, over 16867.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.294, pruned_loss=0.05955, over 3063256.17 frames. ], batch size: 102, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:22:03,333 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 20:22:14,261 INFO [train.py:938] (2/8) Epoch 24, validation: loss=0.1493, simple_loss=0.2618, pruned_loss=0.01837, over 944034.00 frames. 2023-05-01 20:22:14,261 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 20:22:34,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6776, 3.8779, 2.9048, 2.3643, 2.6249, 2.5164, 4.2070, 3.4854], device='cuda:2'), covar=tensor([0.2982, 0.0621, 0.1823, 0.2724, 0.2642, 0.2024, 0.0454, 0.1300], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0272, 0.0308, 0.0320, 0.0301, 0.0266, 0.0299, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 20:23:19,142 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9386, 4.1879, 4.0072, 4.0429, 3.7400, 3.8110, 3.8232, 4.1819], device='cuda:2'), covar=tensor([0.1097, 0.0928, 0.0974, 0.0826, 0.0810, 0.1483, 0.0927, 0.1028], device='cuda:2'), in_proj_covar=tensor([0.0687, 0.0828, 0.0686, 0.0640, 0.0529, 0.0530, 0.0696, 0.0649], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:23:32,275 INFO [train.py:904] (2/8) Epoch 24, batch 6050, loss[loss=0.2317, simple_loss=0.2924, pruned_loss=0.08549, over 11424.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.292, pruned_loss=0.05866, over 3068945.43 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:24:09,880 INFO [optim.py:368] (2/8) 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 20:24:42,760 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-01 20:24:51,434 INFO [train.py:904] (2/8) Epoch 24, batch 6100, loss[loss=0.2174, simple_loss=0.2887, pruned_loss=0.07306, over 11699.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2908, pruned_loss=0.0577, over 3078276.81 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:25:56,136 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239592.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:26:14,091 INFO [train.py:904] (2/8) Epoch 24, batch 6150, loss[loss=0.194, simple_loss=0.2846, pruned_loss=0.05173, over 16472.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2892, pruned_loss=0.05694, over 3079574.17 frames. ], batch size: 75, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:26:45,887 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=239622.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:26:53,237 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239640.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:27:23,914 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 20:27:28,247 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2023-05-01 20:27:34,955 INFO [train.py:904] (2/8) Epoch 24, batch 6200, loss[loss=0.19, simple_loss=0.2765, pruned_loss=0.05173, over 16662.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2862, pruned_loss=0.05553, over 3099135.26 frames. ], batch size: 134, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:28:22,521 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=239683.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:28:52,025 INFO [train.py:904] (2/8) Epoch 24, batch 6250, loss[loss=0.1981, simple_loss=0.2921, pruned_loss=0.05198, over 16718.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2857, pruned_loss=0.05493, over 3121435.48 frames. ], batch size: 124, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:29:13,983 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4126, 2.9530, 2.7177, 2.2990, 2.2785, 2.2933, 2.9754, 2.8377], device='cuda:2'), covar=tensor([0.2508, 0.0719, 0.1594, 0.2587, 0.2337, 0.2199, 0.0545, 0.1433], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0273, 0.0309, 0.0321, 0.0302, 0.0267, 0.0300, 0.0343], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 20:29:29,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7998, 3.8801, 4.1192, 4.0821, 4.1098, 3.8952, 3.9023, 3.9046], device='cuda:2'), covar=tensor([0.0418, 0.0638, 0.0480, 0.0510, 0.0535, 0.0477, 0.0960, 0.0576], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0471, 0.0459, 0.0421, 0.0505, 0.0478, 0.0562, 0.0382], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 20:29:29,985 INFO [optim.py:368] (2/8) 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 20:29:43,107 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7691, 4.5967, 4.8254, 4.9730, 5.1310, 4.6213, 5.1431, 5.1409], device='cuda:2'), covar=tensor([0.2046, 0.1303, 0.1743, 0.0771, 0.0709, 0.1028, 0.0766, 0.0722], device='cuda:2'), in_proj_covar=tensor([0.0644, 0.0791, 0.0912, 0.0802, 0.0612, 0.0633, 0.0664, 0.0772], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:30:05,921 INFO [train.py:904] (2/8) Epoch 24, batch 6300, loss[loss=0.2151, simple_loss=0.2852, pruned_loss=0.07243, over 11617.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2861, pruned_loss=0.05524, over 3118481.08 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:24,179 INFO [train.py:904] (2/8) Epoch 24, batch 6350, loss[loss=0.2539, simple_loss=0.3092, pruned_loss=0.09932, over 11647.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2872, pruned_loss=0.05672, over 3096377.85 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:30,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3285, 4.3885, 4.2088, 3.9385, 3.9423, 4.3328, 4.0207, 4.0759], device='cuda:2'), covar=tensor([0.0567, 0.0521, 0.0295, 0.0304, 0.0788, 0.0447, 0.0716, 0.0558], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0437, 0.0343, 0.0343, 0.0346, 0.0397, 0.0235, 0.0412], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:32:03,926 INFO [optim.py:368] (2/8) 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 20:32:42,128 INFO [train.py:904] (2/8) Epoch 24, batch 6400, loss[loss=0.1721, simple_loss=0.2684, pruned_loss=0.03792, over 16901.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2879, pruned_loss=0.0581, over 3075710.66 frames. ], batch size: 96, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:32:59,819 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0077, 3.2620, 3.1971, 2.1001, 3.0186, 3.2170, 3.0362, 1.9450], device='cuda:2'), covar=tensor([0.0571, 0.0072, 0.0075, 0.0463, 0.0135, 0.0143, 0.0126, 0.0502], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0087, 0.0087, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 20:33:56,440 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 20:33:58,189 INFO [train.py:904] (2/8) Epoch 24, batch 6450, loss[loss=0.2175, simple_loss=0.2855, pruned_loss=0.07472, over 11403.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2875, pruned_loss=0.05719, over 3079800.82 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:34:04,632 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9887, 2.3454, 1.9231, 2.1121, 2.6989, 2.3071, 2.6066, 2.9094], device='cuda:2'), covar=tensor([0.0242, 0.0479, 0.0683, 0.0580, 0.0330, 0.0502, 0.0281, 0.0322], device='cuda:2'), in_proj_covar=tensor([0.0217, 0.0237, 0.0228, 0.0229, 0.0239, 0.0236, 0.0236, 0.0235], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:34:15,309 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 20:34:37,420 INFO [optim.py:368] (2/8) 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 20:35:16,099 INFO [train.py:904] (2/8) Epoch 24, batch 6500, loss[loss=0.1771, simple_loss=0.2742, pruned_loss=0.03997, over 16813.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2853, pruned_loss=0.05602, over 3091250.87 frames. ], batch size: 102, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:35:55,168 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:36:24,750 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7600, 2.5078, 2.2782, 3.9934, 2.2948, 3.9170, 1.6323, 2.7950], device='cuda:2'), covar=tensor([0.1466, 0.0920, 0.1420, 0.0194, 0.0261, 0.0471, 0.1767, 0.0913], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0177, 0.0197, 0.0194, 0.0206, 0.0217, 0.0204, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:36:39,173 INFO [train.py:904] (2/8) Epoch 24, batch 6550, loss[loss=0.1892, simple_loss=0.3006, pruned_loss=0.03893, over 16886.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2878, pruned_loss=0.05697, over 3090410.56 frames. ], batch size: 96, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:37:16,973 INFO [optim.py:368] (2/8) 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 20:37:17,455 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9827, 3.1822, 3.1886, 2.0859, 2.9930, 3.2096, 3.0113, 2.0263], device='cuda:2'), covar=tensor([0.0568, 0.0082, 0.0079, 0.0465, 0.0125, 0.0127, 0.0120, 0.0466], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0086, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 20:37:54,308 INFO [train.py:904] (2/8) Epoch 24, batch 6600, loss[loss=0.2099, simple_loss=0.2988, pruned_loss=0.0605, over 16317.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2897, pruned_loss=0.05731, over 3087102.73 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:37:55,012 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9883, 4.4927, 3.3702, 2.7048, 3.2308, 2.8257, 4.9690, 3.9228], device='cuda:2'), covar=tensor([0.2861, 0.0541, 0.1643, 0.2575, 0.2324, 0.1902, 0.0336, 0.1139], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0272, 0.0306, 0.0319, 0.0300, 0.0265, 0.0299, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 20:38:44,876 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0680, 2.1668, 2.2909, 3.5890, 2.1278, 2.4940, 2.3115, 2.2858], device='cuda:2'), covar=tensor([0.1437, 0.3529, 0.2978, 0.0635, 0.4146, 0.2522, 0.3374, 0.3357], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0457, 0.0374, 0.0330, 0.0439, 0.0524, 0.0428, 0.0534], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:39:11,707 INFO [train.py:904] (2/8) Epoch 24, batch 6650, loss[loss=0.2054, simple_loss=0.2949, pruned_loss=0.05793, over 15312.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2899, pruned_loss=0.05788, over 3100729.74 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:46,091 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7945, 2.5255, 2.2621, 3.2966, 2.3409, 3.4799, 1.6433, 2.6477], device='cuda:2'), covar=tensor([0.1402, 0.0758, 0.1394, 0.0196, 0.0209, 0.0442, 0.1710, 0.0947], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0194, 0.0206, 0.0217, 0.0204, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:39:50,364 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 24, batch 6700, loss[loss=0.1976, simple_loss=0.2842, pruned_loss=0.05554, over 16803.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2895, pruned_loss=0.05838, over 3093122.76 frames. ], batch size: 83, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:41:04,965 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3207, 3.6220, 3.8884, 2.2311, 3.1969, 2.5408, 3.5916, 3.8474], device='cuda:2'), covar=tensor([0.0309, 0.0840, 0.0528, 0.2105, 0.0866, 0.0924, 0.0833, 0.1059], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0170, 0.0155, 0.0147, 0.0132, 0.0144, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:41:37,011 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6741, 3.4610, 4.1069, 2.0598, 4.2854, 4.2539, 3.0258, 3.1458], device='cuda:2'), covar=tensor([0.0805, 0.0289, 0.0176, 0.1183, 0.0064, 0.0141, 0.0473, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0138, 0.0082, 0.0128, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 20:41:45,726 INFO [train.py:904] (2/8) Epoch 24, batch 6750, loss[loss=0.1843, simple_loss=0.2735, pruned_loss=0.04757, over 16765.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2883, pruned_loss=0.05807, over 3121761.21 frames. ], batch size: 83, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:42:23,537 INFO [optim.py:368] (2/8) 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 20:42:51,568 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2535, 3.9520, 3.8960, 2.4424, 3.5335, 3.9348, 3.4976, 2.3935], device='cuda:2'), covar=tensor([0.0590, 0.0056, 0.0061, 0.0464, 0.0119, 0.0124, 0.0110, 0.0441], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0086, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 20:43:01,407 INFO [train.py:904] (2/8) Epoch 24, batch 6800, loss[loss=0.2032, simple_loss=0.2967, pruned_loss=0.0549, over 16302.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2884, pruned_loss=0.05766, over 3123967.27 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:43:42,281 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:44:21,119 INFO [train.py:904] (2/8) Epoch 24, batch 6850, loss[loss=0.2078, simple_loss=0.3055, pruned_loss=0.05507, over 16796.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.29, pruned_loss=0.05822, over 3126459.42 frames. ], batch size: 124, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:44:56,065 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 20:45:17,547 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:45:35,444 INFO [train.py:904] (2/8) Epoch 24, batch 6900, loss[loss=0.2405, simple_loss=0.3257, pruned_loss=0.07764, over 16939.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2919, pruned_loss=0.05754, over 3137484.57 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:46:50,844 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240401.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:46:52,761 INFO [train.py:904] (2/8) Epoch 24, batch 6950, loss[loss=0.2476, simple_loss=0.3182, pruned_loss=0.08849, over 11525.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2939, pruned_loss=0.05967, over 3107537.57 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:47:14,508 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0714, 2.0709, 2.6353, 2.9868, 2.8230, 3.3807, 2.2220, 3.3282], device='cuda:2'), covar=tensor([0.0218, 0.0573, 0.0365, 0.0311, 0.0350, 0.0181, 0.0585, 0.0162], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0194, 0.0181, 0.0184, 0.0200, 0.0160, 0.0198, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:47:33,340 INFO [optim.py:368] (2/8) 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 20:47:40,753 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 20:48:07,678 INFO [train.py:904] (2/8) Epoch 24, batch 7000, loss[loss=0.209, simple_loss=0.3029, pruned_loss=0.05749, over 16635.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.294, pruned_loss=0.05919, over 3096238.80 frames. ], batch size: 57, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:48:56,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9934, 4.7075, 4.5732, 3.1684, 4.0682, 4.6644, 4.0404, 2.6848], device='cuda:2'), covar=tensor([0.0454, 0.0047, 0.0051, 0.0368, 0.0100, 0.0108, 0.0084, 0.0428], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0086, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 20:49:23,902 INFO [train.py:904] (2/8) Epoch 24, batch 7050, loss[loss=0.1946, simple_loss=0.2878, pruned_loss=0.0507, over 16644.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2948, pruned_loss=0.0592, over 3088676.64 frames. ], batch size: 134, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:49:38,592 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1768, 4.0298, 4.2541, 4.3575, 4.4671, 4.0568, 4.4347, 4.4784], device='cuda:2'), covar=tensor([0.1657, 0.1146, 0.1300, 0.0633, 0.0558, 0.1321, 0.0778, 0.0675], device='cuda:2'), in_proj_covar=tensor([0.0634, 0.0783, 0.0901, 0.0789, 0.0606, 0.0626, 0.0658, 0.0763], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:49:57,881 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0804, 5.6320, 5.8440, 5.5137, 5.6212, 6.1460, 5.5659, 5.2655], device='cuda:2'), covar=tensor([0.0908, 0.1623, 0.2109, 0.1806, 0.2065, 0.0842, 0.1628, 0.2574], device='cuda:2'), in_proj_covar=tensor([0.0418, 0.0614, 0.0680, 0.0505, 0.0670, 0.0705, 0.0528, 0.0674], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 20:50:06,655 INFO [optim.py:368] (2/8) 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 20:50:15,806 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1344, 5.4094, 5.1692, 5.1729, 4.9446, 4.9155, 4.8339, 5.5232], device='cuda:2'), covar=tensor([0.1272, 0.0832, 0.0994, 0.0905, 0.0789, 0.0861, 0.1157, 0.0844], device='cuda:2'), in_proj_covar=tensor([0.0693, 0.0833, 0.0690, 0.0646, 0.0532, 0.0535, 0.0701, 0.0655], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:50:42,262 INFO [train.py:904] (2/8) Epoch 24, batch 7100, loss[loss=0.207, simple_loss=0.2926, pruned_loss=0.06064, over 16735.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2931, pruned_loss=0.05917, over 3073616.53 frames. ], batch size: 124, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:51:00,084 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5651, 1.7988, 2.2123, 2.5325, 2.5524, 2.8626, 1.9041, 2.7853], device='cuda:2'), covar=tensor([0.0235, 0.0539, 0.0335, 0.0363, 0.0322, 0.0218, 0.0578, 0.0156], device='cuda:2'), in_proj_covar=tensor([0.0193, 0.0195, 0.0182, 0.0185, 0.0201, 0.0160, 0.0199, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:51:42,350 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7441, 3.7460, 3.9222, 3.6619, 3.8738, 4.1998, 3.8923, 3.5926], device='cuda:2'), covar=tensor([0.2116, 0.2311, 0.2571, 0.2441, 0.2425, 0.1893, 0.1666, 0.2484], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0612, 0.0678, 0.0504, 0.0669, 0.0703, 0.0525, 0.0671], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 20:51:59,029 INFO [train.py:904] (2/8) Epoch 24, batch 7150, loss[loss=0.1941, simple_loss=0.2793, pruned_loss=0.05448, over 16683.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2912, pruned_loss=0.05906, over 3077374.57 frames. ], batch size: 62, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:52:39,163 INFO [optim.py:368] (2/8) 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 20:53:12,641 INFO [train.py:904] (2/8) Epoch 24, batch 7200, loss[loss=0.16, simple_loss=0.2594, pruned_loss=0.03029, over 16864.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2894, pruned_loss=0.05759, over 3056854.72 frames. ], batch size: 102, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:54:20,531 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=240696.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:54:32,007 INFO [train.py:904] (2/8) Epoch 24, batch 7250, loss[loss=0.2193, simple_loss=0.2918, pruned_loss=0.07334, over 11447.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2866, pruned_loss=0.05626, over 3051272.46 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:55:12,201 INFO [optim.py:368] (2/8) 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 20:55:43,014 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2597, 5.2624, 5.1072, 4.6603, 4.7137, 5.1508, 5.0724, 4.8104], device='cuda:2'), covar=tensor([0.0532, 0.0422, 0.0292, 0.0314, 0.1067, 0.0432, 0.0268, 0.0651], device='cuda:2'), in_proj_covar=tensor([0.0289, 0.0435, 0.0339, 0.0339, 0.0344, 0.0393, 0.0234, 0.0409], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 20:55:45,057 INFO [train.py:904] (2/8) Epoch 24, batch 7300, loss[loss=0.2044, simple_loss=0.2907, pruned_loss=0.05907, over 15389.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2859, pruned_loss=0.05595, over 3069863.59 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:57:02,841 INFO [train.py:904] (2/8) Epoch 24, batch 7350, loss[loss=0.1685, simple_loss=0.2604, pruned_loss=0.03823, over 16469.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2866, pruned_loss=0.05663, over 3063023.10 frames. ], batch size: 68, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:57:44,658 INFO [optim.py:368] (2/8) 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 20:58:18,665 INFO [train.py:904] (2/8) Epoch 24, batch 7400, loss[loss=0.2339, simple_loss=0.3042, pruned_loss=0.08181, over 11305.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2874, pruned_loss=0.05787, over 3033881.31 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:59:34,863 INFO [train.py:904] (2/8) Epoch 24, batch 7450, loss[loss=0.2315, simple_loss=0.3137, pruned_loss=0.07463, over 15346.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2888, pruned_loss=0.05869, over 3043926.64 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:00:19,560 INFO [optim.py:368] (2/8) 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 21:00:22,556 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240932.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:00:55,587 INFO [train.py:904] (2/8) Epoch 24, batch 7500, loss[loss=0.1647, simple_loss=0.2591, pruned_loss=0.03515, over 16833.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2888, pruned_loss=0.058, over 3041358.73 frames. ], batch size: 102, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:01:01,913 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8795, 4.8879, 4.6795, 3.9808, 4.7898, 1.8394, 4.5534, 4.4376], device='cuda:2'), covar=tensor([0.0091, 0.0073, 0.0190, 0.0383, 0.0084, 0.2645, 0.0113, 0.0223], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0163, 0.0205, 0.0181, 0.0179, 0.0211, 0.0192, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:01:45,296 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240986.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:01:57,402 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240993.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:02:00,840 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240996.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:02:11,647 INFO [train.py:904] (2/8) Epoch 24, batch 7550, loss[loss=0.2151, simple_loss=0.2853, pruned_loss=0.0725, over 11497.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2878, pruned_loss=0.05789, over 3043700.41 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:02:53,683 INFO [optim.py:368] (2/8) 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 21:03:15,800 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:20,281 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241047.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:29,246 INFO [train.py:904] (2/8) Epoch 24, batch 7600, loss[loss=0.1908, simple_loss=0.2855, pruned_loss=0.04806, over 16734.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2867, pruned_loss=0.05773, over 3073711.83 frames. ], batch size: 83, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:04:36,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9743, 2.3944, 2.3533, 2.9432, 1.9213, 3.2078, 1.7721, 2.6858], device='cuda:2'), covar=tensor([0.1238, 0.0620, 0.1094, 0.0215, 0.0153, 0.0390, 0.1556, 0.0736], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0193, 0.0206, 0.0216, 0.0205, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:04:47,282 INFO [train.py:904] (2/8) Epoch 24, batch 7650, loss[loss=0.2175, simple_loss=0.2932, pruned_loss=0.07091, over 17004.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2871, pruned_loss=0.05797, over 3075857.29 frames. ], batch size: 55, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:04:57,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3100, 3.0824, 3.4015, 1.8978, 3.5845, 3.5984, 2.7501, 2.7382], device='cuda:2'), covar=tensor([0.0885, 0.0294, 0.0221, 0.1215, 0.0089, 0.0215, 0.0523, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0110, 0.0100, 0.0140, 0.0083, 0.0129, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:05:30,616 INFO [optim.py:368] (2/8) 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 21:06:04,719 INFO [train.py:904] (2/8) Epoch 24, batch 7700, loss[loss=0.1784, simple_loss=0.2758, pruned_loss=0.04047, over 16666.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2874, pruned_loss=0.05861, over 3066286.93 frames. ], batch size: 89, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:07:23,956 INFO [train.py:904] (2/8) Epoch 24, batch 7750, loss[loss=0.223, simple_loss=0.2914, pruned_loss=0.07727, over 11898.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2869, pruned_loss=0.05721, over 3103794.10 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:06,386 INFO [optim.py:368] (2/8) 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 21:08:38,514 INFO [train.py:904] (2/8) Epoch 24, batch 7800, loss[loss=0.2179, simple_loss=0.302, pruned_loss=0.06687, over 16723.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2878, pruned_loss=0.0578, over 3106784.78 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:09:34,095 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241288.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:09:40,674 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8238, 3.0041, 2.8251, 4.7134, 3.4807, 4.1568, 1.8957, 3.0163], device='cuda:2'), covar=tensor([0.1423, 0.0773, 0.1159, 0.0196, 0.0324, 0.0401, 0.1610, 0.0860], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0193, 0.0205, 0.0216, 0.0204, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:09:56,311 INFO [train.py:904] (2/8) Epoch 24, batch 7850, loss[loss=0.1847, simple_loss=0.279, pruned_loss=0.04522, over 16718.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2886, pruned_loss=0.05807, over 3097434.96 frames. ], batch size: 89, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:10:08,586 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7736, 2.8827, 2.5689, 4.2539, 3.0460, 3.9770, 1.6275, 2.9196], device='cuda:2'), covar=tensor([0.1363, 0.0694, 0.1202, 0.0192, 0.0235, 0.0391, 0.1659, 0.0816], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0176, 0.0196, 0.0193, 0.0205, 0.0215, 0.0204, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:10:15,465 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5132, 3.3375, 2.7306, 2.1611, 2.2850, 2.3056, 3.5999, 3.1263], device='cuda:2'), covar=tensor([0.2937, 0.0806, 0.1866, 0.2800, 0.2528, 0.2236, 0.0497, 0.1387], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0271, 0.0308, 0.0319, 0.0300, 0.0266, 0.0299, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 21:10:38,475 INFO [optim.py:368] (2/8) 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 21:10:46,761 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 21:10:55,479 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241342.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:11:11,149 INFO [train.py:904] (2/8) Epoch 24, batch 7900, loss[loss=0.2284, simple_loss=0.2998, pruned_loss=0.0785, over 11355.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2882, pruned_loss=0.05765, over 3105983.92 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:11:51,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2415, 3.4187, 3.6803, 2.2161, 3.1578, 2.3617, 3.5931, 3.7467], device='cuda:2'), covar=tensor([0.0288, 0.0854, 0.0582, 0.2130, 0.0860, 0.1010, 0.0681, 0.1005], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0165, 0.0167, 0.0154, 0.0145, 0.0130, 0.0143, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:12:29,004 INFO [train.py:904] (2/8) Epoch 24, batch 7950, loss[loss=0.1987, simple_loss=0.2859, pruned_loss=0.05579, over 16279.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2885, pruned_loss=0.05784, over 3109709.20 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:12:35,022 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241406.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:13:12,617 INFO [optim.py:368] (2/8) 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 21:13:21,761 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6656, 2.4943, 2.3109, 3.3153, 2.2275, 3.5532, 1.5036, 2.7262], device='cuda:2'), covar=tensor([0.1443, 0.0765, 0.1273, 0.0215, 0.0153, 0.0392, 0.1795, 0.0817], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0178, 0.0197, 0.0194, 0.0206, 0.0217, 0.0205, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:13:22,945 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0859, 2.4860, 2.5558, 1.9542, 2.7207, 2.7813, 2.4290, 2.4268], device='cuda:2'), covar=tensor([0.0641, 0.0229, 0.0224, 0.0867, 0.0115, 0.0276, 0.0442, 0.0384], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0138, 0.0082, 0.0127, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:13:28,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6839, 4.3496, 4.2781, 2.9060, 3.7709, 4.3517, 3.7152, 2.5637], device='cuda:2'), covar=tensor([0.0529, 0.0061, 0.0053, 0.0401, 0.0116, 0.0116, 0.0117, 0.0460], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0086, 0.0087, 0.0135, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 21:13:38,985 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3157, 3.2556, 3.2272, 3.4340, 3.4138, 3.2366, 3.3748, 3.4416], device='cuda:2'), covar=tensor([0.1349, 0.1207, 0.1479, 0.0855, 0.1045, 0.3052, 0.1583, 0.1394], device='cuda:2'), in_proj_covar=tensor([0.0640, 0.0790, 0.0905, 0.0794, 0.0610, 0.0629, 0.0662, 0.0767], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:13:46,678 INFO [train.py:904] (2/8) Epoch 24, batch 8000, loss[loss=0.2253, simple_loss=0.2955, pruned_loss=0.07756, over 11132.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2899, pruned_loss=0.05937, over 3075276.11 frames. ], batch size: 249, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:14:09,595 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241467.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:14:31,317 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1544, 4.2225, 4.5381, 4.4929, 4.4972, 4.2140, 4.2081, 4.1825], device='cuda:2'), covar=tensor([0.0367, 0.0539, 0.0389, 0.0407, 0.0496, 0.0476, 0.0930, 0.0565], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0467, 0.0451, 0.0418, 0.0498, 0.0473, 0.0557, 0.0379], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 21:14:34,896 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2327, 4.3723, 4.4964, 4.2563, 4.3534, 4.8346, 4.3773, 4.1048], device='cuda:2'), covar=tensor([0.1788, 0.1925, 0.2301, 0.2260, 0.2565, 0.1093, 0.1684, 0.2579], device='cuda:2'), in_proj_covar=tensor([0.0417, 0.0613, 0.0677, 0.0502, 0.0665, 0.0700, 0.0527, 0.0670], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 21:15:04,288 INFO [train.py:904] (2/8) Epoch 24, batch 8050, loss[loss=0.2288, simple_loss=0.2965, pruned_loss=0.08058, over 11488.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2897, pruned_loss=0.0596, over 3061157.62 frames. ], batch size: 250, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:15:16,521 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9412, 4.1835, 3.9856, 4.0346, 3.7629, 3.8315, 3.8200, 4.1654], device='cuda:2'), covar=tensor([0.1118, 0.0903, 0.1122, 0.0865, 0.0829, 0.1494, 0.0966, 0.1005], device='cuda:2'), in_proj_covar=tensor([0.0694, 0.0829, 0.0693, 0.0646, 0.0528, 0.0535, 0.0701, 0.0652], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:15:47,625 INFO [optim.py:368] (2/8) 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 21:16:02,393 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8913, 2.7163, 2.6713, 1.9169, 2.5648, 2.7251, 2.5988, 1.9642], device='cuda:2'), covar=tensor([0.0474, 0.0102, 0.0096, 0.0410, 0.0150, 0.0129, 0.0129, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0086, 0.0086, 0.0135, 0.0099, 0.0111, 0.0095, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 21:16:08,993 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4028, 4.4243, 4.7926, 4.7454, 4.7706, 4.4715, 4.4693, 4.3577], device='cuda:2'), covar=tensor([0.0366, 0.0573, 0.0368, 0.0406, 0.0488, 0.0412, 0.0965, 0.0508], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0467, 0.0451, 0.0418, 0.0499, 0.0473, 0.0556, 0.0379], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 21:16:10,435 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1008, 3.3996, 3.3938, 2.2105, 3.1718, 3.4155, 3.1600, 2.0123], device='cuda:2'), covar=tensor([0.0578, 0.0068, 0.0071, 0.0473, 0.0115, 0.0109, 0.0116, 0.0482], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0086, 0.0086, 0.0135, 0.0099, 0.0111, 0.0095, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 21:16:22,138 INFO [train.py:904] (2/8) Epoch 24, batch 8100, loss[loss=0.1892, simple_loss=0.2793, pruned_loss=0.04951, over 17138.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2886, pruned_loss=0.05867, over 3053966.20 frames. ], batch size: 48, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:17:14,412 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241588.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:17:26,807 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 21:17:38,324 INFO [train.py:904] (2/8) Epoch 24, batch 8150, loss[loss=0.2105, simple_loss=0.2902, pruned_loss=0.06543, over 16387.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2865, pruned_loss=0.05775, over 3064500.64 frames. ], batch size: 35, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:18:06,777 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:21,043 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 21:18:22,058 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241636.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:18:39,382 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241642.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:55,849 INFO [train.py:904] (2/8) Epoch 24, batch 8200, loss[loss=0.2213, simple_loss=0.2906, pruned_loss=0.07603, over 11798.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2838, pruned_loss=0.0573, over 3059424.69 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:19:15,024 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 21:19:42,891 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:19:56,098 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241690.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:20:16,377 INFO [train.py:904] (2/8) Epoch 24, batch 8250, loss[loss=0.1933, simple_loss=0.2876, pruned_loss=0.04949, over 15452.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2832, pruned_loss=0.05514, over 3051441.44 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:20:29,512 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 21:20:39,820 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4152, 4.5169, 4.6510, 4.3951, 4.5022, 5.0349, 4.5398, 4.1896], device='cuda:2'), covar=tensor([0.1469, 0.2068, 0.2236, 0.2221, 0.2511, 0.1003, 0.1724, 0.2784], device='cuda:2'), in_proj_covar=tensor([0.0415, 0.0608, 0.0673, 0.0500, 0.0660, 0.0695, 0.0523, 0.0666], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 21:21:03,232 INFO [optim.py:368] (2/8) 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 21:21:33,432 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7365, 5.0098, 4.8098, 4.8229, 4.5686, 4.5906, 4.4450, 5.0614], device='cuda:2'), covar=tensor([0.1170, 0.0811, 0.0945, 0.0822, 0.0786, 0.0970, 0.1098, 0.0846], device='cuda:2'), in_proj_covar=tensor([0.0689, 0.0823, 0.0687, 0.0641, 0.0525, 0.0531, 0.0696, 0.0647], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:21:38,998 INFO [train.py:904] (2/8) Epoch 24, batch 8300, loss[loss=0.1838, simple_loss=0.2847, pruned_loss=0.04148, over 16362.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2807, pruned_loss=0.05217, over 3056295.33 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:54,683 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241762.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:23:02,690 INFO [train.py:904] (2/8) Epoch 24, batch 8350, loss[loss=0.202, simple_loss=0.2789, pruned_loss=0.06255, over 12007.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2806, pruned_loss=0.05015, over 3070994.32 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:23:03,964 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241803.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:23:48,197 INFO [optim.py:368] (2/8) 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 21:23:48,803 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241831.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:24:23,524 INFO [train.py:904] (2/8) Epoch 24, batch 8400, loss[loss=0.1746, simple_loss=0.2762, pruned_loss=0.03654, over 16914.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2782, pruned_loss=0.04805, over 3077894.53 frames. ], batch size: 116, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:24:42,781 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241864.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:25:17,141 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4590, 1.6570, 2.0812, 2.4257, 2.4296, 2.7151, 1.9643, 2.6764], device='cuda:2'), covar=tensor([0.0251, 0.0586, 0.0377, 0.0312, 0.0348, 0.0221, 0.0529, 0.0174], device='cuda:2'), in_proj_covar=tensor([0.0189, 0.0193, 0.0179, 0.0182, 0.0197, 0.0158, 0.0196, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:25:28,775 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241892.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:25:45,943 INFO [train.py:904] (2/8) Epoch 24, batch 8450, loss[loss=0.1623, simple_loss=0.2599, pruned_loss=0.03237, over 16700.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2759, pruned_loss=0.04631, over 3059944.67 frames. ], batch size: 89, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:26:31,269 INFO [optim.py:368] (2/8) 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 21:26:43,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9655, 4.2056, 4.0910, 4.0801, 3.7784, 3.8441, 3.8406, 4.2268], device='cuda:2'), covar=tensor([0.1229, 0.1008, 0.0975, 0.0875, 0.0828, 0.1632, 0.0996, 0.1031], device='cuda:2'), in_proj_covar=tensor([0.0684, 0.0821, 0.0683, 0.0638, 0.0522, 0.0529, 0.0692, 0.0644], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:26:45,872 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3893, 4.2073, 4.4107, 4.5526, 4.7477, 4.2889, 4.7052, 4.7377], device='cuda:2'), covar=tensor([0.1882, 0.1376, 0.1706, 0.0908, 0.0659, 0.1155, 0.0788, 0.0824], device='cuda:2'), in_proj_covar=tensor([0.0630, 0.0780, 0.0897, 0.0787, 0.0603, 0.0621, 0.0655, 0.0761], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:27:04,513 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2859, 5.5595, 5.3825, 5.3744, 5.1170, 5.1309, 4.9670, 5.6712], device='cuda:2'), covar=tensor([0.1329, 0.0966, 0.0983, 0.0850, 0.0839, 0.0694, 0.1155, 0.0819], device='cuda:2'), in_proj_covar=tensor([0.0684, 0.0821, 0.0682, 0.0638, 0.0522, 0.0530, 0.0692, 0.0645], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:27:07,283 INFO [train.py:904] (2/8) Epoch 24, batch 8500, loss[loss=0.1444, simple_loss=0.2317, pruned_loss=0.02854, over 11894.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2726, pruned_loss=0.04424, over 3055495.71 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:27:47,065 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:28:33,822 INFO [train.py:904] (2/8) Epoch 24, batch 8550, loss[loss=0.1785, simple_loss=0.2612, pruned_loss=0.04784, over 12014.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2702, pruned_loss=0.04324, over 3046480.40 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:29:26,914 INFO [optim.py:368] (2/8) 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 21:29:54,300 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8442, 5.1457, 4.9398, 4.9168, 4.6826, 4.6695, 4.6028, 5.2395], device='cuda:2'), covar=tensor([0.1134, 0.0946, 0.1004, 0.0885, 0.0842, 0.1093, 0.1112, 0.0990], device='cuda:2'), in_proj_covar=tensor([0.0677, 0.0812, 0.0675, 0.0632, 0.0517, 0.0525, 0.0685, 0.0639], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:30:13,213 INFO [train.py:904] (2/8) Epoch 24, batch 8600, loss[loss=0.1641, simple_loss=0.2491, pruned_loss=0.0396, over 12438.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2699, pruned_loss=0.04212, over 3046297.20 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:30:23,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0485, 5.4264, 5.5977, 5.3086, 5.4456, 5.9242, 5.3582, 5.0657], device='cuda:2'), covar=tensor([0.0845, 0.1649, 0.2188, 0.2066, 0.2115, 0.0818, 0.1524, 0.2320], device='cuda:2'), in_proj_covar=tensor([0.0406, 0.0598, 0.0660, 0.0492, 0.0646, 0.0684, 0.0513, 0.0654], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 21:30:31,476 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:30:49,719 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4958, 3.6318, 2.1297, 4.0116, 2.7167, 3.9383, 2.3833, 2.8943], device='cuda:2'), covar=tensor([0.0297, 0.0348, 0.1746, 0.0255, 0.0881, 0.0526, 0.1545, 0.0822], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0175, 0.0191, 0.0162, 0.0176, 0.0213, 0.0201, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:31:51,604 INFO [train.py:904] (2/8) Epoch 24, batch 8650, loss[loss=0.1791, simple_loss=0.2733, pruned_loss=0.04243, over 16868.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2682, pruned_loss=0.0407, over 3046621.79 frames. ], batch size: 116, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:32:10,578 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242110.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:32:22,427 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 21:32:56,837 INFO [optim.py:368] (2/8) 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 21:33:30,279 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8346, 4.9153, 5.2468, 5.2440, 5.2095, 4.9531, 4.8960, 4.7448], device='cuda:2'), covar=tensor([0.0348, 0.0528, 0.0388, 0.0380, 0.0520, 0.0374, 0.0831, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0413, 0.0464, 0.0450, 0.0414, 0.0495, 0.0469, 0.0552, 0.0377], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 21:33:37,184 INFO [train.py:904] (2/8) Epoch 24, batch 8700, loss[loss=0.1689, simple_loss=0.2631, pruned_loss=0.03733, over 16690.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2651, pruned_loss=0.03938, over 3055075.41 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:33:38,138 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8032, 4.9864, 5.1809, 4.8877, 5.0395, 5.5672, 5.0473, 4.7126], device='cuda:2'), covar=tensor([0.0942, 0.1825, 0.2255, 0.2078, 0.2377, 0.0881, 0.1497, 0.2285], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0596, 0.0656, 0.0489, 0.0643, 0.0682, 0.0511, 0.0651], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 21:33:38,214 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242153.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:33:49,873 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:34:26,704 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9594, 2.2854, 2.2944, 2.9612, 1.6628, 3.2355, 1.8030, 2.7565], device='cuda:2'), covar=tensor([0.1369, 0.0748, 0.1187, 0.0191, 0.0084, 0.0369, 0.1701, 0.0714], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0188, 0.0200, 0.0212, 0.0201, 0.0192], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:34:42,445 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242187.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 21:35:13,454 INFO [train.py:904] (2/8) Epoch 24, batch 8750, loss[loss=0.1622, simple_loss=0.2502, pruned_loss=0.0371, over 12233.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2647, pruned_loss=0.0386, over 3042151.31 frames. ], batch size: 250, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:35:32,218 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-05-01 21:35:42,289 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:10,121 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:21,483 INFO [optim.py:368] (2/8) 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 21:37:05,080 INFO [train.py:904] (2/8) Epoch 24, batch 8800, loss[loss=0.1651, simple_loss=0.2636, pruned_loss=0.03329, over 16243.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2632, pruned_loss=0.03751, over 3063166.22 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:37:09,795 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9775, 2.1289, 2.3880, 3.1847, 2.1657, 2.3306, 2.2994, 2.2374], device='cuda:2'), covar=tensor([0.1321, 0.3766, 0.2833, 0.0728, 0.4543, 0.2624, 0.3521, 0.3833], device='cuda:2'), in_proj_covar=tensor([0.0400, 0.0450, 0.0369, 0.0323, 0.0433, 0.0513, 0.0421, 0.0524], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:37:56,094 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242277.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:17,976 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242287.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:50,061 INFO [train.py:904] (2/8) Epoch 24, batch 8850, loss[loss=0.1692, simple_loss=0.2731, pruned_loss=0.03271, over 15324.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2666, pruned_loss=0.03723, over 3070158.67 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:39:03,905 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242309.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:38,582 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:52,554 INFO [optim.py:368] (2/8) 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 21:40:38,015 INFO [train.py:904] (2/8) Epoch 24, batch 8900, loss[loss=0.1667, simple_loss=0.2629, pruned_loss=0.03526, over 16766.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2665, pruned_loss=0.03681, over 3059868.94 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:41:01,390 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2404, 4.3738, 4.5274, 4.2859, 4.3936, 4.8394, 4.4257, 4.1289], device='cuda:2'), covar=tensor([0.1628, 0.1895, 0.2172, 0.2207, 0.2406, 0.1090, 0.1624, 0.2624], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0597, 0.0657, 0.0489, 0.0644, 0.0682, 0.0511, 0.0652], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 21:41:12,534 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:41:31,368 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8977, 4.8742, 4.6409, 3.8695, 4.7777, 2.0339, 4.4037, 4.4704], device='cuda:2'), covar=tensor([0.0099, 0.0081, 0.0230, 0.0440, 0.0110, 0.2581, 0.0163, 0.0272], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0160, 0.0198, 0.0175, 0.0175, 0.0206, 0.0188, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:42:44,463 INFO [train.py:904] (2/8) Epoch 24, batch 8950, loss[loss=0.1547, simple_loss=0.2535, pruned_loss=0.02791, over 17039.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.266, pruned_loss=0.03686, over 3075560.43 frames. ], batch size: 50, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:43:49,726 INFO [optim.py:368] (2/8) 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 21:44:35,825 INFO [train.py:904] (2/8) Epoch 24, batch 9000, loss[loss=0.1588, simple_loss=0.2471, pruned_loss=0.03526, over 16691.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2628, pruned_loss=0.03584, over 3066587.93 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:44:35,826 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 21:44:45,525 INFO [train.py:938] (2/8) Epoch 24, validation: loss=0.1445, simple_loss=0.2484, pruned_loss=0.02026, over 944034.00 frames. 2023-05-01 21:44:45,526 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 21:44:58,942 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:16,003 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4488, 4.4975, 4.3208, 3.9686, 4.0366, 4.4082, 4.1522, 4.1297], device='cuda:2'), covar=tensor([0.0646, 0.0552, 0.0349, 0.0362, 0.0914, 0.0553, 0.0592, 0.0683], device='cuda:2'), in_proj_covar=tensor([0.0286, 0.0427, 0.0333, 0.0334, 0.0335, 0.0386, 0.0231, 0.0401], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:45:28,220 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-05-01 21:45:48,573 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:58,409 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242487.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:46:30,345 INFO [train.py:904] (2/8) Epoch 24, batch 9050, loss[loss=0.1685, simple_loss=0.2543, pruned_loss=0.04137, over 16795.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.264, pruned_loss=0.0365, over 3061514.00 frames. ], batch size: 124, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:46:38,695 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9625, 2.1713, 2.3918, 3.2271, 2.1476, 2.3665, 2.3102, 2.2898], device='cuda:2'), covar=tensor([0.1313, 0.3552, 0.2726, 0.0714, 0.4442, 0.2485, 0.3569, 0.3344], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0449, 0.0369, 0.0323, 0.0433, 0.0511, 0.0421, 0.0522], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:46:40,247 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:46:45,327 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:47:15,076 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7161, 2.6471, 1.7082, 2.8579, 2.1166, 2.8585, 2.0156, 2.3975], device='cuda:2'), covar=tensor([0.0278, 0.0341, 0.1487, 0.0286, 0.0692, 0.0456, 0.1476, 0.0585], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0172, 0.0189, 0.0160, 0.0174, 0.0210, 0.0199, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 21:47:30,072 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242535.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:48:00,800 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242544.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:02,182 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5687, 3.5570, 3.5296, 2.8593, 3.4638, 2.0574, 3.2983, 2.9050], device='cuda:2'), covar=tensor([0.0159, 0.0150, 0.0200, 0.0236, 0.0119, 0.2490, 0.0153, 0.0264], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0160, 0.0199, 0.0175, 0.0176, 0.0206, 0.0188, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:48:17,898 INFO [train.py:904] (2/8) Epoch 24, batch 9100, loss[loss=0.1583, simple_loss=0.249, pruned_loss=0.03384, over 12259.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2637, pruned_loss=0.03681, over 3072096.14 frames. ], batch size: 247, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:48:23,333 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:27,335 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 21:49:27,657 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242582.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,333 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242602.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,963 INFO [train.py:904] (2/8) Epoch 24, batch 9150, loss[loss=0.1651, simple_loss=0.2528, pruned_loss=0.03874, over 16637.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2641, pruned_loss=0.03689, over 3048266.19 frames. ], batch size: 57, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:50:46,723 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:51:16,813 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5752, 3.5709, 3.5398, 2.8024, 3.4174, 2.0501, 3.2261, 2.8422], device='cuda:2'), covar=tensor([0.0203, 0.0157, 0.0233, 0.0240, 0.0163, 0.2405, 0.0168, 0.0286], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0159, 0.0198, 0.0174, 0.0175, 0.0205, 0.0187, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 21:51:21,271 INFO [optim.py:368] (2/8) 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 21:52:01,741 INFO [train.py:904] (2/8) Epoch 24, batch 9200, loss[loss=0.1307, simple_loss=0.2237, pruned_loss=0.01886, over 17055.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2595, pruned_loss=0.03562, over 3078374.48 frames. ], batch size: 50, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:52:21,715 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:52:24,776 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 24, batch 9250, loss[loss=0.1474, simple_loss=0.2453, pruned_loss=0.02478, over 15256.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.259, pruned_loss=0.03563, over 3062395.96 frames. ], batch size: 190, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:54:40,870 INFO [optim.py:368] (2/8) 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 21:55:27,520 INFO [train.py:904] (2/8) Epoch 24, batch 9300, loss[loss=0.1725, simple_loss=0.2738, pruned_loss=0.0356, over 15566.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2578, pruned_loss=0.03512, over 3076993.71 frames. ], batch size: 194, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:55:30,482 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-05-01 21:56:13,842 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:57:14,574 INFO [train.py:904] (2/8) Epoch 24, batch 9350, loss[loss=0.147, simple_loss=0.2344, pruned_loss=0.02976, over 16242.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2577, pruned_loss=0.03487, over 3090613.82 frames. ], batch size: 35, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:57:27,047 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242809.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:13,359 INFO [optim.py:368] (2/8) 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 21:58:14,172 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:23,837 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 21:58:25,326 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242839.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:54,546 INFO [train.py:904] (2/8) Epoch 24, batch 9400, loss[loss=0.1467, simple_loss=0.2341, pruned_loss=0.02963, over 12478.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2578, pruned_loss=0.03487, over 3066396.26 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:59:03,672 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:59:55,091 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:00:35,987 INFO [train.py:904] (2/8) Epoch 24, batch 9450, loss[loss=0.163, simple_loss=0.2546, pruned_loss=0.03565, over 16998.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2591, pruned_loss=0.03486, over 3054929.84 frames. ], batch size: 109, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:00:51,520 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:13,380 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:32,440 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:37,830 INFO [optim.py:368] (2/8) 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 22:02:16,737 INFO [train.py:904] (2/8) Epoch 24, batch 9500, loss[loss=0.1479, simple_loss=0.2371, pruned_loss=0.02934, over 13152.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2587, pruned_loss=0.03463, over 3063554.41 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:02:30,100 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:02:43,760 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:03:15,680 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242982.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:04:01,172 INFO [train.py:904] (2/8) Epoch 24, batch 9550, loss[loss=0.1735, simple_loss=0.2682, pruned_loss=0.0394, over 16719.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2586, pruned_loss=0.03484, over 3072984.05 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:04:24,733 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243013.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:05:06,082 INFO [optim.py:368] (2/8) 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 22:05:43,246 INFO [train.py:904] (2/8) Epoch 24, batch 9600, loss[loss=0.1887, simple_loss=0.2872, pruned_loss=0.04516, over 15363.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2603, pruned_loss=0.03577, over 3060909.86 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:06:11,034 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9107, 2.7320, 2.6594, 1.9586, 2.5203, 2.7914, 2.6517, 1.8945], device='cuda:2'), covar=tensor([0.0453, 0.0083, 0.0087, 0.0379, 0.0170, 0.0104, 0.0106, 0.0473], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0085, 0.0085, 0.0133, 0.0098, 0.0108, 0.0094, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 22:07:32,090 INFO [train.py:904] (2/8) Epoch 24, batch 9650, loss[loss=0.1766, simple_loss=0.2705, pruned_loss=0.04134, over 16659.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2627, pruned_loss=0.03633, over 3051929.13 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:08:24,605 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1315, 2.5757, 2.6702, 1.9537, 2.8111, 2.8567, 2.5050, 2.4539], device='cuda:2'), covar=tensor([0.0658, 0.0247, 0.0213, 0.0966, 0.0113, 0.0210, 0.0475, 0.0442], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0106, 0.0094, 0.0135, 0.0079, 0.0123, 0.0124, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-01 22:08:29,799 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243128.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:08:37,734 INFO [optim.py:368] (2/8) 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 22:08:50,139 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243139.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:09:19,548 INFO [train.py:904] (2/8) Epoch 24, batch 9700, loss[loss=0.162, simple_loss=0.2556, pruned_loss=0.03423, over 16650.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2608, pruned_loss=0.03587, over 3033912.77 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:10:30,927 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:01,230 INFO [train.py:904] (2/8) Epoch 24, batch 9750, loss[loss=0.1701, simple_loss=0.2678, pruned_loss=0.03623, over 15455.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2596, pruned_loss=0.03587, over 3031203.07 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:11:16,936 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:37,405 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9892, 2.7635, 2.6558, 2.0183, 2.5489, 2.8032, 2.6578, 1.9833], device='cuda:2'), covar=tensor([0.0417, 0.0081, 0.0081, 0.0345, 0.0157, 0.0101, 0.0114, 0.0446], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0084, 0.0084, 0.0132, 0.0098, 0.0107, 0.0093, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 22:12:03,356 INFO [optim.py:368] (2/8) 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 22:12:37,393 INFO [train.py:904] (2/8) Epoch 24, batch 9800, loss[loss=0.1891, simple_loss=0.2935, pruned_loss=0.04232, over 16779.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.26, pruned_loss=0.03517, over 3033844.13 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:12:49,115 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:12:50,362 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:13:10,024 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9062, 2.6532, 2.8306, 2.0315, 2.6722, 2.1287, 2.6639, 2.8389], device='cuda:2'), covar=tensor([0.0293, 0.0924, 0.0594, 0.1954, 0.0848, 0.0989, 0.0675, 0.0869], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0158, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 22:13:22,170 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243277.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 22:13:33,577 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-05-01 22:13:58,435 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=243293.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:14:21,702 INFO [train.py:904] (2/8) Epoch 24, batch 9850, loss[loss=0.1463, simple_loss=0.245, pruned_loss=0.02378, over 16723.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2617, pruned_loss=0.03507, over 3025021.05 frames. ], batch size: 76, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:14:27,023 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5945, 3.6483, 3.4446, 3.0968, 3.3036, 3.5722, 3.3495, 3.4304], device='cuda:2'), covar=tensor([0.0551, 0.0513, 0.0299, 0.0249, 0.0500, 0.0436, 0.1319, 0.0432], device='cuda:2'), in_proj_covar=tensor([0.0285, 0.0421, 0.0330, 0.0331, 0.0331, 0.0382, 0.0228, 0.0396], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:2') 2023-05-01 22:14:28,524 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243306.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:15:22,935 INFO [optim.py:368] (2/8) 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 22:15:54,146 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0302, 2.1608, 2.2335, 3.4467, 2.1168, 2.4009, 2.2741, 2.2742], device='cuda:2'), covar=tensor([0.1371, 0.3951, 0.3235, 0.0632, 0.4613, 0.2832, 0.3890, 0.3838], device='cuda:2'), in_proj_covar=tensor([0.0399, 0.0447, 0.0369, 0.0321, 0.0433, 0.0508, 0.0420, 0.0521], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:16:11,624 INFO [train.py:904] (2/8) Epoch 24, batch 9900, loss[loss=0.1533, simple_loss=0.2591, pruned_loss=0.02371, over 16673.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2614, pruned_loss=0.0346, over 3013935.71 frames. ], batch size: 76, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:16:12,852 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9881, 1.8711, 1.6660, 1.5057, 2.0212, 1.6421, 1.6595, 1.9515], device='cuda:2'), covar=tensor([0.0206, 0.0332, 0.0447, 0.0397, 0.0255, 0.0308, 0.0209, 0.0252], device='cuda:2'), in_proj_covar=tensor([0.0208, 0.0233, 0.0223, 0.0224, 0.0233, 0.0231, 0.0228, 0.0226], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:16:15,140 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=243354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:17:30,330 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 22:18:09,535 INFO [train.py:904] (2/8) Epoch 24, batch 9950, loss[loss=0.176, simple_loss=0.2766, pruned_loss=0.03767, over 16354.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2632, pruned_loss=0.03488, over 3026853.76 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:18:18,329 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8713, 3.1286, 3.4291, 1.9707, 2.9433, 2.2514, 3.4159, 3.3772], device='cuda:2'), covar=tensor([0.0216, 0.0858, 0.0530, 0.2084, 0.0781, 0.0983, 0.0596, 0.0904], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0158, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0170], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 22:19:11,939 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:19:24,675 INFO [optim.py:368] (2/8) 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 22:20:10,932 INFO [train.py:904] (2/8) Epoch 24, batch 10000, loss[loss=0.1641, simple_loss=0.2641, pruned_loss=0.03201, over 15340.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2617, pruned_loss=0.03444, over 3049202.84 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:20:36,494 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 22:20:55,590 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:21:51,123 INFO [train.py:904] (2/8) Epoch 24, batch 10050, loss[loss=0.1705, simple_loss=0.2682, pruned_loss=0.03645, over 16755.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.262, pruned_loss=0.03466, over 3060214.38 frames. ], batch size: 124, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:21:59,237 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 22:22:52,097 INFO [optim.py:368] (2/8) 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 22:23:25,168 INFO [train.py:904] (2/8) Epoch 24, batch 10100, loss[loss=0.1527, simple_loss=0.245, pruned_loss=0.03024, over 15481.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2625, pruned_loss=0.03495, over 3070558.05 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:24:15,452 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243577.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:25:09,202 INFO [train.py:904] (2/8) Epoch 25, batch 0, loss[loss=0.1583, simple_loss=0.2459, pruned_loss=0.0354, over 16803.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.2459, pruned_loss=0.0354, over 16803.00 frames. ], batch size: 42, lr: 2.72e-03, grad_scale: 8.0 2023-05-01 22:25:09,202 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 22:25:16,821 INFO [train.py:938] (2/8) Epoch 25, validation: loss=0.1443, simple_loss=0.2477, pruned_loss=0.02048, over 944034.00 frames. 2023-05-01 22:25:16,821 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 22:25:48,588 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 22:26:21,436 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243649.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:26:26,905 INFO [train.py:904] (2/8) Epoch 25, batch 50, loss[loss=0.189, simple_loss=0.2801, pruned_loss=0.04899, over 16778.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2675, pruned_loss=0.04568, over 745758.06 frames. ], batch size: 57, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:27:19,933 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 22:27:35,803 INFO [train.py:904] (2/8) Epoch 25, batch 100, loss[loss=0.1747, simple_loss=0.2756, pruned_loss=0.03692, over 17113.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.264, pruned_loss=0.04466, over 1313696.42 frames. ], batch size: 48, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:28:22,078 INFO [optim.py:368] (2/8) 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 22:28:45,144 INFO [train.py:904] (2/8) Epoch 25, batch 150, loss[loss=0.2288, simple_loss=0.3014, pruned_loss=0.07811, over 15497.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2604, pruned_loss=0.04318, over 1763584.61 frames. ], batch size: 191, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:28:50,545 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0256, 2.0025, 2.5775, 2.9123, 2.7501, 3.3575, 2.4712, 3.4998], device='cuda:2'), covar=tensor([0.0276, 0.0629, 0.0411, 0.0391, 0.0396, 0.0256, 0.0549, 0.0184], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0190, 0.0178, 0.0181, 0.0196, 0.0155, 0.0194, 0.0153], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:28:57,432 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0225, 5.0375, 5.4561, 5.4107, 5.4418, 5.1454, 5.0581, 4.8802], device='cuda:2'), covar=tensor([0.0376, 0.0519, 0.0400, 0.0463, 0.0551, 0.0427, 0.1000, 0.0475], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0450, 0.0441, 0.0406, 0.0483, 0.0461, 0.0537, 0.0369], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 22:29:55,569 INFO [train.py:904] (2/8) Epoch 25, batch 200, loss[loss=0.2043, simple_loss=0.2756, pruned_loss=0.06648, over 16855.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2615, pruned_loss=0.04376, over 2108329.94 frames. ], batch size: 116, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:30:15,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0318, 4.4574, 3.2119, 2.4371, 2.7724, 2.6362, 4.8143, 3.7551], device='cuda:2'), covar=tensor([0.2701, 0.0551, 0.1813, 0.2958, 0.2910, 0.2196, 0.0291, 0.1368], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0265, 0.0302, 0.0313, 0.0290, 0.0262, 0.0292, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 22:30:29,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2894, 4.5230, 3.4156, 2.7413, 3.1116, 2.9262, 4.9464, 3.8694], device='cuda:2'), covar=tensor([0.2462, 0.0606, 0.1713, 0.2638, 0.2635, 0.1920, 0.0335, 0.1373], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0265, 0.0302, 0.0312, 0.0290, 0.0262, 0.0292, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 22:30:34,309 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 22:30:40,582 INFO [optim.py:368] (2/8) 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 22:31:04,162 INFO [train.py:904] (2/8) Epoch 25, batch 250, loss[loss=0.1798, simple_loss=0.2519, pruned_loss=0.05384, over 16724.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2602, pruned_loss=0.04361, over 2369032.56 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:31:41,950 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2367, 5.2599, 5.0102, 4.5377, 5.0382, 1.8251, 4.8220, 4.9295], device='cuda:2'), covar=tensor([0.0088, 0.0082, 0.0224, 0.0408, 0.0104, 0.2904, 0.0135, 0.0216], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0162, 0.0200, 0.0175, 0.0177, 0.0209, 0.0189, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:31:57,534 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6739, 6.0399, 5.7882, 5.8063, 5.4263, 5.4445, 5.4359, 6.1745], device='cuda:2'), covar=tensor([0.1585, 0.1048, 0.1191, 0.0999, 0.0889, 0.0669, 0.1281, 0.0915], device='cuda:2'), in_proj_covar=tensor([0.0687, 0.0827, 0.0680, 0.0641, 0.0524, 0.0532, 0.0695, 0.0649], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:32:14,686 INFO [train.py:904] (2/8) Epoch 25, batch 300, loss[loss=0.1541, simple_loss=0.2501, pruned_loss=0.02908, over 17122.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2578, pruned_loss=0.04142, over 2587690.56 frames. ], batch size: 47, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:32:53,472 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4567, 5.4126, 5.2892, 4.7254, 4.8830, 5.3396, 5.3233, 4.9149], device='cuda:2'), covar=tensor([0.0590, 0.0539, 0.0327, 0.0367, 0.1096, 0.0452, 0.0232, 0.0778], device='cuda:2'), in_proj_covar=tensor([0.0293, 0.0437, 0.0339, 0.0342, 0.0343, 0.0393, 0.0234, 0.0410], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:33:00,865 INFO [optim.py:368] (2/8) 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 22:33:19,476 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243949.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:33:24,549 INFO [train.py:904] (2/8) Epoch 25, batch 350, loss[loss=0.1438, simple_loss=0.2359, pruned_loss=0.02584, over 16791.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2557, pruned_loss=0.04065, over 2751479.37 frames. ], batch size: 39, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:34:10,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7928, 1.8867, 2.3083, 2.6046, 2.6638, 2.6590, 1.9757, 2.8426], device='cuda:2'), covar=tensor([0.0186, 0.0512, 0.0355, 0.0314, 0.0323, 0.0309, 0.0542, 0.0194], device='cuda:2'), in_proj_covar=tensor([0.0189, 0.0192, 0.0180, 0.0183, 0.0198, 0.0157, 0.0195, 0.0155], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:34:25,842 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243997.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:34:38,353 INFO [train.py:904] (2/8) Epoch 25, batch 400, loss[loss=0.1813, simple_loss=0.257, pruned_loss=0.05279, over 16814.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2543, pruned_loss=0.04117, over 2867034.46 frames. ], batch size: 96, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:35:22,991 INFO [optim.py:368] (2/8) 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 22:35:47,032 INFO [train.py:904] (2/8) Epoch 25, batch 450, loss[loss=0.1622, simple_loss=0.2612, pruned_loss=0.0316, over 17114.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2529, pruned_loss=0.04082, over 2961696.55 frames. ], batch size: 48, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:36:55,204 INFO [train.py:904] (2/8) Epoch 25, batch 500, loss[loss=0.1666, simple_loss=0.239, pruned_loss=0.0471, over 16914.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2504, pruned_loss=0.03957, over 3038810.40 frames. ], batch size: 116, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:37:42,041 INFO [optim.py:368] (2/8) 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 22:38:05,719 INFO [train.py:904] (2/8) Epoch 25, batch 550, loss[loss=0.1503, simple_loss=0.2449, pruned_loss=0.02787, over 17119.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2499, pruned_loss=0.03891, over 3099901.00 frames. ], batch size: 49, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:39:15,774 INFO [train.py:904] (2/8) Epoch 25, batch 600, loss[loss=0.1761, simple_loss=0.2458, pruned_loss=0.05321, over 15522.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2489, pruned_loss=0.03912, over 3142793.27 frames. ], batch size: 190, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:39:27,462 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1990, 3.8601, 4.4226, 2.3289, 4.5430, 4.6425, 3.3821, 3.5637], device='cuda:2'), covar=tensor([0.0636, 0.0270, 0.0179, 0.1126, 0.0077, 0.0157, 0.0403, 0.0387], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0109, 0.0099, 0.0138, 0.0082, 0.0128, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 22:39:44,010 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0347, 5.0431, 5.4493, 5.4281, 5.4836, 5.1547, 5.0890, 4.9201], device='cuda:2'), covar=tensor([0.0358, 0.0656, 0.0501, 0.0466, 0.0452, 0.0447, 0.0934, 0.0454], device='cuda:2'), in_proj_covar=tensor([0.0415, 0.0463, 0.0451, 0.0416, 0.0495, 0.0474, 0.0550, 0.0378], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 22:40:02,986 INFO [optim.py:368] (2/8) 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 22:40:25,349 INFO [train.py:904] (2/8) Epoch 25, batch 650, loss[loss=0.1519, simple_loss=0.2404, pruned_loss=0.03173, over 17168.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2479, pruned_loss=0.03888, over 3169084.53 frames. ], batch size: 46, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:41:33,625 INFO [train.py:904] (2/8) Epoch 25, batch 700, loss[loss=0.1444, simple_loss=0.2354, pruned_loss=0.02664, over 16822.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2472, pruned_loss=0.0378, over 3208314.66 frames. ], batch size: 42, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:41:35,461 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 22:42:20,953 INFO [optim.py:368] (2/8) 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 22:42:42,413 INFO [train.py:904] (2/8) Epoch 25, batch 750, loss[loss=0.167, simple_loss=0.2649, pruned_loss=0.03454, over 17268.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2483, pruned_loss=0.03829, over 3231707.16 frames. ], batch size: 52, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:42:42,794 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:43:28,201 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2960, 4.3091, 4.6381, 4.6294, 4.6657, 4.3838, 4.3812, 4.3174], device='cuda:2'), covar=tensor([0.0390, 0.0793, 0.0426, 0.0428, 0.0506, 0.0462, 0.0864, 0.0548], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0470, 0.0457, 0.0421, 0.0502, 0.0481, 0.0558, 0.0383], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 22:43:52,150 INFO [train.py:904] (2/8) Epoch 25, batch 800, loss[loss=0.1846, simple_loss=0.258, pruned_loss=0.05563, over 16747.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2482, pruned_loss=0.03844, over 3255765.52 frames. ], batch size: 134, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:44:03,132 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 22:44:08,236 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:44:39,329 INFO [optim.py:368] (2/8) 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 22:45:03,354 INFO [train.py:904] (2/8) Epoch 25, batch 850, loss[loss=0.1646, simple_loss=0.2594, pruned_loss=0.03491, over 16686.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2476, pruned_loss=0.03808, over 3267880.67 frames. ], batch size: 57, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:45:11,495 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3740, 2.7128, 2.3224, 2.5619, 3.0038, 2.7949, 3.0728, 3.1655], device='cuda:2'), covar=tensor([0.0239, 0.0431, 0.0583, 0.0478, 0.0324, 0.0412, 0.0323, 0.0309], device='cuda:2'), in_proj_covar=tensor([0.0223, 0.0243, 0.0233, 0.0234, 0.0246, 0.0242, 0.0243, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:46:12,666 INFO [train.py:904] (2/8) Epoch 25, batch 900, loss[loss=0.1678, simple_loss=0.2588, pruned_loss=0.03839, over 16733.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2467, pruned_loss=0.03738, over 3274037.74 frames. ], batch size: 57, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:46:34,898 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3766, 4.1721, 4.4294, 4.5820, 4.6776, 4.2623, 4.5129, 4.6548], device='cuda:2'), covar=tensor([0.1792, 0.1301, 0.1444, 0.0696, 0.0653, 0.1149, 0.2499, 0.0944], device='cuda:2'), in_proj_covar=tensor([0.0665, 0.0818, 0.0941, 0.0828, 0.0632, 0.0649, 0.0683, 0.0793], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:47:00,724 INFO [optim.py:368] (2/8) 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 22:47:23,394 INFO [train.py:904] (2/8) Epoch 25, batch 950, loss[loss=0.1636, simple_loss=0.2524, pruned_loss=0.03742, over 16354.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2471, pruned_loss=0.03746, over 3288421.60 frames. ], batch size: 75, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:48:05,473 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7393, 1.7952, 1.6317, 1.4956, 1.9327, 1.6307, 1.6532, 1.9096], device='cuda:2'), covar=tensor([0.0268, 0.0338, 0.0513, 0.0464, 0.0257, 0.0346, 0.0224, 0.0261], device='cuda:2'), in_proj_covar=tensor([0.0224, 0.0245, 0.0234, 0.0235, 0.0246, 0.0244, 0.0245, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:48:33,896 INFO [train.py:904] (2/8) Epoch 25, batch 1000, loss[loss=0.168, simple_loss=0.2335, pruned_loss=0.05124, over 16767.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2463, pruned_loss=0.03781, over 3287359.96 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:49:21,001 INFO [optim.py:368] (2/8) 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 22:49:42,533 INFO [train.py:904] (2/8) Epoch 25, batch 1050, loss[loss=0.1298, simple_loss=0.2119, pruned_loss=0.02385, over 16774.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2459, pruned_loss=0.03784, over 3289682.93 frames. ], batch size: 39, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:50:53,097 INFO [train.py:904] (2/8) Epoch 25, batch 1100, loss[loss=0.1675, simple_loss=0.2551, pruned_loss=0.03992, over 16669.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2463, pruned_loss=0.03798, over 3298998.16 frames. ], batch size: 57, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:50:56,382 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-01 22:51:01,223 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=244709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:51:13,187 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244717.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:51:28,161 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7101, 5.0385, 4.8354, 4.8193, 4.5881, 4.5291, 4.4436, 5.1131], device='cuda:2'), covar=tensor([0.1411, 0.0963, 0.1075, 0.0927, 0.0787, 0.1279, 0.1221, 0.0904], device='cuda:2'), in_proj_covar=tensor([0.0707, 0.0853, 0.0699, 0.0659, 0.0540, 0.0547, 0.0718, 0.0668], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:51:40,295 INFO [optim.py:368] (2/8) 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 22:52:02,015 INFO [train.py:904] (2/8) Epoch 25, batch 1150, loss[loss=0.1696, simple_loss=0.2479, pruned_loss=0.04562, over 16729.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2461, pruned_loss=0.03747, over 3315062.70 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:52:37,158 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244778.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 22:52:48,213 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1865, 3.3202, 3.3644, 2.2628, 2.9332, 2.4981, 3.6356, 3.6727], device='cuda:2'), covar=tensor([0.0268, 0.0925, 0.0729, 0.2060, 0.0944, 0.1030, 0.0546, 0.0952], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0170, 0.0156, 0.0146, 0.0131, 0.0145, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 22:53:11,753 INFO [train.py:904] (2/8) Epoch 25, batch 1200, loss[loss=0.165, simple_loss=0.2417, pruned_loss=0.0441, over 15639.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2452, pruned_loss=0.03704, over 3312636.67 frames. ], batch size: 190, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:53:57,424 INFO [optim.py:368] (2/8) 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 22:54:19,233 INFO [train.py:904] (2/8) Epoch 25, batch 1250, loss[loss=0.1976, simple_loss=0.2729, pruned_loss=0.06117, over 16447.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2446, pruned_loss=0.03694, over 3301933.47 frames. ], batch size: 146, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:55:01,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8600, 2.0267, 2.3832, 2.7383, 2.7762, 2.7789, 2.1164, 2.9863], device='cuda:2'), covar=tensor([0.0198, 0.0505, 0.0415, 0.0313, 0.0325, 0.0288, 0.0554, 0.0176], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0187, 0.0203, 0.0161, 0.0199, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 22:55:06,291 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:55:26,308 INFO [train.py:904] (2/8) Epoch 25, batch 1300, loss[loss=0.1653, simple_loss=0.2619, pruned_loss=0.03437, over 17123.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2451, pruned_loss=0.03673, over 3312040.67 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:55:54,531 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 22:56:12,137 INFO [optim.py:368] (2/8) 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 22:56:28,218 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244948.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:56:35,068 INFO [train.py:904] (2/8) Epoch 25, batch 1350, loss[loss=0.1803, simple_loss=0.2659, pruned_loss=0.04735, over 16700.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2453, pruned_loss=0.03652, over 3305625.84 frames. ], batch size: 57, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:56:36,660 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:56:39,368 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 22:57:42,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9981, 4.4354, 4.3961, 3.1688, 3.6745, 4.3690, 3.9043, 2.5948], device='cuda:2'), covar=tensor([0.0494, 0.0075, 0.0047, 0.0385, 0.0150, 0.0112, 0.0104, 0.0473], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0136, 0.0100, 0.0111, 0.0097, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 22:57:43,144 INFO [train.py:904] (2/8) Epoch 25, batch 1400, loss[loss=0.1598, simple_loss=0.2556, pruned_loss=0.03198, over 17126.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2453, pruned_loss=0.03647, over 3306718.68 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:57:51,822 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:57:59,986 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245015.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:58:28,647 INFO [optim.py:368] (2/8) 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 22:58:51,283 INFO [train.py:904] (2/8) Epoch 25, batch 1450, loss[loss=0.1475, simple_loss=0.2403, pruned_loss=0.02741, over 17177.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2448, pruned_loss=0.03632, over 3319704.09 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:58:56,914 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:59:00,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9429, 4.0362, 2.7222, 4.6752, 3.2451, 4.5807, 2.7969, 3.4706], device='cuda:2'), covar=tensor([0.0296, 0.0429, 0.1547, 0.0262, 0.0823, 0.0541, 0.1440, 0.0680], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0180, 0.0196, 0.0171, 0.0180, 0.0221, 0.0206, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 22:59:18,826 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245073.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:00:00,089 INFO [train.py:904] (2/8) Epoch 25, batch 1500, loss[loss=0.1695, simple_loss=0.2556, pruned_loss=0.04168, over 17211.00 frames. ], tot_loss[loss=0.1589, simple_loss=0.2446, pruned_loss=0.03657, over 3309164.08 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:00:28,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0098, 4.0175, 3.9584, 3.2814, 3.9642, 1.8190, 3.7387, 3.3697], device='cuda:2'), covar=tensor([0.0165, 0.0138, 0.0216, 0.0265, 0.0112, 0.2865, 0.0157, 0.0275], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0169, 0.0209, 0.0183, 0.0186, 0.0215, 0.0198, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:00:46,575 INFO [optim.py:368] (2/8) 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 23:00:52,146 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5908, 2.5226, 1.8151, 2.6049, 2.0184, 2.7229, 2.0654, 2.2628], device='cuda:2'), covar=tensor([0.0306, 0.0352, 0.1314, 0.0275, 0.0592, 0.0435, 0.1260, 0.0578], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0172, 0.0181, 0.0223, 0.0207, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:01:08,932 INFO [train.py:904] (2/8) Epoch 25, batch 1550, loss[loss=0.2021, simple_loss=0.2691, pruned_loss=0.06757, over 16892.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2452, pruned_loss=0.03775, over 3304075.66 frames. ], batch size: 116, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:01:12,406 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8885, 2.9667, 2.8087, 5.1282, 4.1932, 4.4855, 1.7380, 3.3516], device='cuda:2'), covar=tensor([0.1325, 0.0784, 0.1193, 0.0182, 0.0228, 0.0404, 0.1604, 0.0721], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0196, 0.0204, 0.0218, 0.0206, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:02:19,884 INFO [train.py:904] (2/8) Epoch 25, batch 1600, loss[loss=0.1839, simple_loss=0.2659, pruned_loss=0.0509, over 16179.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.247, pruned_loss=0.03842, over 3297718.73 frames. ], batch size: 164, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:02:25,919 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8198, 2.8925, 2.6992, 4.7859, 3.7543, 4.2289, 1.7406, 3.2065], device='cuda:2'), covar=tensor([0.1435, 0.0791, 0.1256, 0.0212, 0.0222, 0.0436, 0.1649, 0.0809], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0196, 0.0203, 0.0218, 0.0206, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:02:33,831 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0864, 3.8261, 4.3872, 2.2634, 4.4741, 4.5806, 3.4116, 3.6673], device='cuda:2'), covar=tensor([0.0730, 0.0313, 0.0232, 0.1175, 0.0102, 0.0199, 0.0423, 0.0364], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0140, 0.0084, 0.0131, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:02:48,801 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8855, 2.6965, 2.6518, 4.3268, 3.4981, 4.1984, 1.6626, 3.0218], device='cuda:2'), covar=tensor([0.1380, 0.0728, 0.1201, 0.0170, 0.0189, 0.0366, 0.1577, 0.0826], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0196, 0.0204, 0.0218, 0.0206, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:03:06,948 INFO [optim.py:368] (2/8) 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 23:03:16,799 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:03:29,743 INFO [train.py:904] (2/8) Epoch 25, batch 1650, loss[loss=0.1737, simple_loss=0.2546, pruned_loss=0.04638, over 16558.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2494, pruned_loss=0.03957, over 3290581.41 frames. ], batch size: 75, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:03:50,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5391, 3.5506, 2.2212, 3.7595, 2.7271, 3.7366, 2.3841, 2.9044], device='cuda:2'), covar=tensor([0.0256, 0.0394, 0.1542, 0.0355, 0.0803, 0.0729, 0.1314, 0.0697], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0173, 0.0181, 0.0222, 0.0206, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:04:41,926 INFO [train.py:904] (2/8) Epoch 25, batch 1700, loss[loss=0.1654, simple_loss=0.2575, pruned_loss=0.03667, over 17106.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2513, pruned_loss=0.04017, over 3296306.69 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:04:51,855 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:04:52,092 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0423, 3.0667, 3.3004, 2.2018, 2.8989, 2.3106, 3.5689, 3.4571], device='cuda:2'), covar=tensor([0.0224, 0.0970, 0.0630, 0.1864, 0.0885, 0.1003, 0.0493, 0.0848], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0155, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:05:30,603 INFO [optim.py:368] (2/8) 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 23:05:52,657 INFO [train.py:904] (2/8) Epoch 25, batch 1750, loss[loss=0.1697, simple_loss=0.2497, pruned_loss=0.04483, over 16861.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2527, pruned_loss=0.04063, over 3298467.50 frames. ], batch size: 96, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:06:20,890 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245373.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:07:03,956 INFO [train.py:904] (2/8) Epoch 25, batch 1800, loss[loss=0.1862, simple_loss=0.2755, pruned_loss=0.04845, over 16636.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2531, pruned_loss=0.04022, over 3298316.19 frames. ], batch size: 57, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:07:29,362 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245421.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:07:51,077 INFO [optim.py:368] (2/8) 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 23:08:12,915 INFO [train.py:904] (2/8) Epoch 25, batch 1850, loss[loss=0.136, simple_loss=0.2291, pruned_loss=0.02147, over 17221.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2538, pruned_loss=0.04008, over 3296644.21 frames. ], batch size: 43, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:08:22,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4859, 4.5019, 4.8589, 4.8393, 4.9046, 4.5636, 4.5832, 4.4635], device='cuda:2'), covar=tensor([0.0438, 0.0664, 0.0461, 0.0477, 0.0539, 0.0489, 0.0895, 0.0633], device='cuda:2'), in_proj_covar=tensor([0.0433, 0.0482, 0.0469, 0.0432, 0.0516, 0.0495, 0.0573, 0.0392], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 23:08:23,490 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:09:23,471 INFO [train.py:904] (2/8) Epoch 25, batch 1900, loss[loss=0.1921, simple_loss=0.2832, pruned_loss=0.05051, over 12027.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2538, pruned_loss=0.03969, over 3292913.28 frames. ], batch size: 248, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:09:49,178 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245521.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:11,812 INFO [optim.py:368] (2/8) 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 23:10:20,947 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245543.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:25,025 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:33,940 INFO [train.py:904] (2/8) Epoch 25, batch 1950, loss[loss=0.2019, simple_loss=0.3027, pruned_loss=0.05059, over 17035.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2549, pruned_loss=0.03988, over 3289729.22 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:10:35,521 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9765, 4.0794, 4.3390, 4.3154, 4.3464, 4.0904, 4.1166, 4.0896], device='cuda:2'), covar=tensor([0.0413, 0.0595, 0.0394, 0.0380, 0.0532, 0.0439, 0.0735, 0.0558], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0481, 0.0469, 0.0430, 0.0515, 0.0494, 0.0570, 0.0391], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 23:10:40,212 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4090, 4.3867, 4.3148, 3.7137, 4.3950, 1.6230, 4.1083, 3.8188], device='cuda:2'), covar=tensor([0.0149, 0.0139, 0.0212, 0.0333, 0.0104, 0.3260, 0.0172, 0.0282], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0170, 0.0209, 0.0183, 0.0186, 0.0216, 0.0199, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:10:58,262 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7713, 4.1911, 3.0309, 2.3636, 2.7518, 2.5960, 4.4442, 3.5493], device='cuda:2'), covar=tensor([0.2873, 0.0573, 0.1882, 0.2862, 0.2592, 0.2078, 0.0400, 0.1383], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0273, 0.0311, 0.0321, 0.0302, 0.0271, 0.0302, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 23:11:07,219 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 23:11:17,601 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9812, 3.2098, 2.9456, 5.1995, 4.2385, 4.5100, 1.9711, 3.3547], device='cuda:2'), covar=tensor([0.1327, 0.0731, 0.1151, 0.0190, 0.0229, 0.0400, 0.1536, 0.0755], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0196, 0.0203, 0.0217, 0.0205, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:11:26,898 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:11:41,386 INFO [train.py:904] (2/8) Epoch 25, batch 2000, loss[loss=0.1883, simple_loss=0.2654, pruned_loss=0.05554, over 16353.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2547, pruned_loss=0.0397, over 3299985.03 frames. ], batch size: 145, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:49,104 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245607.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:11:53,240 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245610.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:12:28,721 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 23:12:31,548 INFO [optim.py:368] (2/8) 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 23:12:35,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9122, 5.2441, 5.0024, 4.9804, 4.7522, 4.7467, 4.6723, 5.3457], device='cuda:2'), covar=tensor([0.1426, 0.1039, 0.1165, 0.1097, 0.0963, 0.1063, 0.1347, 0.1010], device='cuda:2'), in_proj_covar=tensor([0.0718, 0.0871, 0.0713, 0.0674, 0.0553, 0.0554, 0.0733, 0.0681], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:12:38,924 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6950, 2.6618, 2.3582, 2.6162, 3.0295, 2.8034, 3.2977, 3.2441], device='cuda:2'), covar=tensor([0.0184, 0.0495, 0.0575, 0.0501, 0.0318, 0.0438, 0.0293, 0.0328], device='cuda:2'), in_proj_covar=tensor([0.0228, 0.0247, 0.0235, 0.0236, 0.0248, 0.0246, 0.0247, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:12:50,393 INFO [train.py:904] (2/8) Epoch 25, batch 2050, loss[loss=0.1575, simple_loss=0.244, pruned_loss=0.03552, over 17227.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2542, pruned_loss=0.03973, over 3304143.59 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:12:57,318 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245658.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:12:59,876 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0289, 2.2107, 2.6180, 3.0416, 2.8174, 3.5313, 2.4375, 3.5144], device='cuda:2'), covar=tensor([0.0290, 0.0539, 0.0378, 0.0337, 0.0398, 0.0194, 0.0559, 0.0195], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0197, 0.0184, 0.0188, 0.0205, 0.0162, 0.0200, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:13:34,818 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245686.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:13:45,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6899, 3.6999, 2.3148, 3.9355, 2.9801, 3.8425, 2.3652, 2.9670], device='cuda:2'), covar=tensor([0.0254, 0.0402, 0.1452, 0.0359, 0.0707, 0.0857, 0.1384, 0.0704], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0182, 0.0197, 0.0173, 0.0181, 0.0224, 0.0207, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:13:58,101 INFO [train.py:904] (2/8) Epoch 25, batch 2100, loss[loss=0.1781, simple_loss=0.2641, pruned_loss=0.04606, over 16512.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2546, pruned_loss=0.03994, over 3305463.49 frames. ], batch size: 68, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:14:31,305 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-01 23:14:49,017 INFO [optim.py:368] (2/8) 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 23:14:59,031 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245747.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:15:08,429 INFO [train.py:904] (2/8) Epoch 25, batch 2150, loss[loss=0.1414, simple_loss=0.2325, pruned_loss=0.02516, over 17236.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2551, pruned_loss=0.03992, over 3316368.89 frames. ], batch size: 45, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:00,570 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 25, batch 2200, loss[loss=0.1767, simple_loss=0.2485, pruned_loss=0.05245, over 16647.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.256, pruned_loss=0.04071, over 3308639.04 frames. ], batch size: 89, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:34,702 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245816.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:17:06,932 INFO [optim.py:368] (2/8) 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 23:17:09,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2331, 5.2248, 5.0854, 4.6016, 4.7005, 5.1280, 5.0181, 4.6995], device='cuda:2'), covar=tensor([0.0647, 0.0541, 0.0318, 0.0352, 0.1181, 0.0524, 0.0374, 0.0845], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0469, 0.0364, 0.0368, 0.0369, 0.0423, 0.0249, 0.0441], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 23:17:23,656 INFO [train.py:904] (2/8) Epoch 25, batch 2250, loss[loss=0.1586, simple_loss=0.2485, pruned_loss=0.03437, over 17221.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2568, pruned_loss=0.0413, over 3311648.61 frames. ], batch size: 45, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:17:24,138 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:17:35,287 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0916, 5.0720, 4.9381, 4.4949, 4.5943, 5.0039, 4.8762, 4.6099], device='cuda:2'), covar=tensor([0.0630, 0.0575, 0.0335, 0.0360, 0.1085, 0.0461, 0.0418, 0.0835], device='cuda:2'), in_proj_covar=tensor([0.0315, 0.0470, 0.0365, 0.0369, 0.0369, 0.0424, 0.0250, 0.0443], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-01 23:17:37,005 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5115, 3.5067, 3.4357, 2.7125, 3.3858, 2.0392, 3.1775, 2.8392], device='cuda:2'), covar=tensor([0.0189, 0.0155, 0.0250, 0.0329, 0.0140, 0.2523, 0.0175, 0.0339], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0170, 0.0210, 0.0184, 0.0187, 0.0216, 0.0200, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:18:14,302 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9694, 2.1401, 2.6063, 2.8566, 2.7969, 3.4035, 2.4314, 3.3896], device='cuda:2'), covar=tensor([0.0287, 0.0529, 0.0362, 0.0365, 0.0363, 0.0194, 0.0532, 0.0179], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0198, 0.0185, 0.0190, 0.0205, 0.0162, 0.0201, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:18:16,569 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245891.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:32,159 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245902.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:18:32,939 INFO [train.py:904] (2/8) Epoch 25, batch 2300, loss[loss=0.1777, simple_loss=0.2658, pruned_loss=0.04476, over 16538.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2572, pruned_loss=0.04155, over 3311156.41 frames. ], batch size: 68, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:18:49,943 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3072, 3.4003, 3.6960, 2.5124, 3.4349, 3.8161, 3.4883, 2.0964], device='cuda:2'), covar=tensor([0.0609, 0.0204, 0.0072, 0.0449, 0.0122, 0.0112, 0.0118, 0.0558], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0088, 0.0089, 0.0136, 0.0101, 0.0112, 0.0098, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-01 23:19:24,138 INFO [optim.py:368] (2/8) 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 23:19:42,213 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245952.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:19:42,915 INFO [train.py:904] (2/8) Epoch 25, batch 2350, loss[loss=0.146, simple_loss=0.2372, pruned_loss=0.02738, over 17230.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2567, pruned_loss=0.04141, over 3318365.14 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:19:45,521 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6979, 4.7252, 5.0653, 5.0416, 5.1095, 4.7662, 4.7371, 4.6388], device='cuda:2'), covar=tensor([0.0382, 0.0723, 0.0483, 0.0491, 0.0630, 0.0484, 0.1153, 0.0562], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0483, 0.0472, 0.0433, 0.0517, 0.0497, 0.0575, 0.0392], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 23:19:45,667 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9237, 2.9039, 2.6249, 4.5934, 3.5905, 4.2018, 1.8044, 3.1303], device='cuda:2'), covar=tensor([0.1327, 0.0772, 0.1208, 0.0225, 0.0219, 0.0433, 0.1549, 0.0818], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0178, 0.0197, 0.0197, 0.0204, 0.0217, 0.0206, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:20:33,936 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1855, 3.9774, 4.4338, 2.4133, 4.6284, 4.7404, 3.4692, 3.6665], device='cuda:2'), covar=tensor([0.0726, 0.0294, 0.0254, 0.1170, 0.0082, 0.0166, 0.0429, 0.0417], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0111, 0.0101, 0.0141, 0.0084, 0.0131, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:20:54,771 INFO [train.py:904] (2/8) Epoch 25, batch 2400, loss[loss=0.1461, simple_loss=0.2411, pruned_loss=0.02556, over 16998.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2569, pruned_loss=0.04144, over 3299063.40 frames. ], batch size: 41, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:21:17,951 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 23:21:46,453 INFO [optim.py:368] (2/8) 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 23:21:49,228 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246042.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:22:04,193 INFO [train.py:904] (2/8) Epoch 25, batch 2450, loss[loss=0.1721, simple_loss=0.254, pruned_loss=0.04515, over 16470.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2569, pruned_loss=0.04077, over 3314329.27 frames. ], batch size: 75, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:10,958 INFO [train.py:904] (2/8) Epoch 25, batch 2500, loss[loss=0.1719, simple_loss=0.2587, pruned_loss=0.04254, over 15823.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2564, pruned_loss=0.04068, over 3310673.12 frames. ], batch size: 35, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:21,569 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246111.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:27,375 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 23:23:28,160 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:59,069 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2549, 2.1566, 2.7938, 3.2313, 3.0221, 3.6537, 2.2984, 3.6880], device='cuda:2'), covar=tensor([0.0218, 0.0649, 0.0337, 0.0312, 0.0330, 0.0185, 0.0668, 0.0186], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0198, 0.0185, 0.0190, 0.0205, 0.0163, 0.0202, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:24:01,581 INFO [optim.py:368] (2/8) 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 23:24:12,431 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:18,464 INFO [train.py:904] (2/8) Epoch 25, batch 2550, loss[loss=0.1773, simple_loss=0.2696, pruned_loss=0.04253, over 17076.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2565, pruned_loss=0.04051, over 3317799.24 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:24:24,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1555, 3.7888, 4.3636, 2.1986, 4.4954, 4.6388, 3.3142, 3.5885], device='cuda:2'), covar=tensor([0.0671, 0.0311, 0.0227, 0.1183, 0.0092, 0.0179, 0.0468, 0.0396], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0112, 0.0101, 0.0141, 0.0084, 0.0132, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:24:29,271 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1438, 2.4051, 2.7547, 3.0823, 2.9322, 3.5830, 2.6451, 3.5618], device='cuda:2'), covar=tensor([0.0255, 0.0500, 0.0350, 0.0347, 0.0358, 0.0208, 0.0459, 0.0184], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0197, 0.0185, 0.0189, 0.0204, 0.0163, 0.0201, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:24:35,187 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:44,830 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:25:06,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9205, 2.8249, 2.6642, 4.2503, 3.5224, 4.1408, 1.7324, 2.9878], device='cuda:2'), covar=tensor([0.1311, 0.0660, 0.1098, 0.0166, 0.0103, 0.0369, 0.1535, 0.0835], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0196, 0.0204, 0.0217, 0.0205, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:25:17,133 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246196.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:25:26,251 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246202.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:25:26,989 INFO [train.py:904] (2/8) Epoch 25, batch 2600, loss[loss=0.1659, simple_loss=0.2668, pruned_loss=0.03246, over 17088.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2571, pruned_loss=0.04049, over 3307421.53 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:25:52,244 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:18,782 INFO [optim.py:368] (2/8) 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 23:26:28,308 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:37,198 INFO [train.py:904] (2/8) Epoch 25, batch 2650, loss[loss=0.1559, simple_loss=0.2544, pruned_loss=0.02867, over 17115.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2574, pruned_loss=0.03982, over 3307387.94 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:26:42,344 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246257.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:27:17,016 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:27:44,599 INFO [train.py:904] (2/8) Epoch 25, batch 2700, loss[loss=0.1848, simple_loss=0.2705, pruned_loss=0.04952, over 15510.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2581, pruned_loss=0.03979, over 3317634.35 frames. ], batch size: 190, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:28:34,091 INFO [optim.py:368] (2/8) 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 23:28:36,970 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246342.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:28:52,088 INFO [train.py:904] (2/8) Epoch 25, batch 2750, loss[loss=0.1687, simple_loss=0.2514, pruned_loss=0.04296, over 15727.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2583, pruned_loss=0.03956, over 3323417.80 frames. ], batch size: 190, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:29:28,481 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 23:29:43,588 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246390.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:29:45,538 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3733, 3.9885, 4.4967, 2.4310, 4.7028, 4.7500, 3.3499, 3.7160], device='cuda:2'), covar=tensor([0.0663, 0.0282, 0.0229, 0.1114, 0.0093, 0.0240, 0.0485, 0.0398], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0112, 0.0102, 0.0142, 0.0085, 0.0133, 0.0132, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:30:01,769 INFO [train.py:904] (2/8) Epoch 25, batch 2800, loss[loss=0.1698, simple_loss=0.2578, pruned_loss=0.0409, over 17028.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2581, pruned_loss=0.03933, over 3324346.70 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:30:27,746 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0250, 5.0785, 5.4957, 5.4718, 5.4979, 5.1636, 5.0869, 4.9300], device='cuda:2'), covar=tensor([0.0358, 0.0535, 0.0397, 0.0431, 0.0483, 0.0424, 0.0948, 0.0454], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0484, 0.0472, 0.0433, 0.0517, 0.0497, 0.0575, 0.0392], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 23:30:34,332 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8304, 2.6793, 2.4728, 4.0022, 3.1489, 4.0412, 1.5393, 2.8837], device='cuda:2'), covar=tensor([0.1411, 0.0754, 0.1245, 0.0181, 0.0167, 0.0376, 0.1711, 0.0841], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0198, 0.0198, 0.0206, 0.0219, 0.0206, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-01 23:30:54,340 INFO [optim.py:368] (2/8) 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 23:31:04,175 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:31:10,739 INFO [train.py:904] (2/8) Epoch 25, batch 2850, loss[loss=0.1595, simple_loss=0.2386, pruned_loss=0.04022, over 16812.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2573, pruned_loss=0.03936, over 3325098.39 frames. ], batch size: 96, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:31:31,873 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:10,939 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:20,906 INFO [train.py:904] (2/8) Epoch 25, batch 2900, loss[loss=0.1378, simple_loss=0.2186, pruned_loss=0.02851, over 16959.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2566, pruned_loss=0.03962, over 3321013.48 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:32:31,058 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7779, 4.7631, 5.1744, 5.1573, 5.2063, 4.8633, 4.8137, 4.7071], device='cuda:2'), covar=tensor([0.0383, 0.0664, 0.0430, 0.0419, 0.0619, 0.0489, 0.1139, 0.0565], device='cuda:2'), in_proj_covar=tensor([0.0435, 0.0487, 0.0474, 0.0435, 0.0518, 0.0499, 0.0579, 0.0394], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-01 23:33:14,957 INFO [optim.py:368] (2/8) 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 23:33:25,272 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246547.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:33:25,751 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 23:33:32,608 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246552.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:33:33,511 INFO [train.py:904] (2/8) Epoch 25, batch 2950, loss[loss=0.1759, simple_loss=0.2529, pruned_loss=0.0495, over 16693.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2557, pruned_loss=0.0402, over 3321751.58 frames. ], batch size: 89, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:06,234 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:32,276 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246595.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:42,715 INFO [train.py:904] (2/8) Epoch 25, batch 3000, loss[loss=0.1706, simple_loss=0.2566, pruned_loss=0.0423, over 15953.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2557, pruned_loss=0.04049, over 3324330.40 frames. ], batch size: 35, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:42,716 INFO [train.py:929] (2/8) Computing validation loss 2023-05-01 23:34:52,575 INFO [train.py:938] (2/8) Epoch 25, validation: loss=0.1341, simple_loss=0.239, pruned_loss=0.01457, over 944034.00 frames. 2023-05-01 23:34:52,575 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-01 23:35:46,635 INFO [optim.py:368] (2/8) 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 23:36:05,620 INFO [train.py:904] (2/8) Epoch 25, batch 3050, loss[loss=0.1763, simple_loss=0.262, pruned_loss=0.04525, over 16509.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2552, pruned_loss=0.04052, over 3329760.30 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:36:38,404 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:37:13,553 INFO [train.py:904] (2/8) Epoch 25, batch 3100, loss[loss=0.1512, simple_loss=0.2509, pruned_loss=0.02575, over 17268.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2549, pruned_loss=0.04034, over 3337747.70 frames. ], batch size: 52, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:37:41,145 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 23:38:04,944 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0936, 5.0235, 4.8626, 3.9185, 5.0111, 1.6269, 4.6472, 4.6564], device='cuda:2'), covar=tensor([0.0125, 0.0125, 0.0273, 0.0625, 0.0121, 0.3436, 0.0183, 0.0293], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0172, 0.0211, 0.0186, 0.0189, 0.0217, 0.0201, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:38:05,015 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246738.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:38:08,780 INFO [optim.py:368] (2/8) 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 23:38:25,015 INFO [train.py:904] (2/8) Epoch 25, batch 3150, loss[loss=0.1719, simple_loss=0.2529, pruned_loss=0.04543, over 16166.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.254, pruned_loss=0.0399, over 3328554.44 frames. ], batch size: 164, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:38:44,301 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 25, batch 3200, loss[loss=0.1457, simple_loss=0.2377, pruned_loss=0.02685, over 17191.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2541, pruned_loss=0.03994, over 3318549.85 frames. ], batch size: 46, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:39:51,041 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246815.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:40:27,531 INFO [optim.py:368] (2/8) 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 23:40:42,129 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246852.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:40:42,933 INFO [train.py:904] (2/8) Epoch 25, batch 3250, loss[loss=0.1745, simple_loss=0.2663, pruned_loss=0.04137, over 16645.00 frames. ], tot_loss[loss=0.167, simple_loss=0.254, pruned_loss=0.03998, over 3317875.16 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:41:16,218 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:41:49,125 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 25, batch 3300, loss[loss=0.1302, simple_loss=0.2192, pruned_loss=0.0206, over 16962.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.255, pruned_loss=0.04012, over 3317333.22 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:42:24,635 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:24,747 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4439, 4.2955, 4.5262, 4.6563, 4.7992, 4.3475, 4.6762, 4.7896], device='cuda:2'), covar=tensor([0.1753, 0.1260, 0.1445, 0.0720, 0.0585, 0.1125, 0.1796, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0692, 0.0854, 0.0986, 0.0863, 0.0658, 0.0681, 0.0707, 0.0828], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:42:34,535 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9960, 2.7605, 2.7542, 2.0882, 2.6395, 2.1700, 2.7974, 2.9641], device='cuda:2'), covar=tensor([0.0306, 0.0794, 0.0634, 0.1745, 0.0886, 0.0874, 0.0592, 0.0754], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0169, 0.0171, 0.0156, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-01 23:42:36,873 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:45,876 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3763, 3.3476, 3.4415, 3.5076, 3.5568, 3.2980, 3.5205, 3.6181], device='cuda:2'), covar=tensor([0.1317, 0.0933, 0.1072, 0.0680, 0.0631, 0.2578, 0.1188, 0.0768], device='cuda:2'), in_proj_covar=tensor([0.0692, 0.0855, 0.0987, 0.0864, 0.0659, 0.0682, 0.0708, 0.0829], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:42:46,635 INFO [optim.py:368] (2/8) 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 23:43:00,933 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8708, 2.0744, 2.4968, 2.7944, 2.6764, 3.4049, 2.2945, 3.3555], device='cuda:2'), covar=tensor([0.0309, 0.0535, 0.0387, 0.0391, 0.0401, 0.0195, 0.0558, 0.0184], device='cuda:2'), in_proj_covar=tensor([0.0198, 0.0197, 0.0185, 0.0190, 0.0206, 0.0164, 0.0202, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:43:02,688 INFO [train.py:904] (2/8) Epoch 25, batch 3350, loss[loss=0.1562, simple_loss=0.2576, pruned_loss=0.02742, over 17142.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2557, pruned_loss=0.04043, over 3302621.97 frames. ], batch size: 48, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:43:43,203 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8695, 4.2723, 2.9917, 2.3502, 2.6095, 2.6307, 4.6214, 3.4882], device='cuda:2'), covar=tensor([0.2800, 0.0551, 0.1897, 0.2930, 0.2953, 0.2073, 0.0337, 0.1460], device='cuda:2'), in_proj_covar=tensor([0.0333, 0.0275, 0.0312, 0.0323, 0.0305, 0.0272, 0.0303, 0.0351], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-01 23:44:01,695 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246995.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:44:13,548 INFO [train.py:904] (2/8) Epoch 25, batch 3400, loss[loss=0.2078, simple_loss=0.278, pruned_loss=0.06876, over 16788.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2562, pruned_loss=0.04072, over 3298541.23 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:56,893 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247033.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:45:08,920 INFO [optim.py:368] (2/8) 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 23:45:26,146 INFO [train.py:904] (2/8) Epoch 25, batch 3450, loss[loss=0.164, simple_loss=0.2397, pruned_loss=0.04417, over 16838.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.254, pruned_loss=0.0396, over 3306660.56 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:45:54,750 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 23:46:35,594 INFO [train.py:904] (2/8) Epoch 25, batch 3500, loss[loss=0.191, simple_loss=0.2733, pruned_loss=0.05437, over 11568.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2537, pruned_loss=0.03966, over 3309527.85 frames. ], batch size: 248, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:47:30,016 INFO [optim.py:368] (2/8) 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 23:47:45,773 INFO [train.py:904] (2/8) Epoch 25, batch 3550, loss[loss=0.1671, simple_loss=0.2682, pruned_loss=0.03305, over 17117.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.253, pruned_loss=0.03935, over 3300910.17 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:48:52,336 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 23:48:55,167 INFO [train.py:904] (2/8) Epoch 25, batch 3600, loss[loss=0.1753, simple_loss=0.2495, pruned_loss=0.05053, over 16890.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2519, pruned_loss=0.03922, over 3297257.26 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:49:07,545 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 23:49:12,304 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-01 23:49:49,107 INFO [optim.py:368] (2/8) 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 23:49:54,628 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2402, 3.9645, 3.9878, 2.2231, 3.3632, 3.0070, 3.7016, 4.0352], device='cuda:2'), covar=tensor([0.0420, 0.0857, 0.0556, 0.2033, 0.0826, 0.0853, 0.0873, 0.1010], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0156, 0.0148, 0.0132, 0.0147, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-01 23:50:05,473 INFO [train.py:904] (2/8) Epoch 25, batch 3650, loss[loss=0.1476, simple_loss=0.2279, pruned_loss=0.03364, over 16872.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2504, pruned_loss=0.03978, over 3288132.98 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:50:59,437 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:51:17,359 INFO [train.py:904] (2/8) Epoch 25, batch 3700, loss[loss=0.1693, simple_loss=0.242, pruned_loss=0.04831, over 16774.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2493, pruned_loss=0.04103, over 3263469.11 frames. ], batch size: 89, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:52:01,349 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:52:12,944 INFO [optim.py:368] (2/8) 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 23:52:30,283 INFO [train.py:904] (2/8) Epoch 25, batch 3750, loss[loss=0.177, simple_loss=0.2639, pruned_loss=0.04503, over 16519.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2501, pruned_loss=0.04213, over 3255751.83 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:53:06,842 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:10,240 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:42,912 INFO [train.py:904] (2/8) Epoch 25, batch 3800, loss[loss=0.1718, simple_loss=0.2525, pruned_loss=0.04556, over 16688.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2504, pruned_loss=0.04311, over 3254434.47 frames. ], batch size: 57, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:54:06,914 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0924, 2.1933, 2.3162, 3.8772, 2.1816, 2.4964, 2.2728, 2.3888], device='cuda:2'), covar=tensor([0.1608, 0.3882, 0.3095, 0.0634, 0.4041, 0.2620, 0.4059, 0.3076], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0465, 0.0383, 0.0337, 0.0444, 0.0533, 0.0438, 0.0545], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:54:15,331 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247426.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:35,818 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247440.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:38,773 INFO [optim.py:368] (2/8) 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 23:54:54,708 INFO [train.py:904] (2/8) Epoch 25, batch 3850, loss[loss=0.1713, simple_loss=0.2503, pruned_loss=0.04616, over 16443.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2511, pruned_loss=0.04407, over 3262988.43 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:55:18,752 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7902, 5.1371, 4.6963, 4.9838, 4.7224, 4.5985, 4.6907, 5.2091], device='cuda:2'), covar=tensor([0.2213, 0.1427, 0.2223, 0.1409, 0.1563, 0.1802, 0.2094, 0.1828], device='cuda:2'), in_proj_covar=tensor([0.0725, 0.0874, 0.0720, 0.0679, 0.0556, 0.0559, 0.0739, 0.0686], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:55:22,893 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6084, 4.4772, 4.6376, 4.7864, 4.8895, 4.4118, 4.7744, 4.8729], device='cuda:2'), covar=tensor([0.1700, 0.1210, 0.1406, 0.0761, 0.0672, 0.1134, 0.1773, 0.1055], device='cuda:2'), in_proj_covar=tensor([0.0690, 0.0849, 0.0979, 0.0857, 0.0657, 0.0677, 0.0704, 0.0826], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-01 23:55:43,396 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:05,819 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-01 23:56:06,077 INFO [train.py:904] (2/8) Epoch 25, batch 3900, loss[loss=0.1637, simple_loss=0.2396, pruned_loss=0.04392, over 16781.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2504, pruned_loss=0.04449, over 3273801.23 frames. ], batch size: 102, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:56:39,753 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:03,108 INFO [optim.py:368] (2/8) 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 23:57:17,841 INFO [train.py:904] (2/8) Epoch 25, batch 3950, loss[loss=0.1673, simple_loss=0.2492, pruned_loss=0.04269, over 16523.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2498, pruned_loss=0.04488, over 3279453.71 frames. ], batch size: 75, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:57:19,995 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:25,109 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-05-01 23:57:30,976 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:05,289 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:12,115 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247590.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:30,186 INFO [train.py:904] (2/8) Epoch 25, batch 4000, loss[loss=0.1899, simple_loss=0.257, pruned_loss=0.06138, over 16756.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2501, pruned_loss=0.04509, over 3258160.95 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:58:49,584 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:21,868 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:27,472 INFO [optim.py:368] (2/8) 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 23:59:44,178 INFO [train.py:904] (2/8) Epoch 25, batch 4050, loss[loss=0.1668, simple_loss=0.2537, pruned_loss=0.03992, over 16500.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2508, pruned_loss=0.04454, over 3260857.11 frames. ], batch size: 75, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:59:59,663 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 00:00:52,182 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 00:00:59,191 INFO [train.py:904] (2/8) Epoch 25, batch 4100, loss[loss=0.2055, simple_loss=0.2922, pruned_loss=0.05946, over 17036.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2526, pruned_loss=0.04392, over 3250272.62 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:01:00,706 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-05-02 00:01:25,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8697, 5.1837, 4.7266, 5.0222, 4.7001, 4.5455, 4.6568, 5.2856], device='cuda:2'), covar=tensor([0.2128, 0.1374, 0.2138, 0.1567, 0.1504, 0.1689, 0.2338, 0.1470], device='cuda:2'), in_proj_covar=tensor([0.0723, 0.0872, 0.0718, 0.0676, 0.0554, 0.0557, 0.0738, 0.0683], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:01:43,449 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3286, 3.4755, 3.5983, 3.5728, 3.5858, 3.4348, 3.4752, 3.4694], device='cuda:2'), covar=tensor([0.0397, 0.0622, 0.0450, 0.0459, 0.0511, 0.0489, 0.0708, 0.0531], device='cuda:2'), in_proj_covar=tensor([0.0426, 0.0479, 0.0465, 0.0427, 0.0510, 0.0490, 0.0565, 0.0388], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 00:01:48,423 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247735.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:01:57,927 INFO [optim.py:368] (2/8) 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 00:02:14,951 INFO [train.py:904] (2/8) Epoch 25, batch 4150, loss[loss=0.174, simple_loss=0.2714, pruned_loss=0.03831, over 16896.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2594, pruned_loss=0.04606, over 3231254.51 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:02:41,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6950, 2.8307, 2.8170, 5.0008, 3.9115, 4.1879, 1.6195, 3.0582], device='cuda:2'), covar=tensor([0.1428, 0.0840, 0.1205, 0.0129, 0.0349, 0.0413, 0.1758, 0.0894], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0179, 0.0197, 0.0197, 0.0206, 0.0218, 0.0207, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:03:01,850 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:04,916 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247784.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:32,729 INFO [train.py:904] (2/8) Epoch 25, batch 4200, loss[loss=0.214, simple_loss=0.3016, pruned_loss=0.06315, over 16951.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2665, pruned_loss=0.04757, over 3226633.84 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:30,293 INFO [optim.py:368] (2/8) 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 00:04:35,252 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:04:46,252 INFO [train.py:904] (2/8) Epoch 25, batch 4250, loss[loss=0.1752, simple_loss=0.2654, pruned_loss=0.04252, over 16404.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2703, pruned_loss=0.04817, over 3184796.53 frames. ], batch size: 165, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:51,426 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247856.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:05:28,693 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:05:29,899 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8087, 3.8744, 3.9522, 3.7379, 3.8179, 4.2836, 3.8880, 3.5938], device='cuda:2'), covar=tensor([0.2211, 0.2063, 0.1992, 0.2444, 0.2746, 0.1647, 0.1555, 0.2691], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0625, 0.0680, 0.0508, 0.0675, 0.0709, 0.0531, 0.0679], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 00:05:40,857 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 00:05:41,825 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 00:06:02,531 INFO [train.py:904] (2/8) Epoch 25, batch 4300, loss[loss=0.1847, simple_loss=0.2754, pruned_loss=0.04699, over 11680.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2715, pruned_loss=0.0474, over 3182521.87 frames. ], batch size: 247, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:06:13,452 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:06:16,947 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 00:06:52,758 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7315, 1.8183, 1.6578, 1.4573, 1.9266, 1.5747, 1.6345, 1.9269], device='cuda:2'), covar=tensor([0.0187, 0.0307, 0.0423, 0.0406, 0.0226, 0.0297, 0.0172, 0.0223], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0243, 0.0230, 0.0233, 0.0244, 0.0243, 0.0244, 0.0242], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:07:01,389 INFO [optim.py:368] (2/8) 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 00:07:17,409 INFO [train.py:904] (2/8) Epoch 25, batch 4350, loss[loss=0.185, simple_loss=0.2825, pruned_loss=0.04374, over 17270.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2743, pruned_loss=0.04814, over 3188513.60 frames. ], batch size: 52, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:08:05,203 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247984.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:23,790 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:36,577 INFO [train.py:904] (2/8) Epoch 25, batch 4400, loss[loss=0.1959, simple_loss=0.2895, pruned_loss=0.05117, over 15370.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2763, pruned_loss=0.04926, over 3192724.67 frames. ], batch size: 190, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:08:59,459 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8135, 3.8397, 3.9238, 3.7321, 3.8876, 4.2704, 3.9356, 3.5033], device='cuda:2'), covar=tensor([0.1948, 0.1881, 0.2180, 0.2300, 0.2496, 0.1624, 0.1434, 0.2608], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0620, 0.0675, 0.0504, 0.0669, 0.0705, 0.0526, 0.0673], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 00:09:04,037 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8405, 4.8344, 4.6143, 3.8716, 4.7682, 1.7835, 4.5253, 4.1171], device='cuda:2'), covar=tensor([0.0058, 0.0044, 0.0153, 0.0294, 0.0052, 0.2990, 0.0084, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0169, 0.0209, 0.0186, 0.0187, 0.0215, 0.0199, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:09:16,197 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0932, 5.1308, 4.9740, 4.5928, 4.6885, 5.0346, 4.7959, 4.7471], device='cuda:2'), covar=tensor([0.0429, 0.0238, 0.0196, 0.0226, 0.0688, 0.0288, 0.0301, 0.0501], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0459, 0.0357, 0.0362, 0.0362, 0.0416, 0.0245, 0.0431], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:09:24,817 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:34,454 INFO [optim.py:368] (2/8) 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 00:09:39,348 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248045.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:50,302 INFO [train.py:904] (2/8) Epoch 25, batch 4450, loss[loss=0.2041, simple_loss=0.3, pruned_loss=0.05415, over 15494.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2803, pruned_loss=0.05079, over 3198676.17 frames. ], batch size: 190, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:56,465 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,187 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,205 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:34,922 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248083.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:04,506 INFO [train.py:904] (2/8) Epoch 25, batch 4500, loss[loss=0.1928, simple_loss=0.281, pruned_loss=0.05227, over 16878.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2811, pruned_loss=0.05155, over 3207518.25 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:11:44,502 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:59,978 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:01,968 INFO [optim.py:368] (2/8) 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 00:12:02,456 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.4876, 2.6708, 2.5049, 4.0962, 3.0017, 3.9206, 1.5255, 2.8268], device='cuda:2'), covar=tensor([0.1520, 0.0836, 0.1299, 0.0143, 0.0269, 0.0352, 0.1859, 0.0909], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0196, 0.0205, 0.0216, 0.0206, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:12:03,656 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 25, batch 4550, loss[loss=0.2109, simple_loss=0.2979, pruned_loss=0.06194, over 16680.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2818, pruned_loss=0.05238, over 3214093.44 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:12:22,924 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:00,081 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248181.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:32,556 INFO [train.py:904] (2/8) Epoch 25, batch 4600, loss[loss=0.2059, simple_loss=0.2828, pruned_loss=0.06452, over 11962.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2821, pruned_loss=0.0523, over 3219279.89 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:13:34,235 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248204.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:42,847 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:01,135 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 00:14:11,936 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248229.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:29,187 INFO [optim.py:368] (2/8) 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 00:14:46,950 INFO [train.py:904] (2/8) Epoch 25, batch 4650, loss[loss=0.1871, simple_loss=0.2647, pruned_loss=0.05471, over 16599.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2813, pruned_loss=0.05213, over 3201714.04 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:14:54,066 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:15:07,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9173, 3.0446, 3.3726, 1.9773, 2.9083, 2.0427, 3.3294, 3.2413], device='cuda:2'), covar=tensor([0.0228, 0.0863, 0.0617, 0.2279, 0.0884, 0.1133, 0.0655, 0.1030], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0168, 0.0170, 0.0155, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 00:16:01,796 INFO [train.py:904] (2/8) Epoch 25, batch 4700, loss[loss=0.1626, simple_loss=0.2564, pruned_loss=0.03443, over 16793.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2785, pruned_loss=0.05102, over 3195166.56 frames. ], batch size: 83, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:16:05,183 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0842, 5.5530, 5.7177, 5.5026, 5.5299, 6.0668, 5.5791, 5.2995], device='cuda:2'), covar=tensor([0.0875, 0.1572, 0.1988, 0.1882, 0.2252, 0.0796, 0.1264, 0.2198], device='cuda:2'), in_proj_covar=tensor([0.0424, 0.0624, 0.0680, 0.0507, 0.0674, 0.0711, 0.0529, 0.0678], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 00:16:41,939 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 00:16:44,318 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:56,028 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:57,976 INFO [optim.py:368] (2/8) 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 00:16:59,722 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0396, 4.8719, 5.0574, 5.2672, 5.4378, 4.8683, 5.4305, 5.4785], device='cuda:2'), covar=tensor([0.1810, 0.1328, 0.1862, 0.0781, 0.0584, 0.0762, 0.0633, 0.0600], device='cuda:2'), in_proj_covar=tensor([0.0662, 0.0816, 0.0938, 0.0826, 0.0632, 0.0650, 0.0676, 0.0792], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:17:13,863 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:17:14,734 INFO [train.py:904] (2/8) Epoch 25, batch 4750, loss[loss=0.1389, simple_loss=0.2356, pruned_loss=0.02104, over 16804.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2747, pruned_loss=0.049, over 3202236.74 frames. ], batch size: 102, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:17:59,879 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:13,253 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:29,678 INFO [train.py:904] (2/8) Epoch 25, batch 4800, loss[loss=0.1635, simple_loss=0.2505, pruned_loss=0.03827, over 17001.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2714, pruned_loss=0.04706, over 3192655.05 frames. ], batch size: 53, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:18:41,944 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8749, 2.0468, 2.4820, 2.8853, 2.7527, 3.3347, 2.1234, 3.2626], device='cuda:2'), covar=tensor([0.0223, 0.0532, 0.0375, 0.0342, 0.0373, 0.0190, 0.0601, 0.0169], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0188, 0.0203, 0.0161, 0.0199, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:19:22,190 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:24,756 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248440.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:27,480 INFO [optim.py:368] (2/8) 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 00:19:31,501 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:43,836 INFO [train.py:904] (2/8) Epoch 25, batch 4850, loss[loss=0.1745, simple_loss=0.2702, pruned_loss=0.03939, over 16475.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2718, pruned_loss=0.04623, over 3187252.01 frames. ], batch size: 75, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:20:24,707 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 00:20:36,774 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:20:50,318 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 00:20:58,428 INFO [train.py:904] (2/8) Epoch 25, batch 4900, loss[loss=0.1793, simple_loss=0.2684, pruned_loss=0.04509, over 16826.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2707, pruned_loss=0.04485, over 3188986.00 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:21:08,251 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8058, 1.4167, 1.7059, 1.7278, 1.8400, 1.9194, 1.6927, 1.7871], device='cuda:2'), covar=tensor([0.0253, 0.0413, 0.0232, 0.0326, 0.0289, 0.0188, 0.0426, 0.0162], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0187, 0.0202, 0.0160, 0.0199, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:22:03,445 INFO [optim.py:368] (2/8) 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 00:22:20,179 INFO [train.py:904] (2/8) Epoch 25, batch 4950, loss[loss=0.2124, simple_loss=0.3112, pruned_loss=0.05682, over 16638.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2703, pruned_loss=0.04456, over 3179248.34 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:06,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0556, 2.8023, 3.1630, 1.6826, 3.2524, 3.3164, 2.6821, 2.4465], device='cuda:2'), covar=tensor([0.0906, 0.0340, 0.0174, 0.1325, 0.0101, 0.0181, 0.0491, 0.0550], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0111, 0.0101, 0.0139, 0.0084, 0.0130, 0.0130, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:23:31,264 INFO [train.py:904] (2/8) Epoch 25, batch 5000, loss[loss=0.1736, simple_loss=0.269, pruned_loss=0.03909, over 16703.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2723, pruned_loss=0.04491, over 3177727.08 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:51,579 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5626, 4.4012, 4.2917, 2.7975, 3.8227, 4.3875, 3.7748, 2.4480], device='cuda:2'), covar=tensor([0.0542, 0.0032, 0.0040, 0.0423, 0.0089, 0.0069, 0.0092, 0.0459], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0134, 0.0100, 0.0111, 0.0097, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 00:24:25,383 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248640.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:26,615 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1973, 2.0795, 1.6908, 1.7876, 2.2783, 1.9457, 1.8602, 2.3603], device='cuda:2'), covar=tensor([0.0204, 0.0455, 0.0602, 0.0502, 0.0271, 0.0363, 0.0178, 0.0277], device='cuda:2'), in_proj_covar=tensor([0.0222, 0.0242, 0.0230, 0.0233, 0.0244, 0.0242, 0.0243, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:24:27,922 INFO [optim.py:368] (2/8) 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 00:24:34,801 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 00:24:42,903 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:43,830 INFO [train.py:904] (2/8) Epoch 25, batch 5050, loss[loss=0.1839, simple_loss=0.2759, pruned_loss=0.04589, over 16836.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2729, pruned_loss=0.04477, over 3187641.62 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:25:35,138 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:35,152 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:39,548 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:57,288 INFO [train.py:904] (2/8) Epoch 25, batch 5100, loss[loss=0.2018, simple_loss=0.289, pruned_loss=0.05732, over 12111.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2711, pruned_loss=0.04416, over 3185343.58 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:26:15,840 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:45,319 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6392, 1.9081, 2.2381, 2.6593, 2.6213, 3.0399, 2.0490, 2.9236], device='cuda:2'), covar=tensor([0.0230, 0.0532, 0.0395, 0.0356, 0.0339, 0.0181, 0.0562, 0.0132], device='cuda:2'), in_proj_covar=tensor([0.0193, 0.0194, 0.0182, 0.0187, 0.0201, 0.0159, 0.0198, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:26:49,998 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:52,085 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248739.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:55,412 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5914, 3.8170, 2.8148, 2.2548, 2.4690, 2.5208, 3.9304, 3.3626], device='cuda:2'), covar=tensor([0.2791, 0.0525, 0.1842, 0.2789, 0.2538, 0.1910, 0.0462, 0.1182], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0272, 0.0310, 0.0319, 0.0302, 0.0269, 0.0301, 0.0346], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 00:26:56,587 INFO [optim.py:368] (2/8) 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 00:27:12,690 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:27:13,424 INFO [train.py:904] (2/8) Epoch 25, batch 5150, loss[loss=0.1558, simple_loss=0.2518, pruned_loss=0.02983, over 16506.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2711, pruned_loss=0.04326, over 3191586.03 frames. ], batch size: 68, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:27:41,963 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6403, 3.6884, 1.9950, 4.2265, 2.6791, 4.0886, 2.2610, 2.9625], device='cuda:2'), covar=tensor([0.0268, 0.0347, 0.1874, 0.0193, 0.0887, 0.0540, 0.1710, 0.0822], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0179, 0.0194, 0.0168, 0.0177, 0.0218, 0.0202, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:27:44,174 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8273, 3.7650, 3.8908, 4.0122, 4.0879, 3.6914, 4.0451, 4.1187], device='cuda:2'), covar=tensor([0.1545, 0.1075, 0.1327, 0.0684, 0.0564, 0.1876, 0.0847, 0.0735], device='cuda:2'), in_proj_covar=tensor([0.0657, 0.0808, 0.0930, 0.0818, 0.0625, 0.0642, 0.0671, 0.0784], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:27:47,494 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:27:57,842 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 00:28:01,899 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6121, 2.5779, 1.8723, 2.7061, 2.0009, 2.7874, 2.1358, 2.3831], device='cuda:2'), covar=tensor([0.0276, 0.0333, 0.1278, 0.0228, 0.0701, 0.0595, 0.1171, 0.0603], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0168, 0.0177, 0.0219, 0.0202, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:28:03,431 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:18,022 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-02 00:28:28,598 INFO [train.py:904] (2/8) Epoch 25, batch 5200, loss[loss=0.1944, simple_loss=0.2911, pruned_loss=0.0489, over 11987.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2702, pruned_loss=0.04334, over 3186724.68 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:29:25,194 INFO [optim.py:368] (2/8) 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 00:29:41,113 INFO [train.py:904] (2/8) Epoch 25, batch 5250, loss[loss=0.1639, simple_loss=0.2474, pruned_loss=0.04024, over 17024.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2679, pruned_loss=0.04303, over 3183164.14 frames. ], batch size: 55, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:30:53,686 INFO [train.py:904] (2/8) Epoch 25, batch 5300, loss[loss=0.1698, simple_loss=0.2576, pruned_loss=0.04102, over 15489.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2645, pruned_loss=0.0419, over 3193811.71 frames. ], batch size: 191, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:31:51,341 INFO [optim.py:368] (2/8) 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 00:32:04,273 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1677, 2.3378, 2.7472, 3.1550, 3.0443, 3.8177, 2.4195, 3.6186], device='cuda:2'), covar=tensor([0.0254, 0.0530, 0.0381, 0.0339, 0.0339, 0.0152, 0.0529, 0.0144], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0187, 0.0202, 0.0160, 0.0199, 0.0161], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:32:07,998 INFO [train.py:904] (2/8) Epoch 25, batch 5350, loss[loss=0.1794, simple_loss=0.273, pruned_loss=0.04286, over 16908.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2632, pruned_loss=0.04132, over 3205597.96 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:32:21,412 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0690, 3.9758, 4.1183, 4.2543, 4.3937, 3.9906, 4.3416, 4.3961], device='cuda:2'), covar=tensor([0.1646, 0.1140, 0.1488, 0.0730, 0.0497, 0.1310, 0.0800, 0.0691], device='cuda:2'), in_proj_covar=tensor([0.0659, 0.0811, 0.0933, 0.0820, 0.0625, 0.0644, 0.0673, 0.0786], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:32:29,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9855, 2.2680, 2.3583, 2.7265, 1.9138, 3.2354, 1.7930, 2.7836], device='cuda:2'), covar=tensor([0.1158, 0.0663, 0.1033, 0.0153, 0.0088, 0.0323, 0.1441, 0.0650], device='cuda:2'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0193, 0.0203, 0.0214, 0.0204, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:33:00,481 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248988.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:33:08,891 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8912, 2.8121, 2.4079, 2.6978, 3.1346, 2.8513, 3.3399, 3.4095], device='cuda:2'), covar=tensor([0.0083, 0.0395, 0.0509, 0.0431, 0.0293, 0.0367, 0.0247, 0.0260], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0240, 0.0228, 0.0230, 0.0241, 0.0240, 0.0239, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:33:18,196 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 00:33:22,803 INFO [train.py:904] (2/8) Epoch 25, batch 5400, loss[loss=0.1854, simple_loss=0.2853, pruned_loss=0.04279, over 15434.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2655, pruned_loss=0.04207, over 3187221.39 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:34:11,327 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:17,372 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249039.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:20,469 INFO [optim.py:368] (2/8) 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 00:34:29,102 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249047.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:40,330 INFO [train.py:904] (2/8) Epoch 25, batch 5450, loss[loss=0.1836, simple_loss=0.2781, pruned_loss=0.04454, over 16441.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2681, pruned_loss=0.04333, over 3169615.06 frames. ], batch size: 75, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:34:46,622 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0050, 4.2945, 3.3296, 2.7957, 3.2712, 2.9470, 4.8385, 3.8598], device='cuda:2'), covar=tensor([0.2695, 0.0679, 0.1645, 0.2319, 0.2210, 0.1689, 0.0361, 0.1131], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0273, 0.0310, 0.0320, 0.0302, 0.0269, 0.0302, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 00:34:52,842 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7870, 2.0465, 2.4553, 2.7511, 2.6825, 3.2701, 2.0984, 3.1590], device='cuda:2'), covar=tensor([0.0231, 0.0481, 0.0297, 0.0336, 0.0310, 0.0184, 0.0531, 0.0151], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0194, 0.0182, 0.0187, 0.0201, 0.0160, 0.0199, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:35:08,684 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:34,360 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249087.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:58,372 INFO [train.py:904] (2/8) Epoch 25, batch 5500, loss[loss=0.2515, simple_loss=0.3225, pruned_loss=0.09027, over 11664.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2746, pruned_loss=0.04733, over 3144208.07 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:36:46,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6952, 4.2842, 4.2179, 2.8590, 3.8014, 4.3159, 3.8075, 2.5096], device='cuda:2'), covar=tensor([0.0499, 0.0053, 0.0059, 0.0413, 0.0114, 0.0126, 0.0102, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 00:37:00,872 INFO [optim.py:368] (2/8) 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 00:37:17,248 INFO [train.py:904] (2/8) Epoch 25, batch 5550, loss[loss=0.2205, simple_loss=0.3076, pruned_loss=0.06668, over 16749.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.282, pruned_loss=0.05233, over 3117726.66 frames. ], batch size: 124, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:38:34,929 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6128, 2.4039, 2.3140, 3.5705, 2.1619, 3.6238, 1.6087, 2.5189], device='cuda:2'), covar=tensor([0.1562, 0.0983, 0.1473, 0.0230, 0.0208, 0.0515, 0.1868, 0.1064], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0205, 0.0215, 0.0206, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:38:41,395 INFO [train.py:904] (2/8) Epoch 25, batch 5600, loss[loss=0.2653, simple_loss=0.3286, pruned_loss=0.101, over 11373.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2872, pruned_loss=0.05692, over 3074626.56 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:39:02,214 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:39:32,647 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 00:39:47,608 INFO [optim.py:368] (2/8) 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 00:39:49,283 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2729, 3.4468, 3.5833, 3.5665, 3.5798, 3.4161, 3.4400, 3.4943], device='cuda:2'), covar=tensor([0.0477, 0.0824, 0.0535, 0.0514, 0.0563, 0.0698, 0.0922, 0.0622], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0469, 0.0458, 0.0417, 0.0501, 0.0479, 0.0558, 0.0381], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 00:40:04,580 INFO [train.py:904] (2/8) Epoch 25, batch 5650, loss[loss=0.2364, simple_loss=0.3149, pruned_loss=0.07896, over 15265.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2921, pruned_loss=0.06088, over 3048386.11 frames. ], batch size: 191, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:40:41,216 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:40:42,307 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7807, 3.6539, 3.8574, 3.6198, 3.8272, 4.2078, 3.8782, 3.5888], device='cuda:2'), covar=tensor([0.1986, 0.2635, 0.2988, 0.2682, 0.2692, 0.1917, 0.1809, 0.2793], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0618, 0.0676, 0.0502, 0.0669, 0.0707, 0.0526, 0.0675], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 00:41:21,944 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 00:41:22,857 INFO [train.py:904] (2/8) Epoch 25, batch 5700, loss[loss=0.2339, simple_loss=0.2972, pruned_loss=0.08531, over 11374.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2932, pruned_loss=0.06186, over 3054125.08 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:42:02,857 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5622, 1.7626, 2.2530, 2.4934, 2.4839, 2.9306, 1.9592, 2.8213], device='cuda:2'), covar=tensor([0.0244, 0.0536, 0.0328, 0.0331, 0.0329, 0.0192, 0.0560, 0.0158], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0193, 0.0181, 0.0186, 0.0200, 0.0159, 0.0198, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:42:25,334 INFO [optim.py:368] (2/8) 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 00:42:34,471 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:42:43,861 INFO [train.py:904] (2/8) Epoch 25, batch 5750, loss[loss=0.1811, simple_loss=0.2673, pruned_loss=0.04747, over 17029.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2955, pruned_loss=0.06332, over 3026825.59 frames. ], batch size: 55, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:43:13,865 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:43:19,920 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-05-02 00:43:53,995 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:44:01,014 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6341, 2.5170, 2.3731, 4.4282, 3.0080, 3.9928, 1.5844, 2.9152], device='cuda:2'), covar=tensor([0.1671, 0.0979, 0.1470, 0.0224, 0.0280, 0.0458, 0.1967, 0.0925], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0205, 0.0215, 0.0206, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:44:07,472 INFO [train.py:904] (2/8) Epoch 25, batch 5800, loss[loss=0.2004, simple_loss=0.295, pruned_loss=0.05288, over 16765.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2957, pruned_loss=0.06306, over 3006735.98 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:44:32,820 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 00:45:26,284 INFO [train.py:904] (2/8) Epoch 25, batch 5850, loss[loss=0.2082, simple_loss=0.2852, pruned_loss=0.06557, over 11232.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2932, pruned_loss=0.06083, over 3026669.05 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:46:46,937 INFO [train.py:904] (2/8) Epoch 25, batch 5900, loss[loss=0.1821, simple_loss=0.2775, pruned_loss=0.04336, over 17220.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2926, pruned_loss=0.06044, over 3046886.74 frames. ], batch size: 45, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:47:19,127 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 00:47:52,231 INFO [optim.py:368] (2/8) 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 00:48:08,075 INFO [train.py:904] (2/8) Epoch 25, batch 5950, loss[loss=0.1799, simple_loss=0.2756, pruned_loss=0.04205, over 16836.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2929, pruned_loss=0.05931, over 3044320.63 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:48:36,950 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:49:28,142 INFO [train.py:904] (2/8) Epoch 25, batch 6000, loss[loss=0.1964, simple_loss=0.28, pruned_loss=0.05638, over 16548.00 frames. ], tot_loss[loss=0.205, simple_loss=0.292, pruned_loss=0.05903, over 3038202.97 frames. ], batch size: 68, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:49:28,143 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 00:49:38,591 INFO [train.py:938] (2/8) Epoch 25, validation: loss=0.1487, simple_loss=0.2613, pruned_loss=0.01811, over 944034.00 frames. 2023-05-02 00:49:38,592 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 00:50:36,538 INFO [optim.py:368] (2/8) 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 00:50:37,727 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2455, 3.7191, 3.7260, 2.3645, 3.4464, 3.7882, 3.4833, 2.0956], device='cuda:2'), covar=tensor([0.0604, 0.0068, 0.0077, 0.0493, 0.0121, 0.0148, 0.0111, 0.0514], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0136, 0.0100, 0.0113, 0.0097, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 00:50:46,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5077, 3.5819, 3.3731, 3.0502, 3.1795, 3.4754, 3.2852, 3.2975], device='cuda:2'), covar=tensor([0.0597, 0.0648, 0.0317, 0.0279, 0.0558, 0.0481, 0.1386, 0.0500], device='cuda:2'), in_proj_covar=tensor([0.0300, 0.0451, 0.0350, 0.0355, 0.0355, 0.0409, 0.0240, 0.0424], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:50:53,940 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9631, 2.1597, 2.2824, 3.4175, 2.1231, 2.4404, 2.2726, 2.2866], device='cuda:2'), covar=tensor([0.1589, 0.3422, 0.2937, 0.0684, 0.4290, 0.2443, 0.3599, 0.3318], device='cuda:2'), in_proj_covar=tensor([0.0413, 0.0463, 0.0379, 0.0332, 0.0442, 0.0529, 0.0433, 0.0539], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:50:54,725 INFO [train.py:904] (2/8) Epoch 25, batch 6050, loss[loss=0.2122, simple_loss=0.3002, pruned_loss=0.0621, over 15176.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2904, pruned_loss=0.05831, over 3035570.86 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:51:05,143 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5571, 1.6888, 2.1471, 2.4683, 2.5211, 2.7597, 1.6908, 2.7062], device='cuda:2'), covar=tensor([0.0210, 0.0603, 0.0334, 0.0336, 0.0315, 0.0192, 0.0766, 0.0165], device='cuda:2'), in_proj_covar=tensor([0.0193, 0.0193, 0.0182, 0.0186, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:52:12,397 INFO [train.py:904] (2/8) Epoch 25, batch 6100, loss[loss=0.2026, simple_loss=0.2939, pruned_loss=0.05563, over 16941.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2896, pruned_loss=0.05742, over 3050854.37 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:53:15,555 INFO [optim.py:368] (2/8) 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 00:53:29,950 INFO [train.py:904] (2/8) Epoch 25, batch 6150, loss[loss=0.1876, simple_loss=0.2787, pruned_loss=0.04822, over 16761.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2874, pruned_loss=0.0565, over 3063718.28 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:54:20,285 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7548, 4.8560, 4.6315, 4.2601, 4.2942, 4.7399, 4.5541, 4.4063], device='cuda:2'), covar=tensor([0.0684, 0.0610, 0.0333, 0.0356, 0.1001, 0.0590, 0.0429, 0.0717], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0449, 0.0348, 0.0353, 0.0354, 0.0407, 0.0239, 0.0422], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:54:51,553 INFO [train.py:904] (2/8) Epoch 25, batch 6200, loss[loss=0.1899, simple_loss=0.2768, pruned_loss=0.05149, over 15320.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2858, pruned_loss=0.05601, over 3071973.59 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:55:09,855 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8189, 5.1089, 4.8515, 4.8945, 4.5899, 4.5944, 4.5617, 5.2062], device='cuda:2'), covar=tensor([0.1303, 0.0908, 0.1020, 0.0978, 0.0885, 0.1057, 0.1180, 0.0898], device='cuda:2'), in_proj_covar=tensor([0.0692, 0.0835, 0.0690, 0.0646, 0.0531, 0.0532, 0.0702, 0.0654], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:55:09,899 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:55:36,651 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3272, 2.3719, 2.3737, 4.0379, 2.3647, 2.7558, 2.4313, 2.5462], device='cuda:2'), covar=tensor([0.1327, 0.3419, 0.2910, 0.0525, 0.3826, 0.2519, 0.3432, 0.3239], device='cuda:2'), in_proj_covar=tensor([0.0414, 0.0463, 0.0380, 0.0332, 0.0442, 0.0530, 0.0433, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 00:55:55,230 INFO [optim.py:368] (2/8) 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 00:56:10,146 INFO [train.py:904] (2/8) Epoch 25, batch 6250, loss[loss=0.1652, simple_loss=0.2577, pruned_loss=0.03632, over 17218.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2853, pruned_loss=0.05553, over 3084492.61 frames. ], batch size: 43, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:56:10,621 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0027, 4.0290, 4.3233, 4.3050, 4.3171, 4.0750, 4.0834, 4.0786], device='cuda:2'), covar=tensor([0.0376, 0.0696, 0.0432, 0.0422, 0.0470, 0.0434, 0.0939, 0.0511], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0469, 0.0456, 0.0417, 0.0500, 0.0477, 0.0555, 0.0381], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 00:56:20,464 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7095, 2.5201, 2.3544, 3.3018, 2.1738, 3.5922, 1.4846, 2.7381], device='cuda:2'), covar=tensor([0.1465, 0.0797, 0.1312, 0.0190, 0.0153, 0.0397, 0.1869, 0.0877], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0197, 0.0196, 0.0206, 0.0217, 0.0207, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:56:38,294 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249871.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:56:44,481 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:57:26,272 INFO [train.py:904] (2/8) Epoch 25, batch 6300, loss[loss=0.2259, simple_loss=0.2943, pruned_loss=0.07877, over 11621.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2862, pruned_loss=0.05603, over 3069041.64 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:57:38,477 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-05-02 00:57:52,580 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249919.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:58:25,151 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 00:58:29,093 INFO [optim.py:368] (2/8) 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 00:58:45,042 INFO [train.py:904] (2/8) Epoch 25, batch 6350, loss[loss=0.2305, simple_loss=0.2959, pruned_loss=0.08253, over 11407.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2868, pruned_loss=0.05631, over 3079765.88 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:58:48,582 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249955.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 00:58:56,639 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0354, 3.0630, 1.8239, 3.2570, 2.2111, 3.3006, 2.1372, 2.6011], device='cuda:2'), covar=tensor([0.0327, 0.0426, 0.1643, 0.0267, 0.0900, 0.0664, 0.1448, 0.0737], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0181, 0.0196, 0.0169, 0.0179, 0.0220, 0.0204, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 00:59:20,549 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249976.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:00:04,121 INFO [train.py:904] (2/8) Epoch 25, batch 6400, loss[loss=0.2265, simple_loss=0.299, pruned_loss=0.07701, over 11209.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.287, pruned_loss=0.05731, over 3073872.36 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:00:23,851 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250016.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:00:56,407 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:01:05,937 INFO [optim.py:368] (2/8) 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 01:01:09,151 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8451, 3.9493, 4.0978, 4.1044, 4.1285, 3.9364, 3.9395, 3.9115], device='cuda:2'), covar=tensor([0.0341, 0.0556, 0.0501, 0.0454, 0.0410, 0.0433, 0.0796, 0.0511], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0473, 0.0460, 0.0420, 0.0506, 0.0482, 0.0559, 0.0384], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 01:01:20,199 INFO [train.py:904] (2/8) Epoch 25, batch 6450, loss[loss=0.1907, simple_loss=0.278, pruned_loss=0.05168, over 16393.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.287, pruned_loss=0.05666, over 3085390.25 frames. ], batch size: 146, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:01:27,545 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7516, 4.6264, 4.8128, 4.9395, 5.1059, 4.5521, 5.1111, 5.1178], device='cuda:2'), covar=tensor([0.1929, 0.1255, 0.1527, 0.0763, 0.0599, 0.1031, 0.0724, 0.0606], device='cuda:2'), in_proj_covar=tensor([0.0652, 0.0802, 0.0926, 0.0811, 0.0621, 0.0639, 0.0670, 0.0780], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:02:36,771 INFO [train.py:904] (2/8) Epoch 25, batch 6500, loss[loss=0.2633, simple_loss=0.3155, pruned_loss=0.1055, over 11200.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2852, pruned_loss=0.05617, over 3077443.77 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:02:43,266 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 01:03:36,087 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 01:03:40,562 INFO [optim.py:368] (2/8) 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 01:03:52,679 INFO [train.py:904] (2/8) Epoch 25, batch 6550, loss[loss=0.1976, simple_loss=0.2993, pruned_loss=0.04796, over 16188.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2877, pruned_loss=0.05701, over 3080198.22 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 4.0 2023-05-02 01:04:18,236 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:04:40,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9611, 2.1524, 2.2861, 3.4309, 2.0893, 2.3983, 2.2420, 2.2796], device='cuda:2'), covar=tensor([0.1481, 0.3366, 0.2836, 0.0651, 0.4125, 0.2436, 0.3400, 0.3330], device='cuda:2'), in_proj_covar=tensor([0.0413, 0.0462, 0.0379, 0.0332, 0.0442, 0.0530, 0.0433, 0.0540], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:04:43,144 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-05-02 01:05:05,603 INFO [train.py:904] (2/8) Epoch 25, batch 6600, loss[loss=0.2014, simple_loss=0.2844, pruned_loss=0.05915, over 16735.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2898, pruned_loss=0.05759, over 3076870.75 frames. ], batch size: 57, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:06:08,493 INFO [optim.py:368] (2/8) 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 01:06:21,898 INFO [train.py:904] (2/8) Epoch 25, batch 6650, loss[loss=0.1691, simple_loss=0.2543, pruned_loss=0.04194, over 16557.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2903, pruned_loss=0.05855, over 3066752.96 frames. ], batch size: 68, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:37,544 INFO [train.py:904] (2/8) Epoch 25, batch 6700, loss[loss=0.2225, simple_loss=0.2915, pruned_loss=0.07674, over 11911.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2896, pruned_loss=0.05923, over 3045398.57 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:50,662 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250311.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 01:08:08,450 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2734, 2.4473, 2.1138, 2.3150, 2.8205, 2.4631, 2.8424, 3.0569], device='cuda:2'), covar=tensor([0.0142, 0.0474, 0.0567, 0.0487, 0.0303, 0.0441, 0.0224, 0.0266], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0239, 0.0229, 0.0230, 0.0240, 0.0238, 0.0239, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:08:23,195 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:08:41,137 INFO [optim.py:368] (2/8) 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 01:08:53,999 INFO [train.py:904] (2/8) Epoch 25, batch 6750, loss[loss=0.1878, simple_loss=0.2735, pruned_loss=0.0511, over 16876.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2879, pruned_loss=0.0588, over 3058526.28 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:09:29,695 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6914, 1.9426, 2.3052, 2.7076, 2.6447, 3.0624, 1.9345, 3.0177], device='cuda:2'), covar=tensor([0.0229, 0.0559, 0.0348, 0.0327, 0.0318, 0.0202, 0.0581, 0.0150], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0194, 0.0182, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:10:10,572 INFO [train.py:904] (2/8) Epoch 25, batch 6800, loss[loss=0.1926, simple_loss=0.2888, pruned_loss=0.04821, over 16731.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2875, pruned_loss=0.05801, over 3084173.18 frames. ], batch size: 89, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:16,510 INFO [optim.py:368] (2/8) 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 01:11:27,491 INFO [train.py:904] (2/8) Epoch 25, batch 6850, loss[loss=0.2075, simple_loss=0.3024, pruned_loss=0.05626, over 16242.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2889, pruned_loss=0.05854, over 3068962.12 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:43,167 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2390, 2.0024, 2.6632, 3.2311, 2.9204, 3.5964, 2.2014, 3.6561], device='cuda:2'), covar=tensor([0.0191, 0.0610, 0.0380, 0.0270, 0.0375, 0.0166, 0.0671, 0.0146], device='cuda:2'), in_proj_covar=tensor([0.0193, 0.0195, 0.0182, 0.0186, 0.0201, 0.0161, 0.0199, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:11:52,825 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 25, batch 6900, loss[loss=0.1841, simple_loss=0.282, pruned_loss=0.04307, over 16849.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2907, pruned_loss=0.05721, over 3098754.48 frames. ], batch size: 42, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:13:06,142 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:13:47,067 INFO [optim.py:368] (2/8) 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 01:14:00,190 INFO [train.py:904] (2/8) Epoch 25, batch 6950, loss[loss=0.1991, simple_loss=0.2901, pruned_loss=0.0541, over 16433.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2929, pruned_loss=0.05937, over 3078047.75 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:15:03,053 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7921, 4.5938, 4.7891, 4.9971, 5.2241, 4.6654, 5.1923, 5.1918], device='cuda:2'), covar=tensor([0.2312, 0.1530, 0.2275, 0.0985, 0.0834, 0.1085, 0.1029, 0.0741], device='cuda:2'), in_proj_covar=tensor([0.0650, 0.0800, 0.0925, 0.0809, 0.0621, 0.0639, 0.0671, 0.0780], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:15:18,385 INFO [train.py:904] (2/8) Epoch 25, batch 7000, loss[loss=0.2002, simple_loss=0.3009, pruned_loss=0.04971, over 16483.00 frames. ], tot_loss[loss=0.205, simple_loss=0.293, pruned_loss=0.05848, over 3091635.08 frames. ], batch size: 75, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:15:21,960 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6207, 2.5640, 1.8123, 2.7074, 2.0530, 2.7804, 2.1247, 2.3949], device='cuda:2'), covar=tensor([0.0352, 0.0457, 0.1400, 0.0272, 0.0734, 0.0592, 0.1278, 0.0627], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0180, 0.0196, 0.0168, 0.0178, 0.0219, 0.0203, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:15:31,352 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250611.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:16:03,414 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:16:22,009 INFO [optim.py:368] (2/8) 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 01:16:35,090 INFO [train.py:904] (2/8) Epoch 25, batch 7050, loss[loss=0.192, simple_loss=0.2854, pruned_loss=0.04932, over 16720.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2943, pruned_loss=0.05853, over 3099630.47 frames. ], batch size: 124, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:16:44,699 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250659.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:16:49,616 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250662.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:17,438 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250680.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:51,094 INFO [train.py:904] (2/8) Epoch 25, batch 7100, loss[loss=0.2414, simple_loss=0.3055, pruned_loss=0.08869, over 11228.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2931, pruned_loss=0.05895, over 3070112.00 frames. ], batch size: 250, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:18:23,459 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250723.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:18:56,841 INFO [optim.py:368] (2/8) 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 01:19:09,288 INFO [train.py:904] (2/8) Epoch 25, batch 7150, loss[loss=0.1942, simple_loss=0.279, pruned_loss=0.05473, over 17046.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2906, pruned_loss=0.05801, over 3104383.88 frames. ], batch size: 55, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:20:22,681 INFO [train.py:904] (2/8) Epoch 25, batch 7200, loss[loss=0.1843, simple_loss=0.2654, pruned_loss=0.05155, over 11390.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2883, pruned_loss=0.05622, over 3100639.09 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:21:01,687 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 01:21:08,881 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,112 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,840 INFO [optim.py:368] (2/8) 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 01:21:41,038 INFO [train.py:904] (2/8) Epoch 25, batch 7250, loss[loss=0.1927, simple_loss=0.2766, pruned_loss=0.05444, over 15351.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2858, pruned_loss=0.05508, over 3089372.44 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:22:12,968 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8301, 3.9429, 2.5166, 4.6601, 3.0589, 4.5761, 2.7614, 3.2424], device='cuda:2'), covar=tensor([0.0303, 0.0378, 0.1744, 0.0240, 0.0838, 0.0484, 0.1435, 0.0786], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0180, 0.0196, 0.0168, 0.0178, 0.0219, 0.0204, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:22:42,091 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:22:55,036 INFO [train.py:904] (2/8) Epoch 25, batch 7300, loss[loss=0.1751, simple_loss=0.2723, pruned_loss=0.03895, over 16693.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.286, pruned_loss=0.0551, over 3091118.86 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:23:00,832 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250906.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 01:23:59,649 INFO [optim.py:368] (2/8) 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 01:24:09,701 INFO [train.py:904] (2/8) Epoch 25, batch 7350, loss[loss=0.2331, simple_loss=0.2987, pruned_loss=0.08379, over 11084.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2869, pruned_loss=0.05622, over 3086311.11 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:21,491 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9280, 2.7310, 2.6589, 1.9936, 2.5865, 2.7220, 2.6079, 1.9671], device='cuda:2'), covar=tensor([0.0406, 0.0097, 0.0093, 0.0361, 0.0149, 0.0147, 0.0131, 0.0396], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0087, 0.0087, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 01:25:27,063 INFO [train.py:904] (2/8) Epoch 25, batch 7400, loss[loss=0.1962, simple_loss=0.2996, pruned_loss=0.0464, over 16796.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2874, pruned_loss=0.0565, over 3097086.38 frames. ], batch size: 102, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:50,965 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:26:35,986 INFO [optim.py:368] (2/8) 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 01:26:46,723 INFO [train.py:904] (2/8) Epoch 25, batch 7450, loss[loss=0.2026, simple_loss=0.293, pruned_loss=0.05607, over 16862.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2884, pruned_loss=0.05732, over 3088424.13 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:28:06,270 INFO [train.py:904] (2/8) Epoch 25, batch 7500, loss[loss=0.2158, simple_loss=0.3056, pruned_loss=0.06304, over 16261.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2883, pruned_loss=0.05654, over 3087459.52 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:28:12,261 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0615, 2.4671, 2.5891, 1.9623, 2.6940, 2.7899, 2.3943, 2.3654], device='cuda:2'), covar=tensor([0.0745, 0.0276, 0.0269, 0.0974, 0.0146, 0.0343, 0.0481, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0139, 0.0084, 0.0129, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:29:12,775 INFO [optim.py:368] (2/8) 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 01:29:23,952 INFO [train.py:904] (2/8) Epoch 25, batch 7550, loss[loss=0.185, simple_loss=0.2766, pruned_loss=0.04671, over 16210.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2878, pruned_loss=0.0573, over 3070972.16 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:29:27,060 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5539, 3.5723, 2.7457, 2.2034, 2.3895, 2.4002, 3.7737, 3.2653], device='cuda:2'), covar=tensor([0.3053, 0.0635, 0.1915, 0.2806, 0.2710, 0.2169, 0.0449, 0.1334], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0273, 0.0311, 0.0321, 0.0304, 0.0270, 0.0301, 0.0345], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 01:30:19,158 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:32,668 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:37,926 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251201.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 01:30:39,949 INFO [train.py:904] (2/8) Epoch 25, batch 7600, loss[loss=0.1881, simple_loss=0.2758, pruned_loss=0.05019, over 16335.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2877, pruned_loss=0.05758, over 3078501.96 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:30:50,322 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9828, 4.9539, 4.7812, 4.0504, 4.8953, 1.8216, 4.6247, 4.4050], device='cuda:2'), covar=tensor([0.0107, 0.0094, 0.0193, 0.0393, 0.0086, 0.2801, 0.0136, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0164, 0.0204, 0.0181, 0.0181, 0.0211, 0.0193, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:30:51,960 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-02 01:31:43,658 INFO [optim.py:368] (2/8) 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 01:31:53,103 INFO [train.py:904] (2/8) Epoch 25, batch 7650, loss[loss=0.2027, simple_loss=0.29, pruned_loss=0.05769, over 16363.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2876, pruned_loss=0.05781, over 3081017.32 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:31:54,091 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 01:32:02,524 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251259.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:32:34,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1726, 3.0548, 3.2506, 1.6775, 3.3364, 3.4589, 2.8251, 2.5637], device='cuda:2'), covar=tensor([0.0931, 0.0308, 0.0235, 0.1420, 0.0138, 0.0246, 0.0451, 0.0556], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0139, 0.0084, 0.0130, 0.0128, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:32:35,756 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 01:33:06,978 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:33:08,935 INFO [train.py:904] (2/8) Epoch 25, batch 7700, loss[loss=0.1878, simple_loss=0.2743, pruned_loss=0.05068, over 17198.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2873, pruned_loss=0.05825, over 3079551.37 frames. ], batch size: 45, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:33:15,435 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 01:33:31,769 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:34:14,481 INFO [optim.py:368] (2/8) 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 01:34:25,531 INFO [train.py:904] (2/8) Epoch 25, batch 7750, loss[loss=0.2077, simple_loss=0.2879, pruned_loss=0.06376, over 16642.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2871, pruned_loss=0.05791, over 3076121.28 frames. ], batch size: 57, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:34:39,706 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:34:45,579 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:35:40,759 INFO [train.py:904] (2/8) Epoch 25, batch 7800, loss[loss=0.1763, simple_loss=0.2679, pruned_loss=0.04234, over 16588.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2881, pruned_loss=0.05887, over 3064813.00 frames. ], batch size: 68, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:35:42,325 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251404.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:36:45,151 INFO [optim.py:368] (2/8) 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 01:36:55,301 INFO [train.py:904] (2/8) Epoch 25, batch 7850, loss[loss=0.1803, simple_loss=0.2813, pruned_loss=0.0397, over 16887.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2885, pruned_loss=0.05817, over 3068818.07 frames. ], batch size: 96, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:37:13,938 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251465.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:37:50,393 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:37:55,408 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3796, 3.2868, 2.6232, 2.0854, 2.1287, 2.1965, 3.3964, 2.9301], device='cuda:2'), covar=tensor([0.3375, 0.0718, 0.1973, 0.3002, 0.2989, 0.2511, 0.0503, 0.1602], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0274, 0.0313, 0.0322, 0.0305, 0.0271, 0.0301, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 01:38:07,364 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251501.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:38:08,678 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-05-02 01:38:09,238 INFO [train.py:904] (2/8) Epoch 25, batch 7900, loss[loss=0.2432, simple_loss=0.308, pruned_loss=0.08921, over 11701.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2877, pruned_loss=0.05822, over 3051911.11 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:38:17,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3466, 2.8637, 3.1158, 1.9471, 2.7859, 2.0835, 3.0392, 3.1586], device='cuda:2'), covar=tensor([0.0281, 0.0823, 0.0592, 0.2222, 0.0846, 0.1102, 0.0691, 0.0853], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0167, 0.0169, 0.0155, 0.0147, 0.0131, 0.0144, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:38:37,406 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2421, 4.2000, 4.1301, 3.3290, 4.1740, 1.7279, 3.9505, 3.7174], device='cuda:2'), covar=tensor([0.0137, 0.0124, 0.0208, 0.0374, 0.0103, 0.2987, 0.0152, 0.0289], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0163, 0.0203, 0.0180, 0.0179, 0.0209, 0.0191, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:39:04,226 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:08,719 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0595, 3.0567, 1.6244, 3.2439, 2.2186, 3.2806, 1.8936, 2.5177], device='cuda:2'), covar=tensor([0.0330, 0.0447, 0.2106, 0.0281, 0.0926, 0.0648, 0.1888, 0.0838], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0180, 0.0197, 0.0168, 0.0178, 0.0220, 0.0205, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:39:17,278 INFO [optim.py:368] (2/8) 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 01:39:22,129 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:28,550 INFO [train.py:904] (2/8) Epoch 25, batch 7950, loss[loss=0.1924, simple_loss=0.2778, pruned_loss=0.05353, over 16530.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2884, pruned_loss=0.05871, over 3044947.95 frames. ], batch size: 68, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:39:30,108 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:40:27,537 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3849, 2.4996, 1.8971, 2.0977, 2.7812, 2.4351, 2.9179, 3.0934], device='cuda:2'), covar=tensor([0.0170, 0.0515, 0.0746, 0.0633, 0.0363, 0.0521, 0.0321, 0.0300], device='cuda:2'), in_proj_covar=tensor([0.0215, 0.0235, 0.0225, 0.0227, 0.0236, 0.0235, 0.0234, 0.0233], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:40:46,595 INFO [train.py:904] (2/8) Epoch 25, batch 8000, loss[loss=0.2126, simple_loss=0.2834, pruned_loss=0.07095, over 11345.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2894, pruned_loss=0.05939, over 3045165.64 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:41:43,311 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3378, 3.4423, 3.6034, 3.5925, 3.5987, 3.4012, 3.4712, 3.4774], device='cuda:2'), covar=tensor([0.0406, 0.0745, 0.0483, 0.0441, 0.0533, 0.0634, 0.0781, 0.0558], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0477, 0.0461, 0.0424, 0.0507, 0.0487, 0.0562, 0.0388], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 01:41:44,820 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 01:42:01,887 INFO [train.py:904] (2/8) Epoch 25, batch 8050, loss[loss=0.2118, simple_loss=0.3068, pruned_loss=0.05836, over 16705.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2891, pruned_loss=0.05905, over 3045267.86 frames. ], batch size: 76, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:42:09,017 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:42:09,117 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1658, 4.0437, 4.2273, 4.3504, 4.4934, 4.0697, 4.3854, 4.5021], device='cuda:2'), covar=tensor([0.1767, 0.1208, 0.1489, 0.0767, 0.0626, 0.1278, 0.0975, 0.0801], device='cuda:2'), in_proj_covar=tensor([0.0644, 0.0793, 0.0914, 0.0802, 0.0616, 0.0634, 0.0669, 0.0774], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:42:22,386 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5979, 1.8150, 2.1853, 2.5448, 2.5278, 2.9088, 1.9062, 2.8533], device='cuda:2'), covar=tensor([0.0230, 0.0554, 0.0393, 0.0380, 0.0364, 0.0216, 0.0569, 0.0168], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0195, 0.0182, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:43:16,983 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:43:17,637 INFO [train.py:904] (2/8) Epoch 25, batch 8100, loss[loss=0.1923, simple_loss=0.2844, pruned_loss=0.05009, over 16792.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2889, pruned_loss=0.05843, over 3059213.16 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:43:27,358 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251709.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:44:16,505 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4908, 3.2791, 3.7389, 1.9997, 3.8434, 3.8822, 2.9776, 2.9109], device='cuda:2'), covar=tensor([0.0791, 0.0311, 0.0182, 0.1142, 0.0089, 0.0193, 0.0439, 0.0471], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0084, 0.0129, 0.0129, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:44:22,960 INFO [optim.py:368] (2/8) 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 01:44:33,794 INFO [train.py:904] (2/8) Epoch 25, batch 8150, loss[loss=0.1791, simple_loss=0.2641, pruned_loss=0.04703, over 16588.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2858, pruned_loss=0.0573, over 3082262.54 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:44:44,516 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251760.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:00,594 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:50,905 INFO [train.py:904] (2/8) Epoch 25, batch 8200, loss[loss=0.229, simple_loss=0.2904, pruned_loss=0.08384, over 11479.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2836, pruned_loss=0.0567, over 3072356.80 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:46:59,192 INFO [optim.py:368] (2/8) 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] (2/8) attn_weights_entropy = tensor([1.9751, 1.5736, 1.9259, 2.0337, 2.2257, 2.3318, 1.8235, 2.2582], device='cuda:2'), covar=tensor([0.0267, 0.0526, 0.0333, 0.0394, 0.0353, 0.0238, 0.0534, 0.0191], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0194, 0.0181, 0.0184, 0.0200, 0.0159, 0.0197, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:47:10,659 INFO [train.py:904] (2/8) Epoch 25, batch 8250, loss[loss=0.1621, simple_loss=0.2517, pruned_loss=0.03622, over 12130.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2824, pruned_loss=0.05398, over 3067300.94 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:47:12,505 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:29,883 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:30,708 INFO [train.py:904] (2/8) Epoch 25, batch 8300, loss[loss=0.1756, simple_loss=0.2732, pruned_loss=0.03895, over 15134.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2797, pruned_loss=0.05117, over 3050550.98 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:41,162 INFO [optim.py:368] (2/8) 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 01:49:52,636 INFO [train.py:904] (2/8) Epoch 25, batch 8350, loss[loss=0.1805, simple_loss=0.2752, pruned_loss=0.04287, over 15169.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2791, pruned_loss=0.04915, over 3061873.48 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:59,394 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:50:55,675 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1931, 2.4299, 2.5907, 2.0095, 2.7307, 2.7762, 2.5447, 2.5166], device='cuda:2'), covar=tensor([0.0619, 0.0273, 0.0238, 0.0925, 0.0125, 0.0309, 0.0402, 0.0396], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0109, 0.0098, 0.0137, 0.0083, 0.0127, 0.0127, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 01:51:05,349 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251997.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:51:17,314 INFO [train.py:904] (2/8) Epoch 25, batch 8400, loss[loss=0.1585, simple_loss=0.258, pruned_loss=0.02949, over 16835.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2762, pruned_loss=0.04709, over 3043210.40 frames. ], batch size: 102, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:51:22,059 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252005.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:51:48,673 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1199, 3.7352, 4.1704, 2.2724, 4.3683, 4.3930, 3.3732, 3.5273], device='cuda:2'), covar=tensor([0.0565, 0.0232, 0.0193, 0.1055, 0.0062, 0.0145, 0.0366, 0.0318], device='cuda:2'), in_proj_covar=tensor([0.0146, 0.0109, 0.0098, 0.0137, 0.0083, 0.0127, 0.0127, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 01:51:54,256 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 01:52:04,307 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8744, 3.2544, 3.5384, 2.0965, 3.0410, 2.2389, 3.5016, 3.4551], device='cuda:2'), covar=tensor([0.0299, 0.0868, 0.0502, 0.2142, 0.0786, 0.1074, 0.0561, 0.0924], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0164, 0.0166, 0.0152, 0.0144, 0.0128, 0.0142, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:52:27,817 INFO [optim.py:368] (2/8) 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 01:52:40,178 INFO [train.py:904] (2/8) Epoch 25, batch 8450, loss[loss=0.1723, simple_loss=0.2709, pruned_loss=0.03684, over 16387.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2741, pruned_loss=0.04517, over 3049951.23 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:52:51,890 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252060.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:52:59,910 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:53:58,681 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2698, 4.3478, 4.5129, 4.2730, 4.3594, 4.8568, 4.4208, 4.0803], device='cuda:2'), covar=tensor([0.1613, 0.2054, 0.2231, 0.2158, 0.2665, 0.1035, 0.1593, 0.2585], device='cuda:2'), in_proj_covar=tensor([0.0411, 0.0600, 0.0665, 0.0493, 0.0654, 0.0689, 0.0517, 0.0661], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 01:54:01,927 INFO [train.py:904] (2/8) Epoch 25, batch 8500, loss[loss=0.1549, simple_loss=0.2537, pruned_loss=0.02801, over 16772.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.271, pruned_loss=0.04331, over 3041103.01 frames. ], batch size: 102, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:54:11,539 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252108.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:55:15,323 INFO [optim.py:368] (2/8) 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 01:55:28,836 INFO [train.py:904] (2/8) Epoch 25, batch 8550, loss[loss=0.1637, simple_loss=0.2456, pruned_loss=0.04086, over 12070.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2687, pruned_loss=0.0421, over 3041610.48 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:56:18,078 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2613, 3.4796, 3.9585, 2.0531, 3.2335, 2.5381, 3.7527, 3.7241], device='cuda:2'), covar=tensor([0.0228, 0.0807, 0.0480, 0.2190, 0.0753, 0.0977, 0.0568, 0.0873], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0164, 0.0166, 0.0152, 0.0144, 0.0129, 0.0142, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 01:57:09,035 INFO [train.py:904] (2/8) Epoch 25, batch 8600, loss[loss=0.1617, simple_loss=0.2578, pruned_loss=0.03279, over 16519.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2687, pruned_loss=0.04088, over 3039527.27 frames. ], batch size: 68, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:57:24,006 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0247, 1.8640, 1.6900, 1.5139, 2.0048, 1.5977, 1.6264, 1.9439], device='cuda:2'), covar=tensor([0.0233, 0.0372, 0.0462, 0.0436, 0.0256, 0.0332, 0.0213, 0.0259], device='cuda:2'), in_proj_covar=tensor([0.0212, 0.0233, 0.0223, 0.0224, 0.0233, 0.0232, 0.0230, 0.0230], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:57:29,983 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0858, 5.0768, 4.8660, 4.3062, 4.9502, 2.0967, 4.6774, 4.7093], device='cuda:2'), covar=tensor([0.0082, 0.0082, 0.0176, 0.0301, 0.0079, 0.2428, 0.0125, 0.0192], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0163, 0.0202, 0.0179, 0.0179, 0.0210, 0.0191, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 01:58:34,358 INFO [optim.py:368] (2/8) 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 01:58:37,356 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9505, 4.4928, 3.3463, 2.4184, 2.8405, 2.8780, 4.8194, 3.8370], device='cuda:2'), covar=tensor([0.2945, 0.0532, 0.1807, 0.3294, 0.3028, 0.1988, 0.0351, 0.1234], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0267, 0.0305, 0.0314, 0.0297, 0.0266, 0.0295, 0.0338], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 01:58:42,124 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 01:58:48,812 INFO [train.py:904] (2/8) Epoch 25, batch 8650, loss[loss=0.1728, simple_loss=0.2617, pruned_loss=0.04194, over 12213.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2668, pruned_loss=0.0395, over 3040622.98 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:00:23,894 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252297.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:00:33,683 INFO [train.py:904] (2/8) Epoch 25, batch 8700, loss[loss=0.1616, simple_loss=0.2529, pruned_loss=0.03517, over 17036.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2641, pruned_loss=0.03844, over 3052523.36 frames. ], batch size: 55, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:01:02,650 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3041, 3.3546, 1.9799, 3.6315, 2.5013, 3.6251, 2.3325, 2.8002], device='cuda:2'), covar=tensor([0.0337, 0.0395, 0.1804, 0.0236, 0.0893, 0.0543, 0.1546, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0163, 0.0174, 0.0213, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 02:01:53,003 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:01:54,399 INFO [optim.py:368] (2/8) 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 02:02:09,353 INFO [train.py:904] (2/8) Epoch 25, batch 8750, loss[loss=0.1756, simple_loss=0.2683, pruned_loss=0.04152, over 12101.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2639, pruned_loss=0.03774, over 3053881.32 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:02:41,187 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252365.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:03:07,979 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 02:04:01,511 INFO [train.py:904] (2/8) Epoch 25, batch 8800, loss[loss=0.1717, simple_loss=0.2642, pruned_loss=0.03955, over 16832.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2625, pruned_loss=0.03704, over 3046406.10 frames. ], batch size: 124, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:04:13,666 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8067, 1.4164, 1.7525, 1.6827, 1.8576, 1.9276, 1.6635, 1.8845], device='cuda:2'), covar=tensor([0.0303, 0.0466, 0.0251, 0.0327, 0.0355, 0.0223, 0.0451, 0.0155], device='cuda:2'), in_proj_covar=tensor([0.0189, 0.0192, 0.0181, 0.0182, 0.0199, 0.0158, 0.0196, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:04:21,826 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:30,817 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252417.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:07,192 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:17,821 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9898, 4.2505, 4.1357, 4.1570, 3.8346, 3.8805, 3.8932, 4.2688], device='cuda:2'), covar=tensor([0.1046, 0.0863, 0.0843, 0.0744, 0.0693, 0.1762, 0.0950, 0.0920], device='cuda:2'), in_proj_covar=tensor([0.0676, 0.0817, 0.0675, 0.0631, 0.0518, 0.0526, 0.0685, 0.0638], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:05:28,895 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-05-02 02:05:31,485 INFO [optim.py:368] (2/8) 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 02:05:46,036 INFO [train.py:904] (2/8) Epoch 25, batch 8850, loss[loss=0.1568, simple_loss=0.2453, pruned_loss=0.0342, over 12474.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2654, pruned_loss=0.03671, over 3039818.62 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:06:16,152 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 02:06:35,641 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9260, 2.2327, 2.3323, 3.0089, 1.8507, 3.2874, 1.6375, 2.7209], device='cuda:2'), covar=tensor([0.1262, 0.0760, 0.1144, 0.0169, 0.0082, 0.0362, 0.1664, 0.0725], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0176, 0.0194, 0.0192, 0.0201, 0.0213, 0.0204, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 02:06:39,694 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252478.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:07,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1217, 3.2439, 3.2228, 2.2735, 2.9693, 3.2855, 3.1633, 1.9986], device='cuda:2'), covar=tensor([0.0493, 0.0056, 0.0060, 0.0385, 0.0111, 0.0082, 0.0087, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0084, 0.0086, 0.0132, 0.0098, 0.0109, 0.0094, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 02:07:16,272 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:16,635 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 02:07:31,607 INFO [train.py:904] (2/8) Epoch 25, batch 8900, loss[loss=0.1799, simple_loss=0.272, pruned_loss=0.04397, over 16787.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2657, pruned_loss=0.03596, over 3053788.02 frames. ], batch size: 124, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:09:18,926 INFO [optim.py:368] (2/8) 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 02:09:34,562 INFO [train.py:904] (2/8) Epoch 25, batch 8950, loss[loss=0.1637, simple_loss=0.2587, pruned_loss=0.03431, over 16349.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2651, pruned_loss=0.036, over 3054468.47 frames. ], batch size: 166, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:10:59,618 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252592.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:11:21,906 INFO [train.py:904] (2/8) Epoch 25, batch 9000, loss[loss=0.1641, simple_loss=0.2538, pruned_loss=0.03724, over 16866.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.262, pruned_loss=0.03493, over 3060814.15 frames. ], batch size: 116, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:11:21,906 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 02:11:31,586 INFO [train.py:938] (2/8) Epoch 25, validation: loss=0.1442, simple_loss=0.248, pruned_loss=0.02014, over 944034.00 frames. 2023-05-02 02:11:31,587 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 02:11:53,538 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:01,733 INFO [optim.py:368] (2/8) 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 02:13:14,919 INFO [train.py:904] (2/8) Epoch 25, batch 9050, loss[loss=0.1476, simple_loss=0.2431, pruned_loss=0.02607, over 16289.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2633, pruned_loss=0.03554, over 3064701.78 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:13:15,916 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252653.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:18,386 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5149, 3.2139, 3.4543, 1.7697, 3.6006, 3.6820, 3.0092, 2.8443], device='cuda:2'), covar=tensor([0.0718, 0.0277, 0.0193, 0.1390, 0.0098, 0.0180, 0.0419, 0.0474], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0107, 0.0095, 0.0135, 0.0081, 0.0124, 0.0125, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 02:13:47,173 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:56,512 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:14:58,382 INFO [train.py:904] (2/8) Epoch 25, batch 9100, loss[loss=0.1912, simple_loss=0.2853, pruned_loss=0.04862, over 15371.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2629, pruned_loss=0.036, over 3056516.30 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:15:50,529 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9821, 2.1128, 2.2121, 3.4548, 2.0581, 2.3600, 2.2580, 2.2095], device='cuda:2'), covar=tensor([0.1431, 0.3823, 0.3240, 0.0643, 0.4508, 0.2760, 0.3701, 0.3534], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0454, 0.0373, 0.0324, 0.0434, 0.0517, 0.0425, 0.0528], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:16:01,346 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:16:41,919 INFO [optim.py:368] (2/8) 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 02:16:55,035 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 02:16:56,608 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 02:16:57,785 INFO [train.py:904] (2/8) Epoch 25, batch 9150, loss[loss=0.163, simple_loss=0.2532, pruned_loss=0.0364, over 12046.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.263, pruned_loss=0.03582, over 3053674.74 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:17:42,773 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:20,263 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:20,657 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 02:18:43,278 INFO [train.py:904] (2/8) Epoch 25, batch 9200, loss[loss=0.1652, simple_loss=0.2627, pruned_loss=0.03389, over 16376.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2586, pruned_loss=0.03498, over 3048547.88 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:20:05,202 INFO [optim.py:368] (2/8) 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 02:20:19,098 INFO [train.py:904] (2/8) Epoch 25, batch 9250, loss[loss=0.1508, simple_loss=0.2336, pruned_loss=0.03401, over 12499.00 frames. ], tot_loss[loss=0.164, simple_loss=0.258, pruned_loss=0.03494, over 3037246.99 frames. ], batch size: 250, lr: 2.67e-03, grad_scale: 16.0 2023-05-02 02:20:28,212 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252857.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:20:37,652 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 02:22:11,611 INFO [train.py:904] (2/8) Epoch 25, batch 9300, loss[loss=0.1326, simple_loss=0.2285, pruned_loss=0.0184, over 16711.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2568, pruned_loss=0.03455, over 3042051.95 frames. ], batch size: 76, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:22:47,849 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7196, 4.5989, 4.7982, 4.9538, 5.1016, 4.6007, 5.1220, 5.1445], device='cuda:2'), covar=tensor([0.2066, 0.1174, 0.1581, 0.0718, 0.0613, 0.0843, 0.0569, 0.0647], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0764, 0.0879, 0.0777, 0.0595, 0.0614, 0.0644, 0.0747], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:22:47,930 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:22:56,580 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 02:23:45,074 INFO [optim.py:368] (2/8) 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 02:23:47,878 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252948.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:23:55,315 INFO [train.py:904] (2/8) Epoch 25, batch 9350, loss[loss=0.1801, simple_loss=0.2699, pruned_loss=0.04512, over 15416.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2565, pruned_loss=0.03443, over 3051393.48 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:24:29,928 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:24:50,097 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4686, 3.0769, 2.7034, 2.2478, 2.1727, 2.2891, 3.0760, 2.8362], device='cuda:2'), covar=tensor([0.2808, 0.0725, 0.1789, 0.3305, 0.3135, 0.2373, 0.0511, 0.1591], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0264, 0.0303, 0.0312, 0.0292, 0.0263, 0.0293, 0.0335], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 02:25:16,679 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252993.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:25:36,656 INFO [train.py:904] (2/8) Epoch 25, batch 9400, loss[loss=0.1694, simple_loss=0.2695, pruned_loss=0.0347, over 16124.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2567, pruned_loss=0.03448, over 3035920.58 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:25:42,676 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2654, 4.0918, 4.3119, 4.4474, 4.5518, 4.1660, 4.5514, 4.6000], device='cuda:2'), covar=tensor([0.1750, 0.1210, 0.1583, 0.0766, 0.0592, 0.1191, 0.0736, 0.0750], device='cuda:2'), in_proj_covar=tensor([0.0622, 0.0764, 0.0880, 0.0777, 0.0596, 0.0616, 0.0644, 0.0747], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:26:19,760 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:26:44,075 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0729, 1.5428, 1.9460, 2.0328, 2.1035, 2.3540, 1.7788, 2.2567], device='cuda:2'), covar=tensor([0.0316, 0.0569, 0.0368, 0.0450, 0.0452, 0.0291, 0.0578, 0.0176], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0192, 0.0180, 0.0182, 0.0198, 0.0157, 0.0195, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:27:02,463 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 02:27:05,785 INFO [optim.py:368] (2/8) 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 02:27:17,025 INFO [train.py:904] (2/8) Epoch 25, batch 9450, loss[loss=0.1423, simple_loss=0.2418, pruned_loss=0.02143, over 16879.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2586, pruned_loss=0.03457, over 3038031.99 frames. ], batch size: 102, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:27:19,777 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:27:57,740 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:20,139 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0870, 5.3930, 5.1831, 5.2062, 4.9202, 4.9367, 4.7441, 5.5084], device='cuda:2'), covar=tensor([0.1360, 0.1015, 0.1013, 0.0899, 0.0807, 0.0796, 0.1302, 0.1000], device='cuda:2'), in_proj_covar=tensor([0.0676, 0.0815, 0.0673, 0.0631, 0.0518, 0.0526, 0.0683, 0.0640], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:28:31,312 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:54,659 INFO [train.py:904] (2/8) Epoch 25, batch 9500, loss[loss=0.1692, simple_loss=0.2598, pruned_loss=0.03925, over 15445.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2576, pruned_loss=0.03428, over 3021659.84 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:29:33,158 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253121.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:06,151 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:26,751 INFO [optim.py:368] (2/8) 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 02:30:40,902 INFO [train.py:904] (2/8) Epoch 25, batch 9550, loss[loss=0.1673, simple_loss=0.2697, pruned_loss=0.03252, over 16691.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2578, pruned_loss=0.0342, over 3036965.87 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:30:48,739 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9829, 4.2745, 4.1462, 4.1109, 3.7870, 3.8559, 3.8622, 4.2839], device='cuda:2'), covar=tensor([0.1238, 0.0930, 0.0945, 0.0868, 0.0817, 0.1785, 0.1047, 0.0920], device='cuda:2'), in_proj_covar=tensor([0.0676, 0.0814, 0.0672, 0.0631, 0.0518, 0.0525, 0.0682, 0.0639], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:32:22,177 INFO [train.py:904] (2/8) Epoch 25, batch 9600, loss[loss=0.1744, simple_loss=0.274, pruned_loss=0.03742, over 16927.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2597, pruned_loss=0.0349, over 3048335.48 frames. ], batch size: 109, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:42,632 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253213.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:33:29,252 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 02:33:55,685 INFO [optim.py:368] (2/8) 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 02:33:59,638 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253248.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:34:07,803 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2032, 3.5327, 3.4735, 2.3931, 3.2183, 3.5774, 3.3444, 1.9943], device='cuda:2'), covar=tensor([0.0575, 0.0053, 0.0064, 0.0431, 0.0114, 0.0085, 0.0097, 0.0570], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0084, 0.0085, 0.0132, 0.0098, 0.0108, 0.0094, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 02:34:10,799 INFO [train.py:904] (2/8) Epoch 25, batch 9650, loss[loss=0.1684, simple_loss=0.2616, pruned_loss=0.03762, over 16638.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2621, pruned_loss=0.03545, over 3056716.49 frames. ], batch size: 57, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:34:52,658 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253296.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:00,315 INFO [train.py:904] (2/8) Epoch 25, batch 9700, loss[loss=0.1534, simple_loss=0.2468, pruned_loss=0.02997, over 12494.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2606, pruned_loss=0.03493, over 3054768.02 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:36:27,591 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:42,578 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:56,118 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253329.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:37:34,480 INFO [optim.py:368] (2/8) 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 02:37:37,167 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253349.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:37:44,442 INFO [train.py:904] (2/8) Epoch 25, batch 9750, loss[loss=0.1605, simple_loss=0.2549, pruned_loss=0.03306, over 16687.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.26, pruned_loss=0.03513, over 3069185.30 frames. ], batch size: 83, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:38:21,278 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:38:38,963 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253380.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:00,978 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253390.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:23,675 INFO [train.py:904] (2/8) Epoch 25, batch 9800, loss[loss=0.1397, simple_loss=0.2298, pruned_loss=0.02481, over 12247.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2593, pruned_loss=0.03426, over 3069042.71 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:39:53,002 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:40:37,888 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:40:55,716 INFO [optim.py:368] (2/8) 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 02:41:06,296 INFO [train.py:904] (2/8) Epoch 25, batch 9850, loss[loss=0.1457, simple_loss=0.2379, pruned_loss=0.02677, over 12734.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2603, pruned_loss=0.03405, over 3063413.30 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:41:16,648 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9587, 2.6200, 2.8714, 2.0524, 2.7212, 2.1354, 2.7781, 2.7922], device='cuda:2'), covar=tensor([0.0338, 0.0949, 0.0515, 0.1908, 0.0784, 0.0921, 0.0700, 0.0890], device='cuda:2'), in_proj_covar=tensor([0.0153, 0.0160, 0.0164, 0.0151, 0.0142, 0.0128, 0.0140, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 02:41:41,658 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4066, 4.6694, 4.5243, 4.5354, 4.2139, 4.2082, 4.2231, 4.7191], device='cuda:2'), covar=tensor([0.1217, 0.0940, 0.1030, 0.0883, 0.0843, 0.1580, 0.1144, 0.1017], device='cuda:2'), in_proj_covar=tensor([0.0676, 0.0815, 0.0669, 0.0630, 0.0518, 0.0524, 0.0681, 0.0637], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:42:03,943 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:00,285 INFO [train.py:904] (2/8) Epoch 25, batch 9900, loss[loss=0.1577, simple_loss=0.2479, pruned_loss=0.03375, over 12684.00 frames. ], tot_loss[loss=0.164, simple_loss=0.26, pruned_loss=0.03402, over 3049928.58 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:43:25,452 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253513.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:41,975 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:44:39,454 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9266, 2.7626, 2.9273, 2.0815, 2.7462, 2.2019, 2.7222, 2.9713], device='cuda:2'), covar=tensor([0.0338, 0.0920, 0.0473, 0.1854, 0.0794, 0.0896, 0.0733, 0.0915], device='cuda:2'), in_proj_covar=tensor([0.0152, 0.0159, 0.0163, 0.0150, 0.0141, 0.0127, 0.0139, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 02:44:46,254 INFO [optim.py:368] (2/8) 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 02:44:59,720 INFO [train.py:904] (2/8) Epoch 25, batch 9950, loss[loss=0.1551, simple_loss=0.2579, pruned_loss=0.02614, over 16919.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2622, pruned_loss=0.03459, over 3046121.59 frames. ], batch size: 96, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:45:18,703 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:45:34,749 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:46:14,339 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:47:02,499 INFO [train.py:904] (2/8) Epoch 25, batch 10000, loss[loss=0.1651, simple_loss=0.265, pruned_loss=0.0326, over 15209.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2613, pruned_loss=0.03434, over 3075243.89 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:47:38,249 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6369, 2.6805, 1.8585, 2.7913, 2.0903, 2.8292, 2.1356, 2.3708], device='cuda:2'), covar=tensor([0.0263, 0.0340, 0.1377, 0.0296, 0.0681, 0.0485, 0.1169, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0167, 0.0172, 0.0191, 0.0160, 0.0172, 0.0207, 0.0198, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 02:47:52,958 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253628.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:12,494 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3548, 4.3390, 4.6977, 4.6638, 4.6663, 4.4303, 4.4049, 4.3512], device='cuda:2'), covar=tensor([0.0319, 0.0614, 0.0405, 0.0446, 0.0464, 0.0417, 0.0792, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0454, 0.0444, 0.0405, 0.0488, 0.0466, 0.0535, 0.0372], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 02:48:36,791 INFO [optim.py:368] (2/8) 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 02:48:39,882 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:41,187 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253650.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:46,076 INFO [train.py:904] (2/8) Epoch 25, batch 10050, loss[loss=0.1684, simple_loss=0.2598, pruned_loss=0.03851, over 12291.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2612, pruned_loss=0.03403, over 3077968.73 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:48:54,365 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7813, 3.7394, 3.9115, 3.7374, 3.8514, 4.2578, 3.9345, 3.6579], device='cuda:2'), covar=tensor([0.1917, 0.2237, 0.2319, 0.2383, 0.2935, 0.1621, 0.1623, 0.2661], device='cuda:2'), in_proj_covar=tensor([0.0394, 0.0580, 0.0646, 0.0479, 0.0633, 0.0669, 0.0504, 0.0638], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 02:49:48,421 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:11,405 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253697.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:20,891 INFO [train.py:904] (2/8) Epoch 25, batch 10100, loss[loss=0.1561, simple_loss=0.2497, pruned_loss=0.03126, over 15254.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2612, pruned_loss=0.03407, over 3068396.97 frames. ], batch size: 191, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:50:29,465 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 02:50:35,685 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253711.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:47,363 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7495, 4.5459, 4.8184, 4.9376, 5.0965, 4.5695, 5.1444, 5.1266], device='cuda:2'), covar=tensor([0.1993, 0.1332, 0.1620, 0.0782, 0.0576, 0.0899, 0.0512, 0.0749], device='cuda:2'), in_proj_covar=tensor([0.0619, 0.0758, 0.0871, 0.0772, 0.0592, 0.0610, 0.0641, 0.0743], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:51:21,521 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253736.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:51:33,766 INFO [optim.py:368] (2/8) 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 02:52:07,300 INFO [train.py:904] (2/8) Epoch 26, batch 0, loss[loss=0.1582, simple_loss=0.2379, pruned_loss=0.03923, over 15806.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2379, pruned_loss=0.03923, over 15806.00 frames. ], batch size: 35, lr: 2.61e-03, grad_scale: 8.0 2023-05-02 02:52:07,300 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 02:52:14,671 INFO [train.py:938] (2/8) Epoch 26, validation: loss=0.1437, simple_loss=0.2472, pruned_loss=0.02009, over 944034.00 frames. 2023-05-02 02:52:14,672 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 02:52:42,126 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3528, 3.9621, 4.4519, 2.3326, 4.6469, 4.7293, 3.4015, 3.5946], device='cuda:2'), covar=tensor([0.0580, 0.0251, 0.0205, 0.1084, 0.0070, 0.0129, 0.0413, 0.0375], device='cuda:2'), in_proj_covar=tensor([0.0143, 0.0107, 0.0094, 0.0135, 0.0080, 0.0123, 0.0125, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 02:52:44,971 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253775.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:53:23,651 INFO [train.py:904] (2/8) Epoch 26, batch 50, loss[loss=0.1817, simple_loss=0.2756, pruned_loss=0.0439, over 16646.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2692, pruned_loss=0.0476, over 744385.83 frames. ], batch size: 62, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:53:34,917 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1112, 4.8752, 5.1213, 5.2597, 5.4888, 4.7684, 5.4236, 5.4746], device='cuda:2'), covar=tensor([0.2126, 0.1442, 0.1978, 0.0958, 0.0677, 0.0994, 0.0719, 0.0733], device='cuda:2'), in_proj_covar=tensor([0.0625, 0.0763, 0.0879, 0.0779, 0.0597, 0.0616, 0.0647, 0.0750], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:53:47,693 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 02:54:28,065 INFO [optim.py:368] (2/8) 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 02:54:30,988 INFO [train.py:904] (2/8) Epoch 26, batch 100, loss[loss=0.1563, simple_loss=0.2556, pruned_loss=0.02854, over 17063.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2649, pruned_loss=0.04529, over 1309276.26 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:55:03,715 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:55:39,319 INFO [train.py:904] (2/8) Epoch 26, batch 150, loss[loss=0.1977, simple_loss=0.2698, pruned_loss=0.06284, over 16743.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2622, pruned_loss=0.04398, over 1751624.56 frames. ], batch size: 124, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:08,342 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:56:11,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8754, 4.0773, 2.6554, 4.5950, 3.1862, 4.4900, 2.7529, 3.3697], device='cuda:2'), covar=tensor([0.0315, 0.0418, 0.1571, 0.0281, 0.0795, 0.0584, 0.1494, 0.0737], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0175, 0.0194, 0.0165, 0.0176, 0.0213, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 02:56:46,629 INFO [optim.py:368] (2/8) 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 02:56:49,068 INFO [train.py:904] (2/8) Epoch 26, batch 200, loss[loss=0.1677, simple_loss=0.265, pruned_loss=0.03522, over 17244.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2625, pruned_loss=0.04349, over 2101028.89 frames. ], batch size: 52, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:55,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2840, 5.2074, 5.1535, 4.6083, 4.7767, 5.2059, 5.1419, 4.7818], device='cuda:2'), covar=tensor([0.0611, 0.0507, 0.0350, 0.0402, 0.1208, 0.0420, 0.0323, 0.0872], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0438, 0.0342, 0.0345, 0.0344, 0.0394, 0.0235, 0.0409], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:57:33,664 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253985.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:03,910 INFO [train.py:904] (2/8) Epoch 26, batch 250, loss[loss=0.1314, simple_loss=0.2192, pruned_loss=0.02177, over 16747.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2594, pruned_loss=0.04282, over 2373663.06 frames. ], batch size: 39, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:58:07,775 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:42,509 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2440, 4.0486, 4.3003, 4.4263, 4.5354, 4.1559, 4.3718, 4.5229], device='cuda:2'), covar=tensor([0.1690, 0.1264, 0.1334, 0.0711, 0.0602, 0.1142, 0.2113, 0.0753], device='cuda:2'), in_proj_covar=tensor([0.0637, 0.0779, 0.0898, 0.0793, 0.0606, 0.0628, 0.0662, 0.0763], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 02:58:47,175 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:50,830 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:10,748 INFO [optim.py:368] (2/8) 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 02:59:14,066 INFO [train.py:904] (2/8) Epoch 26, batch 300, loss[loss=0.1551, simple_loss=0.2554, pruned_loss=0.02739, over 17018.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2574, pruned_loss=0.04187, over 2585410.15 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:59:45,482 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254075.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:57,061 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:00:23,586 INFO [train.py:904] (2/8) Epoch 26, batch 350, loss[loss=0.1535, simple_loss=0.2489, pruned_loss=0.02906, over 17123.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2559, pruned_loss=0.04114, over 2747089.64 frames. ], batch size: 47, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 03:00:24,056 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9851, 1.8564, 2.5272, 2.8817, 2.8014, 2.9904, 1.8441, 3.0821], device='cuda:2'), covar=tensor([0.0205, 0.0647, 0.0365, 0.0309, 0.0330, 0.0263, 0.0750, 0.0212], device='cuda:2'), in_proj_covar=tensor([0.0191, 0.0195, 0.0183, 0.0185, 0.0202, 0.0159, 0.0198, 0.0159], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:00:51,986 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:01:30,800 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 26, batch 400, loss[loss=0.1679, simple_loss=0.2439, pruned_loss=0.04597, over 16468.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2547, pruned_loss=0.04085, over 2878536.67 frames. ], batch size: 146, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:02:07,820 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:02:14,778 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0626, 4.0612, 3.9721, 3.6698, 3.7458, 4.0740, 3.7285, 3.8755], device='cuda:2'), covar=tensor([0.0610, 0.0723, 0.0351, 0.0319, 0.0694, 0.0462, 0.1120, 0.0596], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0446, 0.0349, 0.0352, 0.0350, 0.0403, 0.0239, 0.0416], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:02:44,706 INFO [train.py:904] (2/8) Epoch 26, batch 450, loss[loss=0.1581, simple_loss=0.2534, pruned_loss=0.0314, over 17112.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2528, pruned_loss=0.03943, over 2978021.27 frames. ], batch size: 47, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:03:12,887 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:03:15,095 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254225.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:03:50,701 INFO [optim.py:368] (2/8) 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 03:03:52,995 INFO [train.py:904] (2/8) Epoch 26, batch 500, loss[loss=0.156, simple_loss=0.249, pruned_loss=0.03145, over 16604.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2508, pruned_loss=0.03893, over 3038672.73 frames. ], batch size: 68, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:04:18,987 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254271.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:04:57,414 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6633, 2.5902, 1.9515, 2.6934, 2.1282, 2.7971, 2.1672, 2.4115], device='cuda:2'), covar=tensor([0.0340, 0.0389, 0.1353, 0.0349, 0.0714, 0.0484, 0.1227, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0179, 0.0197, 0.0169, 0.0178, 0.0217, 0.0204, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 03:05:01,741 INFO [train.py:904] (2/8) Epoch 26, batch 550, loss[loss=0.1598, simple_loss=0.2549, pruned_loss=0.03233, over 17140.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2492, pruned_loss=0.0385, over 3093258.41 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:05:05,513 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:05:46,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0319, 4.0437, 3.9737, 3.3692, 3.9764, 1.8432, 3.7641, 3.4112], device='cuda:2'), covar=tensor([0.0165, 0.0147, 0.0196, 0.0254, 0.0099, 0.2779, 0.0143, 0.0280], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0165, 0.0202, 0.0178, 0.0179, 0.0210, 0.0192, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:06:08,396 INFO [optim.py:368] (2/8) 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 03:06:11,644 INFO [train.py:904] (2/8) Epoch 26, batch 600, loss[loss=0.1683, simple_loss=0.2398, pruned_loss=0.04838, over 16901.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2489, pruned_loss=0.03904, over 3140501.21 frames. ], batch size: 96, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:06:13,036 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:07:21,730 INFO [train.py:904] (2/8) Epoch 26, batch 650, loss[loss=0.1493, simple_loss=0.2237, pruned_loss=0.03743, over 16830.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2476, pruned_loss=0.03875, over 3177732.53 frames. ], batch size: 96, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:08:28,766 INFO [optim.py:368] (2/8) 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 03:08:30,991 INFO [train.py:904] (2/8) Epoch 26, batch 700, loss[loss=0.1637, simple_loss=0.2482, pruned_loss=0.03965, over 17278.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2474, pruned_loss=0.03818, over 3211144.64 frames. ], batch size: 45, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:09:07,547 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0344, 4.9391, 4.9147, 4.5110, 4.6125, 4.9772, 4.7753, 4.6947], device='cuda:2'), covar=tensor([0.0642, 0.0887, 0.0330, 0.0355, 0.0903, 0.0540, 0.0447, 0.0670], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0456, 0.0356, 0.0360, 0.0357, 0.0412, 0.0245, 0.0427], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:09:41,471 INFO [train.py:904] (2/8) Epoch 26, batch 750, loss[loss=0.1745, simple_loss=0.2537, pruned_loss=0.04765, over 12040.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2473, pruned_loss=0.03836, over 3238961.70 frames. ], batch size: 246, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:10:08,857 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 03:10:27,536 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:10:46,992 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 26, batch 800, loss[loss=0.1402, simple_loss=0.2322, pruned_loss=0.0241, over 17139.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2466, pruned_loss=0.03805, over 3264728.92 frames. ], batch size: 47, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:11:14,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1874, 2.1261, 2.2685, 3.8924, 2.1975, 2.4672, 2.2490, 2.3097], device='cuda:2'), covar=tensor([0.1581, 0.4242, 0.3340, 0.0668, 0.4413, 0.2915, 0.4349, 0.3498], device='cuda:2'), in_proj_covar=tensor([0.0417, 0.0466, 0.0385, 0.0335, 0.0447, 0.0533, 0.0438, 0.0546], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:11:20,415 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254574.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:51,479 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:59,169 INFO [train.py:904] (2/8) Epoch 26, batch 850, loss[loss=0.1744, simple_loss=0.259, pruned_loss=0.0449, over 16700.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2456, pruned_loss=0.03724, over 3278707.07 frames. ], batch size: 89, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:12:19,106 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3736, 2.4696, 2.4143, 4.0985, 2.3642, 2.8142, 2.5268, 2.6284], device='cuda:2'), covar=tensor([0.1417, 0.3632, 0.3138, 0.0638, 0.4129, 0.2567, 0.3535, 0.3639], device='cuda:2'), in_proj_covar=tensor([0.0418, 0.0467, 0.0385, 0.0336, 0.0447, 0.0534, 0.0439, 0.0547], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:12:45,137 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:13:08,051 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 26, batch 900, loss[loss=0.1471, simple_loss=0.2302, pruned_loss=0.03203, over 16805.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2449, pruned_loss=0.03671, over 3285129.94 frames. ], batch size: 39, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:13:16,400 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0614, 4.4993, 4.4961, 3.3195, 3.7456, 4.4821, 3.9875, 3.0062], device='cuda:2'), covar=tensor([0.0463, 0.0057, 0.0046, 0.0343, 0.0152, 0.0099, 0.0100, 0.0401], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0087, 0.0088, 0.0135, 0.0101, 0.0112, 0.0097, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 03:13:39,431 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9134, 3.5650, 3.9525, 2.1480, 4.0598, 4.1189, 3.2472, 3.0701], device='cuda:2'), covar=tensor([0.0689, 0.0287, 0.0213, 0.1143, 0.0109, 0.0210, 0.0397, 0.0476], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0111, 0.0099, 0.0139, 0.0085, 0.0129, 0.0129, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 03:14:19,297 INFO [train.py:904] (2/8) Epoch 26, batch 950, loss[loss=0.1821, simple_loss=0.2744, pruned_loss=0.04491, over 17038.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2453, pruned_loss=0.03678, over 3282170.42 frames. ], batch size: 53, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:14:25,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9246, 5.0277, 5.4337, 5.4044, 5.4102, 5.0700, 4.9980, 4.8006], device='cuda:2'), covar=tensor([0.0395, 0.0547, 0.0430, 0.0403, 0.0526, 0.0446, 0.0977, 0.0515], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0481, 0.0469, 0.0430, 0.0516, 0.0493, 0.0568, 0.0393], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 03:14:40,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5733, 2.4973, 2.0065, 2.2377, 2.8947, 2.5838, 3.2605, 3.1701], device='cuda:2'), covar=tensor([0.0236, 0.0609, 0.0810, 0.0666, 0.0422, 0.0557, 0.0324, 0.0369], device='cuda:2'), in_proj_covar=tensor([0.0230, 0.0249, 0.0237, 0.0238, 0.0249, 0.0248, 0.0245, 0.0245], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:15:24,092 INFO [optim.py:368] (2/8) 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 03:15:27,107 INFO [train.py:904] (2/8) Epoch 26, batch 1000, loss[loss=0.1415, simple_loss=0.2274, pruned_loss=0.0278, over 17208.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2448, pruned_loss=0.0373, over 3282837.32 frames. ], batch size: 45, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:15:31,104 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:10,365 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 03:16:28,368 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:35,898 INFO [train.py:904] (2/8) Epoch 26, batch 1050, loss[loss=0.1498, simple_loss=0.2335, pruned_loss=0.03302, over 16247.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2442, pruned_loss=0.03701, over 3279311.79 frames. ], batch size: 36, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:16:55,066 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:42,933 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254851.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:43,590 INFO [optim.py:368] (2/8) 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 03:17:44,777 INFO [train.py:904] (2/8) Epoch 26, batch 1100, loss[loss=0.1533, simple_loss=0.2437, pruned_loss=0.0315, over 17225.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2441, pruned_loss=0.03675, over 3281691.33 frames. ], batch size: 45, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:17:51,530 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2423, 4.2577, 4.5679, 4.5523, 4.5990, 4.3037, 4.3132, 4.2491], device='cuda:2'), covar=tensor([0.0418, 0.0616, 0.0437, 0.0453, 0.0523, 0.0458, 0.0831, 0.0586], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0482, 0.0470, 0.0430, 0.0516, 0.0494, 0.0569, 0.0393], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 03:17:53,553 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:36,861 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254892.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:53,028 INFO [train.py:904] (2/8) Epoch 26, batch 1150, loss[loss=0.1415, simple_loss=0.2381, pruned_loss=0.0224, over 17199.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.244, pruned_loss=0.03672, over 3294777.62 frames. ], batch size: 46, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:19:04,330 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254912.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:29,216 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:59,610 INFO [optim.py:368] (2/8) 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 03:20:00,665 INFO [train.py:904] (2/8) Epoch 26, batch 1200, loss[loss=0.1885, simple_loss=0.2834, pruned_loss=0.04673, over 17133.00 frames. ], tot_loss[loss=0.1584, simple_loss=0.2443, pruned_loss=0.03628, over 3297322.50 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:10,674 INFO [train.py:904] (2/8) Epoch 26, batch 1250, loss[loss=0.1508, simple_loss=0.2517, pruned_loss=0.02501, over 17028.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2449, pruned_loss=0.0371, over 3298602.92 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:27,897 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4379, 2.7593, 3.0975, 2.0733, 2.8043, 2.1626, 3.0734, 3.0229], device='cuda:2'), covar=tensor([0.0301, 0.1089, 0.0576, 0.2079, 0.0904, 0.1061, 0.0704, 0.1116], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0169, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 03:22:00,646 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8141, 2.5521, 2.5276, 3.9617, 3.2480, 3.9885, 1.5993, 2.8874], device='cuda:2'), covar=tensor([0.1556, 0.0803, 0.1238, 0.0225, 0.0164, 0.0362, 0.1798, 0.0874], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0179, 0.0198, 0.0198, 0.0203, 0.0218, 0.0209, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 03:22:19,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7996, 4.8080, 4.6540, 4.0652, 4.7533, 1.8841, 4.4812, 4.2328], device='cuda:2'), covar=tensor([0.0156, 0.0143, 0.0257, 0.0442, 0.0115, 0.2906, 0.0185, 0.0342], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0171, 0.0209, 0.0184, 0.0185, 0.0217, 0.0198, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:22:19,974 INFO [optim.py:368] (2/8) 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 03:22:21,123 INFO [train.py:904] (2/8) Epoch 26, batch 1300, loss[loss=0.1712, simple_loss=0.2508, pruned_loss=0.04578, over 16892.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2449, pruned_loss=0.03721, over 3306142.98 frames. ], batch size: 109, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:22:29,862 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3561, 4.1251, 4.1273, 4.5317, 4.6415, 4.2992, 4.4834, 4.6018], device='cuda:2'), covar=tensor([0.1592, 0.1421, 0.2208, 0.0908, 0.0793, 0.1538, 0.2773, 0.1166], device='cuda:2'), in_proj_covar=tensor([0.0675, 0.0826, 0.0955, 0.0844, 0.0640, 0.0663, 0.0697, 0.0809], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:22:41,305 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 03:23:04,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4466, 3.4217, 3.4913, 3.5554, 3.6166, 3.3208, 3.5312, 3.6706], device='cuda:2'), covar=tensor([0.1372, 0.0957, 0.1122, 0.0652, 0.0650, 0.2317, 0.1244, 0.0788], device='cuda:2'), in_proj_covar=tensor([0.0674, 0.0824, 0.0954, 0.0842, 0.0638, 0.0662, 0.0695, 0.0807], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:23:30,835 INFO [train.py:904] (2/8) Epoch 26, batch 1350, loss[loss=0.1716, simple_loss=0.2693, pruned_loss=0.03696, over 17031.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2453, pruned_loss=0.03705, over 3292863.26 frames. ], batch size: 53, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:42,884 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:24:38,899 INFO [optim.py:368] (2/8) 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 03:24:40,853 INFO [train.py:904] (2/8) Epoch 26, batch 1400, loss[loss=0.1423, simple_loss=0.2284, pruned_loss=0.02817, over 16978.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2456, pruned_loss=0.0371, over 3295474.76 frames. ], batch size: 41, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:24:41,767 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:23,408 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8334, 1.9664, 2.5224, 2.8714, 2.6455, 3.4197, 2.4106, 3.4482], device='cuda:2'), covar=tensor([0.0350, 0.0631, 0.0430, 0.0408, 0.0450, 0.0218, 0.0514, 0.0180], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0200, 0.0188, 0.0190, 0.0206, 0.0164, 0.0202, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:25:34,092 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:48,426 INFO [train.py:904] (2/8) Epoch 26, batch 1450, loss[loss=0.1726, simple_loss=0.2595, pruned_loss=0.04282, over 16391.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2447, pruned_loss=0.03684, over 3294254.82 frames. ], batch size: 68, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:25:54,149 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:25,715 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8799, 1.9439, 2.4733, 2.7278, 2.7878, 2.8073, 2.1372, 2.9917], device='cuda:2'), covar=tensor([0.0204, 0.0557, 0.0393, 0.0292, 0.0339, 0.0333, 0.0595, 0.0234], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0200, 0.0188, 0.0190, 0.0206, 0.0164, 0.0202, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:26:26,866 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255230.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:40,750 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:56,773 INFO [optim.py:368] (2/8) 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 03:26:57,914 INFO [train.py:904] (2/8) Epoch 26, batch 1500, loss[loss=0.1557, simple_loss=0.2495, pruned_loss=0.03102, over 17194.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2442, pruned_loss=0.03703, over 3295048.56 frames. ], batch size: 44, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:27:31,540 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:27:36,190 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9964, 4.7572, 4.9506, 5.1670, 5.4265, 4.7044, 5.3726, 5.4087], device='cuda:2'), covar=tensor([0.1940, 0.1518, 0.2161, 0.0961, 0.0592, 0.1037, 0.0552, 0.0780], device='cuda:2'), in_proj_covar=tensor([0.0683, 0.0835, 0.0964, 0.0851, 0.0645, 0.0670, 0.0704, 0.0816], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:28:04,972 INFO [train.py:904] (2/8) Epoch 26, batch 1550, loss[loss=0.1522, simple_loss=0.2444, pruned_loss=0.02998, over 17139.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2455, pruned_loss=0.03758, over 3306514.44 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:29:12,928 INFO [optim.py:368] (2/8) 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 03:29:14,102 INFO [train.py:904] (2/8) Epoch 26, batch 1600, loss[loss=0.142, simple_loss=0.2314, pruned_loss=0.02628, over 16862.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2469, pruned_loss=0.03834, over 3313044.52 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:30:03,520 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:30:12,078 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-05-02 03:30:23,417 INFO [train.py:904] (2/8) Epoch 26, batch 1650, loss[loss=0.1718, simple_loss=0.2648, pruned_loss=0.03936, over 16663.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2485, pruned_loss=0.03847, over 3320394.21 frames. ], batch size: 57, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:30:35,813 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:09,139 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2722, 5.8045, 5.9744, 5.6283, 5.8109, 6.3214, 5.8286, 5.5013], device='cuda:2'), covar=tensor([0.0918, 0.1892, 0.2516, 0.2315, 0.2515, 0.0922, 0.1614, 0.2458], device='cuda:2'), in_proj_covar=tensor([0.0428, 0.0633, 0.0701, 0.0519, 0.0686, 0.0726, 0.0545, 0.0690], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 03:31:16,728 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1458, 4.1766, 4.5035, 4.4826, 4.5061, 4.2372, 4.2545, 4.1441], device='cuda:2'), covar=tensor([0.0424, 0.0662, 0.0400, 0.0413, 0.0569, 0.0464, 0.0839, 0.0745], device='cuda:2'), in_proj_covar=tensor([0.0436, 0.0490, 0.0476, 0.0435, 0.0523, 0.0501, 0.0577, 0.0400], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 03:31:28,970 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=255449.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:32,745 INFO [optim.py:368] (2/8) 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 03:31:34,037 INFO [train.py:904] (2/8) Epoch 26, batch 1700, loss[loss=0.1513, simple_loss=0.2498, pruned_loss=0.02642, over 17168.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2504, pruned_loss=0.03905, over 3322464.55 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:31:34,444 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:43,179 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255459.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:47,939 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 03:32:20,110 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 03:32:29,829 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7222, 4.2543, 4.2533, 2.9976, 3.6508, 4.2298, 3.8262, 2.5430], device='cuda:2'), covar=tensor([0.0530, 0.0100, 0.0052, 0.0406, 0.0145, 0.0111, 0.0109, 0.0489], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0137, 0.0102, 0.0113, 0.0098, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 03:32:40,139 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:32:42,742 INFO [train.py:904] (2/8) Epoch 26, batch 1750, loss[loss=0.1674, simple_loss=0.2726, pruned_loss=0.03111, over 17033.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2513, pruned_loss=0.03923, over 3318134.68 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:32:47,885 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255507.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:32:59,206 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9797, 5.1417, 4.9512, 4.5953, 4.2164, 5.1496, 5.0301, 4.6677], device='cuda:2'), covar=tensor([0.1029, 0.0797, 0.0527, 0.0497, 0.2135, 0.0559, 0.0371, 0.0945], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0473, 0.0369, 0.0373, 0.0372, 0.0427, 0.0254, 0.0445], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 03:33:32,816 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-05-02 03:33:49,125 INFO [optim.py:368] (2/8) 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 03:33:51,396 INFO [train.py:904] (2/8) Epoch 26, batch 1800, loss[loss=0.1819, simple_loss=0.2643, pruned_loss=0.04975, over 16733.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2523, pruned_loss=0.03876, over 3327555.51 frames. ], batch size: 134, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:33:53,374 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 03:33:54,880 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:34:35,547 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7373, 2.7337, 2.3236, 2.6917, 3.0170, 2.7847, 3.3078, 3.2198], device='cuda:2'), covar=tensor([0.0197, 0.0504, 0.0635, 0.0471, 0.0346, 0.0446, 0.0299, 0.0326], device='cuda:2'), in_proj_covar=tensor([0.0231, 0.0249, 0.0236, 0.0237, 0.0248, 0.0247, 0.0246, 0.0246], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:34:53,036 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6932, 3.2338, 3.6694, 1.9996, 3.7461, 3.7751, 3.1169, 2.8526], device='cuda:2'), covar=tensor([0.0700, 0.0291, 0.0217, 0.1185, 0.0122, 0.0222, 0.0434, 0.0459], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0113, 0.0100, 0.0141, 0.0086, 0.0132, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 03:34:59,034 INFO [train.py:904] (2/8) Epoch 26, batch 1850, loss[loss=0.1693, simple_loss=0.2649, pruned_loss=0.03688, over 16696.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2538, pruned_loss=0.03947, over 3326398.08 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:35:08,525 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 03:35:19,810 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5904, 2.5492, 2.5687, 4.5429, 2.4886, 2.9657, 2.5989, 2.7300], device='cuda:2'), covar=tensor([0.1292, 0.3694, 0.3142, 0.0481, 0.4075, 0.2453, 0.3507, 0.3469], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0470, 0.0386, 0.0338, 0.0446, 0.0536, 0.0441, 0.0548], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:36:05,021 INFO [optim.py:368] (2/8) 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 03:36:06,158 INFO [train.py:904] (2/8) Epoch 26, batch 1900, loss[loss=0.1703, simple_loss=0.2635, pruned_loss=0.03857, over 17250.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2531, pruned_loss=0.0387, over 3322352.44 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:36:32,133 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7752, 2.5513, 2.4988, 4.2059, 3.4415, 4.1251, 1.5809, 2.9877], device='cuda:2'), covar=tensor([0.1413, 0.0764, 0.1225, 0.0164, 0.0149, 0.0389, 0.1658, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0199, 0.0205, 0.0220, 0.0209, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 03:37:15,846 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7284, 2.9239, 2.8699, 5.0742, 4.0475, 4.3775, 1.7794, 3.2666], device='cuda:2'), covar=tensor([0.1396, 0.0805, 0.1171, 0.0175, 0.0214, 0.0413, 0.1610, 0.0782], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0199, 0.0206, 0.0220, 0.0210, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 03:37:16,452 INFO [train.py:904] (2/8) Epoch 26, batch 1950, loss[loss=0.148, simple_loss=0.2434, pruned_loss=0.02631, over 17193.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2523, pruned_loss=0.03785, over 3332893.27 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:37:42,562 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 03:38:12,956 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255744.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:38:24,266 INFO [optim.py:368] (2/8) 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 03:38:25,464 INFO [train.py:904] (2/8) Epoch 26, batch 2000, loss[loss=0.1442, simple_loss=0.2311, pruned_loss=0.02867, over 16773.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2516, pruned_loss=0.03772, over 3334724.60 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:39:35,298 INFO [train.py:904] (2/8) Epoch 26, batch 2050, loss[loss=0.1865, simple_loss=0.2604, pruned_loss=0.05631, over 16910.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2517, pruned_loss=0.0381, over 3334190.47 frames. ], batch size: 116, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:40:10,691 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2567, 3.2355, 2.1379, 3.4451, 2.5585, 3.4268, 2.1756, 2.6900], device='cuda:2'), covar=tensor([0.0340, 0.0485, 0.1585, 0.0402, 0.0834, 0.0893, 0.1525, 0.0811], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0184, 0.0201, 0.0176, 0.0182, 0.0223, 0.0208, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 03:40:29,605 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2602, 2.3759, 2.4116, 4.0290, 2.3091, 2.7692, 2.4683, 2.5414], device='cuda:2'), covar=tensor([0.1615, 0.3817, 0.3147, 0.0720, 0.4282, 0.2542, 0.3546, 0.3520], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0470, 0.0385, 0.0338, 0.0446, 0.0536, 0.0440, 0.0548], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:40:44,483 INFO [optim.py:368] (2/8) 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 03:40:45,698 INFO [train.py:904] (2/8) Epoch 26, batch 2100, loss[loss=0.182, simple_loss=0.2689, pruned_loss=0.04755, over 15388.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.253, pruned_loss=0.03894, over 3325617.68 frames. ], batch size: 190, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:41:54,461 INFO [train.py:904] (2/8) Epoch 26, batch 2150, loss[loss=0.1737, simple_loss=0.2698, pruned_loss=0.03879, over 17253.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2546, pruned_loss=0.03978, over 3322656.97 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:42:06,145 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 03:42:52,687 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1538, 5.1360, 5.0384, 4.5883, 4.7340, 5.0922, 4.9632, 4.7324], device='cuda:2'), covar=tensor([0.0614, 0.0611, 0.0300, 0.0365, 0.1015, 0.0483, 0.0390, 0.0728], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0474, 0.0369, 0.0374, 0.0372, 0.0427, 0.0254, 0.0446], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 03:43:04,641 INFO [optim.py:368] (2/8) 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 03:43:05,922 INFO [train.py:904] (2/8) Epoch 26, batch 2200, loss[loss=0.1616, simple_loss=0.2415, pruned_loss=0.04088, over 16806.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2552, pruned_loss=0.04031, over 3311514.69 frames. ], batch size: 96, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:43:10,483 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255956.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:44:19,344 INFO [train.py:904] (2/8) Epoch 26, batch 2250, loss[loss=0.2309, simple_loss=0.3055, pruned_loss=0.07812, over 11612.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2561, pruned_loss=0.04093, over 3301141.65 frames. ], batch size: 246, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:44:39,482 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:09,607 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256038.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:17,743 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:27,437 INFO [optim.py:368] (2/8) 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 03:45:29,154 INFO [train.py:904] (2/8) Epoch 26, batch 2300, loss[loss=0.1516, simple_loss=0.2361, pruned_loss=0.03349, over 16752.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2562, pruned_loss=0.04059, over 3309775.59 frames. ], batch size: 89, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:46:09,886 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:14,146 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:24,422 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256092.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:34,521 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:39,718 INFO [train.py:904] (2/8) Epoch 26, batch 2350, loss[loss=0.1614, simple_loss=0.2504, pruned_loss=0.03621, over 16968.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2567, pruned_loss=0.04059, over 3307705.67 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:46:44,210 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0480, 5.5040, 5.6426, 5.3157, 5.4171, 6.0412, 5.4292, 5.1257], device='cuda:2'), covar=tensor([0.1005, 0.1833, 0.2509, 0.2041, 0.2535, 0.0942, 0.1472, 0.2446], device='cuda:2'), in_proj_covar=tensor([0.0433, 0.0638, 0.0705, 0.0523, 0.0690, 0.0727, 0.0548, 0.0696], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 03:47:17,651 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7155, 2.6474, 2.2341, 2.5768, 2.9705, 2.7743, 3.2937, 3.2186], device='cuda:2'), covar=tensor([0.0189, 0.0488, 0.0605, 0.0515, 0.0341, 0.0469, 0.0274, 0.0302], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0250, 0.0237, 0.0238, 0.0250, 0.0249, 0.0248, 0.0247], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:47:19,452 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:32,027 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:37,013 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:45,351 INFO [optim.py:368] (2/8) 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 03:47:46,517 INFO [train.py:904] (2/8) Epoch 26, batch 2400, loss[loss=0.185, simple_loss=0.2627, pruned_loss=0.0536, over 16699.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2574, pruned_loss=0.04068, over 3306087.70 frames. ], batch size: 124, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:48:42,730 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256193.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 03:48:55,676 INFO [train.py:904] (2/8) Epoch 26, batch 2450, loss[loss=0.1431, simple_loss=0.2304, pruned_loss=0.02794, over 16782.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2577, pruned_loss=0.04072, over 3315444.61 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:49:33,744 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2453, 2.4230, 2.4106, 3.9879, 2.2533, 2.7840, 2.4669, 2.5326], device='cuda:2'), covar=tensor([0.1471, 0.3632, 0.3145, 0.0670, 0.4179, 0.2629, 0.3532, 0.3523], device='cuda:2'), in_proj_covar=tensor([0.0420, 0.0470, 0.0386, 0.0339, 0.0447, 0.0537, 0.0441, 0.0550], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:50:01,726 INFO [optim.py:368] (2/8) 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 03:50:03,673 INFO [train.py:904] (2/8) Epoch 26, batch 2500, loss[loss=0.1502, simple_loss=0.245, pruned_loss=0.02771, over 16823.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2572, pruned_loss=0.04044, over 3318163.48 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:20,843 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6946, 1.8960, 2.2979, 2.5297, 2.6040, 2.6694, 1.9899, 2.8391], device='cuda:2'), covar=tensor([0.0215, 0.0516, 0.0380, 0.0311, 0.0353, 0.0299, 0.0517, 0.0190], device='cuda:2'), in_proj_covar=tensor([0.0198, 0.0201, 0.0189, 0.0192, 0.0208, 0.0166, 0.0203, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:50:48,275 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256285.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:51:13,207 INFO [train.py:904] (2/8) Epoch 26, batch 2550, loss[loss=0.1865, simple_loss=0.2768, pruned_loss=0.04805, over 16417.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2568, pruned_loss=0.04023, over 3327303.08 frames. ], batch size: 68, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:51:26,211 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256312.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:14,088 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256346.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:22,558 INFO [optim.py:368] (2/8) 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 03:52:23,667 INFO [train.py:904] (2/8) Epoch 26, batch 2600, loss[loss=0.1528, simple_loss=0.2523, pruned_loss=0.02659, over 17098.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2563, pruned_loss=0.0399, over 3317804.21 frames. ], batch size: 49, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:52:51,381 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:59,151 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 03:53:21,466 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256394.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:33,341 INFO [train.py:904] (2/8) Epoch 26, batch 2650, loss[loss=0.1437, simple_loss=0.2359, pruned_loss=0.02572, over 16787.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2566, pruned_loss=0.03943, over 3322013.13 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:53:39,657 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8957, 4.6806, 4.9484, 5.1136, 5.3307, 4.6868, 5.3372, 5.3287], device='cuda:2'), covar=tensor([0.2079, 0.1534, 0.1882, 0.0863, 0.0581, 0.0915, 0.0570, 0.0632], device='cuda:2'), in_proj_covar=tensor([0.0689, 0.0848, 0.0980, 0.0860, 0.0652, 0.0682, 0.0713, 0.0825], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:53:58,425 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5394, 4.4501, 4.4361, 4.1321, 4.1830, 4.5001, 4.2235, 4.2636], device='cuda:2'), covar=tensor([0.0796, 0.0943, 0.0392, 0.0368, 0.0860, 0.0567, 0.0549, 0.0731], device='cuda:2'), in_proj_covar=tensor([0.0317, 0.0471, 0.0368, 0.0373, 0.0370, 0.0425, 0.0252, 0.0443], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 03:54:07,613 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 03:54:14,528 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256433.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:20,784 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256440.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:39,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4111, 4.3799, 4.2973, 3.6852, 4.3581, 1.6761, 4.0903, 3.8225], device='cuda:2'), covar=tensor([0.0152, 0.0144, 0.0216, 0.0309, 0.0102, 0.3217, 0.0159, 0.0287], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0173, 0.0210, 0.0186, 0.0187, 0.0218, 0.0200, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:54:40,362 INFO [optim.py:368] (2/8) 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 03:54:41,406 INFO [train.py:904] (2/8) Epoch 26, batch 2700, loss[loss=0.1893, simple_loss=0.2916, pruned_loss=0.0435, over 17261.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2575, pruned_loss=0.0392, over 3325854.01 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:55:06,491 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:55:28,932 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256488.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 03:55:48,614 INFO [train.py:904] (2/8) Epoch 26, batch 2750, loss[loss=0.1917, simple_loss=0.2883, pruned_loss=0.04749, over 17085.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2578, pruned_loss=0.03934, over 3325977.29 frames. ], batch size: 53, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:56:29,975 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256532.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:56:56,715 INFO [optim.py:368] (2/8) 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 03:56:58,713 INFO [train.py:904] (2/8) Epoch 26, batch 2800, loss[loss=0.1626, simple_loss=0.2571, pruned_loss=0.03409, over 17035.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2577, pruned_loss=0.03923, over 3331647.60 frames. ], batch size: 53, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:57:36,520 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8672, 2.8637, 2.6139, 2.7136, 3.1266, 2.9005, 3.4468, 3.3362], device='cuda:2'), covar=tensor([0.0185, 0.0463, 0.0534, 0.0505, 0.0323, 0.0471, 0.0300, 0.0340], device='cuda:2'), in_proj_covar=tensor([0.0234, 0.0250, 0.0238, 0.0238, 0.0250, 0.0248, 0.0248, 0.0248], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 03:58:07,641 INFO [train.py:904] (2/8) Epoch 26, batch 2850, loss[loss=0.1461, simple_loss=0.2335, pruned_loss=0.02931, over 17250.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2561, pruned_loss=0.03941, over 3314069.40 frames. ], batch size: 43, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:58:21,132 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256612.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:00,638 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256641.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:15,040 INFO [optim.py:368] (2/8) 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 03:59:16,872 INFO [train.py:904] (2/8) Epoch 26, batch 2900, loss[loss=0.1449, simple_loss=0.2358, pruned_loss=0.02703, over 17178.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2558, pruned_loss=0.03974, over 3312437.65 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:59:27,187 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256660.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:27,475 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5241, 3.7149, 4.1732, 2.3075, 3.2739, 2.5608, 3.9700, 3.8798], device='cuda:2'), covar=tensor([0.0265, 0.0874, 0.0444, 0.2034, 0.0784, 0.1006, 0.0571, 0.1044], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0157, 0.0147, 0.0132, 0.0146, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 03:59:35,568 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1971, 4.1809, 4.1186, 3.5892, 4.1757, 1.7873, 3.9743, 3.7044], device='cuda:2'), covar=tensor([0.0155, 0.0143, 0.0204, 0.0284, 0.0114, 0.2875, 0.0157, 0.0250], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0173, 0.0210, 0.0186, 0.0187, 0.0217, 0.0200, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:00:13,260 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 26, batch 2950, loss[loss=0.1687, simple_loss=0.251, pruned_loss=0.04321, over 16720.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2558, pruned_loss=0.04091, over 3297128.27 frames. ], batch size: 134, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:01:01,149 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256728.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:04,209 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 04:01:14,567 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:18,014 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256740.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:19,400 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9417, 2.0059, 2.6815, 2.9508, 2.7973, 3.4885, 2.2936, 3.5255], device='cuda:2'), covar=tensor([0.0340, 0.0616, 0.0364, 0.0367, 0.0398, 0.0213, 0.0577, 0.0168], device='cuda:2'), in_proj_covar=tensor([0.0198, 0.0199, 0.0189, 0.0192, 0.0206, 0.0166, 0.0203, 0.0165], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:01:20,853 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256742.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:35,075 INFO [optim.py:368] (2/8) 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 04:01:35,090 INFO [train.py:904] (2/8) Epoch 26, batch 3000, loss[loss=0.184, simple_loss=0.2651, pruned_loss=0.05144, over 16934.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2561, pruned_loss=0.04147, over 3300446.12 frames. ], batch size: 90, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:01:35,091 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 04:01:43,995 INFO [train.py:938] (2/8) Epoch 26, validation: loss=0.1339, simple_loss=0.2392, pruned_loss=0.01435, over 944034.00 frames. 2023-05-02 04:01:43,995 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 04:02:06,701 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 04:02:27,416 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,513 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,562 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 26, batch 3050, loss[loss=0.1861, simple_loss=0.2662, pruned_loss=0.05305, over 12480.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2559, pruned_loss=0.04141, over 3303353.69 frames. ], batch size: 247, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:03:27,116 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256827.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:03:27,259 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256827.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:03:31,302 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1573, 5.8240, 5.9275, 5.5853, 5.7372, 6.2871, 5.8523, 5.5177], device='cuda:2'), covar=tensor([0.1030, 0.1987, 0.2508, 0.2078, 0.2564, 0.0908, 0.1531, 0.2429], device='cuda:2'), in_proj_covar=tensor([0.0437, 0.0645, 0.0713, 0.0528, 0.0698, 0.0736, 0.0551, 0.0705], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 04:03:38,906 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:04:02,821 INFO [optim.py:368] (2/8) 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 04:04:02,837 INFO [train.py:904] (2/8) Epoch 26, batch 3100, loss[loss=0.1334, simple_loss=0.2232, pruned_loss=0.02175, over 17219.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2544, pruned_loss=0.04075, over 3306385.66 frames. ], batch size: 43, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:04:47,242 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256883.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:04:53,585 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256888.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:05:13,355 INFO [train.py:904] (2/8) Epoch 26, batch 3150, loss[loss=0.1674, simple_loss=0.258, pruned_loss=0.03841, over 16709.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2534, pruned_loss=0.04075, over 3306615.83 frames. ], batch size: 89, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:05:33,273 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-05-02 04:06:05,226 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256941.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:06:09,402 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256944.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:06:21,403 INFO [optim.py:368] (2/8) 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 04:06:21,418 INFO [train.py:904] (2/8) Epoch 26, batch 3200, loss[loss=0.147, simple_loss=0.2412, pruned_loss=0.02644, over 17034.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2528, pruned_loss=0.04027, over 3313477.87 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:07:11,954 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:07:29,905 INFO [train.py:904] (2/8) Epoch 26, batch 3250, loss[loss=0.1385, simple_loss=0.225, pruned_loss=0.02604, over 16981.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2535, pruned_loss=0.0409, over 3311451.45 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:07:43,679 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:03,382 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:38,566 INFO [optim.py:368] (2/8) 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 04:08:38,588 INFO [train.py:904] (2/8) Epoch 26, batch 3300, loss[loss=0.1834, simple_loss=0.2709, pruned_loss=0.04797, over 15370.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2553, pruned_loss=0.04128, over 3314655.80 frames. ], batch size: 190, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:09:07,498 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257074.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:09,485 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257076.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:48,084 INFO [train.py:904] (2/8) Epoch 26, batch 3350, loss[loss=0.1898, simple_loss=0.2616, pruned_loss=0.05901, over 16903.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2555, pruned_loss=0.04087, over 3316022.43 frames. ], batch size: 109, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:09:56,293 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0297, 2.5835, 2.1319, 2.3339, 2.9034, 2.6896, 2.9435, 3.0366], device='cuda:2'), covar=tensor([0.0280, 0.0473, 0.0586, 0.0511, 0.0315, 0.0420, 0.0285, 0.0351], device='cuda:2'), in_proj_covar=tensor([0.0236, 0.0250, 0.0238, 0.0239, 0.0251, 0.0250, 0.0250, 0.0249], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:10:20,876 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 04:10:56,591 INFO [train.py:904] (2/8) Epoch 26, batch 3400, loss[loss=0.1772, simple_loss=0.2599, pruned_loss=0.0473, over 16754.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2553, pruned_loss=0.04056, over 3313655.15 frames. ], batch size: 124, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:11:13,561 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257165.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:27,518 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257175.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:39,234 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257183.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:12:06,426 INFO [train.py:904] (2/8) Epoch 26, batch 3450, loss[loss=0.141, simple_loss=0.224, pruned_loss=0.02896, over 16961.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2533, pruned_loss=0.0399, over 3319227.45 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:12:33,474 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:39,826 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:57,435 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257239.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:12:59,791 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:13:16,457 INFO [optim.py:368] (2/8) 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 04:13:16,472 INFO [train.py:904] (2/8) Epoch 26, batch 3500, loss[loss=0.1699, simple_loss=0.2694, pruned_loss=0.03518, over 17030.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2521, pruned_loss=0.03967, over 3320676.36 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:13:36,900 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2479, 5.2095, 5.0174, 4.4495, 5.0707, 1.6701, 4.8098, 4.8504], device='cuda:2'), covar=tensor([0.0096, 0.0096, 0.0256, 0.0455, 0.0115, 0.3262, 0.0167, 0.0263], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0188, 0.0189, 0.0219, 0.0202, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:13:58,989 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:25,652 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:26,372 INFO [train.py:904] (2/8) Epoch 26, batch 3550, loss[loss=0.1489, simple_loss=0.2322, pruned_loss=0.03282, over 16936.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2519, pruned_loss=0.03922, over 3321295.75 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:15:02,432 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 04:15:34,857 INFO [train.py:904] (2/8) Epoch 26, batch 3600, loss[loss=0.1551, simple_loss=0.2348, pruned_loss=0.0377, over 16697.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2509, pruned_loss=0.03858, over 3325307.57 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:15:35,973 INFO [optim.py:368] (2/8) 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 04:15:58,479 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257369.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:16:47,819 INFO [train.py:904] (2/8) Epoch 26, batch 3650, loss[loss=0.1578, simple_loss=0.2305, pruned_loss=0.04257, over 16465.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2489, pruned_loss=0.03876, over 3320414.63 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:24,407 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2214, 2.2754, 2.4728, 4.0049, 2.3206, 2.6418, 2.3750, 2.4593], device='cuda:2'), covar=tensor([0.1574, 0.3814, 0.3250, 0.0640, 0.3984, 0.2666, 0.3926, 0.3330], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0471, 0.0386, 0.0340, 0.0447, 0.0539, 0.0442, 0.0551], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:17:25,522 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7815, 3.5332, 3.8975, 2.1332, 3.9865, 4.0118, 3.2629, 2.9625], device='cuda:2'), covar=tensor([0.0733, 0.0288, 0.0189, 0.1196, 0.0102, 0.0205, 0.0372, 0.0480], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0112, 0.0100, 0.0139, 0.0086, 0.0131, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 04:17:58,062 INFO [train.py:904] (2/8) Epoch 26, batch 3700, loss[loss=0.1771, simple_loss=0.2559, pruned_loss=0.04912, over 11665.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2483, pruned_loss=0.04007, over 3301586.47 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:59,893 INFO [optim.py:368] (2/8) 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 04:18:18,521 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:18:41,397 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257483.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:19:09,128 INFO [train.py:904] (2/8) Epoch 26, batch 3750, loss[loss=0.1825, simple_loss=0.2761, pruned_loss=0.04443, over 16645.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2488, pruned_loss=0.04133, over 3280569.20 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:19:33,996 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:19:44,869 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257528.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:19:48,975 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257531.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:20:01,274 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257539.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:20:20,023 INFO [train.py:904] (2/8) Epoch 26, batch 3800, loss[loss=0.1847, simple_loss=0.2552, pruned_loss=0.05708, over 16883.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2498, pruned_loss=0.04259, over 3283951.69 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:20:22,177 INFO [optim.py:368] (2/8) 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 04:20:55,722 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:20:55,803 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257578.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:21:09,654 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257587.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:21:22,613 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:21:30,542 INFO [train.py:904] (2/8) Epoch 26, batch 3850, loss[loss=0.1436, simple_loss=0.2247, pruned_loss=0.0312, over 16699.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2499, pruned_loss=0.04321, over 3285799.64 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:21:49,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8534, 1.4300, 1.7520, 1.7300, 1.8536, 2.0083, 1.6472, 1.8736], device='cuda:2'), covar=tensor([0.0268, 0.0469, 0.0247, 0.0360, 0.0303, 0.0192, 0.0492, 0.0140], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0199, 0.0189, 0.0192, 0.0207, 0.0166, 0.0204, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:22:20,347 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257639.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:22:36,040 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5510, 5.5307, 5.2914, 4.6960, 5.5234, 2.1237, 5.2347, 4.8483], device='cuda:2'), covar=tensor([0.0053, 0.0046, 0.0176, 0.0309, 0.0055, 0.2879, 0.0098, 0.0262], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0173, 0.0211, 0.0186, 0.0187, 0.0216, 0.0200, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:22:41,228 INFO [train.py:904] (2/8) Epoch 26, batch 3900, loss[loss=0.181, simple_loss=0.2592, pruned_loss=0.05142, over 16264.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.25, pruned_loss=0.04395, over 3285261.25 frames. ], batch size: 165, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:42,469 INFO [optim.py:368] (2/8) 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 04:23:04,599 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:23:29,612 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1370, 2.1837, 1.7821, 1.8836, 2.3483, 2.1061, 2.1445, 2.4526], device='cuda:2'), covar=tensor([0.0315, 0.0371, 0.0550, 0.0467, 0.0240, 0.0359, 0.0212, 0.0283], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0246, 0.0233, 0.0235, 0.0246, 0.0245, 0.0246, 0.0245], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:23:51,422 INFO [train.py:904] (2/8) Epoch 26, batch 3950, loss[loss=0.1488, simple_loss=0.2213, pruned_loss=0.03819, over 16435.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2496, pruned_loss=0.04422, over 3284603.84 frames. ], batch size: 75, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:24:12,553 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:24:36,778 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3169, 5.3635, 5.6980, 5.6510, 5.7560, 5.4144, 5.3084, 5.1046], device='cuda:2'), covar=tensor([0.0299, 0.0494, 0.0373, 0.0416, 0.0410, 0.0339, 0.0933, 0.0470], device='cuda:2'), in_proj_covar=tensor([0.0442, 0.0498, 0.0482, 0.0442, 0.0531, 0.0508, 0.0588, 0.0407], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-02 04:25:02,901 INFO [train.py:904] (2/8) Epoch 26, batch 4000, loss[loss=0.1781, simple_loss=0.2706, pruned_loss=0.04279, over 15519.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2495, pruned_loss=0.04476, over 3284621.09 frames. ], batch size: 191, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:25:03,988 INFO [optim.py:368] (2/8) 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 04:25:41,163 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1884, 3.4053, 3.6000, 2.0538, 3.0840, 2.3649, 3.6639, 3.6953], device='cuda:2'), covar=tensor([0.0190, 0.0779, 0.0570, 0.2189, 0.0851, 0.0940, 0.0507, 0.0846], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 04:25:44,540 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6954, 6.0639, 5.5931, 6.0286, 5.5292, 5.1271, 5.6776, 6.1845], device='cuda:2'), covar=tensor([0.2272, 0.1181, 0.2151, 0.1232, 0.1342, 0.1168, 0.2011, 0.1362], device='cuda:2'), in_proj_covar=tensor([0.0721, 0.0875, 0.0715, 0.0677, 0.0556, 0.0555, 0.0736, 0.0684], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:26:13,221 INFO [train.py:904] (2/8) Epoch 26, batch 4050, loss[loss=0.1679, simple_loss=0.2524, pruned_loss=0.04169, over 16871.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2502, pruned_loss=0.04392, over 3286647.98 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:26:40,023 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:26:42,417 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257823.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:26:53,127 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 04:27:25,363 INFO [train.py:904] (2/8) Epoch 26, batch 4100, loss[loss=0.167, simple_loss=0.2524, pruned_loss=0.04077, over 16531.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.252, pruned_loss=0.04356, over 3281871.05 frames. ], batch size: 75, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:27:26,539 INFO [optim.py:368] (2/8) 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 04:27:35,029 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257859.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:27:49,110 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257869.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:27:51,009 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4034, 5.6946, 5.4464, 5.4601, 5.1486, 5.0720, 5.0978, 5.8329], device='cuda:2'), covar=tensor([0.1216, 0.0799, 0.1200, 0.0861, 0.0827, 0.0721, 0.1187, 0.0733], device='cuda:2'), in_proj_covar=tensor([0.0720, 0.0875, 0.0714, 0.0676, 0.0556, 0.0555, 0.0736, 0.0683], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:28:02,794 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:10,475 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257883.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:30,684 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:39,711 INFO [train.py:904] (2/8) Epoch 26, batch 4150, loss[loss=0.1703, simple_loss=0.2695, pruned_loss=0.03552, over 16829.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2587, pruned_loss=0.0457, over 3231759.78 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:07,225 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:28,831 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257934.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:29:44,224 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257944.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:29:45,298 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:52,799 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:56,806 INFO [train.py:904] (2/8) Epoch 26, batch 4200, loss[loss=0.21, simple_loss=0.3026, pruned_loss=0.0587, over 16549.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2656, pruned_loss=0.0471, over 3202132.26 frames. ], batch size: 35, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:58,475 INFO [optim.py:368] (2/8) 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 04:31:00,222 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4513, 3.0328, 2.6074, 2.3293, 2.2478, 2.1212, 2.9960, 2.7873], device='cuda:2'), covar=tensor([0.2588, 0.0793, 0.1838, 0.2667, 0.2919, 0.2542, 0.0691, 0.1513], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0273, 0.0308, 0.0320, 0.0304, 0.0270, 0.0300, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 04:31:03,069 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0287, 2.3851, 2.3613, 2.6809, 1.9574, 3.2093, 1.7977, 2.7751], device='cuda:2'), covar=tensor([0.1160, 0.0696, 0.1174, 0.0171, 0.0114, 0.0467, 0.1592, 0.0760], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0181, 0.0199, 0.0200, 0.0207, 0.0219, 0.0208, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 04:31:15,317 INFO [train.py:904] (2/8) Epoch 26, batch 4250, loss[loss=0.1969, simple_loss=0.289, pruned_loss=0.05246, over 17052.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.269, pruned_loss=0.04702, over 3177067.59 frames. ], batch size: 55, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:31:29,172 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:32:29,606 INFO [train.py:904] (2/8) Epoch 26, batch 4300, loss[loss=0.1886, simple_loss=0.285, pruned_loss=0.04611, over 16493.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2706, pruned_loss=0.0463, over 3190659.32 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:32:31,421 INFO [optim.py:368] (2/8) 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 04:33:15,606 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5764, 5.5710, 5.3512, 4.7741, 5.5855, 2.1500, 5.2816, 5.0764], device='cuda:2'), covar=tensor([0.0053, 0.0048, 0.0152, 0.0339, 0.0055, 0.2863, 0.0094, 0.0203], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0173, 0.0210, 0.0186, 0.0187, 0.0217, 0.0200, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:33:45,780 INFO [train.py:904] (2/8) Epoch 26, batch 4350, loss[loss=0.1985, simple_loss=0.2874, pruned_loss=0.05485, over 17230.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2741, pruned_loss=0.04726, over 3193539.74 frames. ], batch size: 44, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:33:51,083 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 04:34:15,547 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:34:46,338 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9711, 2.2147, 2.3303, 3.4039, 2.0824, 2.4727, 2.3270, 2.3235], device='cuda:2'), covar=tensor([0.1469, 0.3217, 0.2852, 0.0691, 0.4033, 0.2335, 0.3132, 0.3277], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0470, 0.0383, 0.0337, 0.0445, 0.0539, 0.0441, 0.0550], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:34:58,989 INFO [train.py:904] (2/8) Epoch 26, batch 4400, loss[loss=0.1781, simple_loss=0.2679, pruned_loss=0.04414, over 16551.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2759, pruned_loss=0.04813, over 3195401.19 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:35:00,098 INFO [optim.py:368] (2/8) 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 04:35:25,196 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258171.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:35:59,764 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0968, 3.4388, 3.5736, 2.0056, 3.0658, 2.1311, 3.4644, 3.6066], device='cuda:2'), covar=tensor([0.0231, 0.0749, 0.0531, 0.2203, 0.0788, 0.1062, 0.0586, 0.0909], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0156, 0.0148, 0.0132, 0.0146, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 04:36:11,040 INFO [train.py:904] (2/8) Epoch 26, batch 4450, loss[loss=0.1983, simple_loss=0.2895, pruned_loss=0.05351, over 16491.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2797, pruned_loss=0.04957, over 3222520.66 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:36:28,144 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:36:55,328 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258234.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:37:03,325 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258239.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:37:22,666 INFO [train.py:904] (2/8) Epoch 26, batch 4500, loss[loss=0.1823, simple_loss=0.2699, pruned_loss=0.04728, over 16965.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.28, pruned_loss=0.0504, over 3220017.38 frames. ], batch size: 41, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:37:23,848 INFO [optim.py:368] (2/8) 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 04:37:27,059 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3654, 3.3846, 2.1626, 3.8884, 2.5892, 3.8715, 2.3000, 2.7529], device='cuda:2'), covar=tensor([0.0325, 0.0371, 0.1602, 0.0149, 0.0817, 0.0398, 0.1488, 0.0795], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0179, 0.0195, 0.0171, 0.0178, 0.0218, 0.0203, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 04:37:33,876 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-02 04:37:54,769 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 04:37:57,336 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 04:38:05,641 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258282.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:38:35,271 INFO [train.py:904] (2/8) Epoch 26, batch 4550, loss[loss=0.2089, simple_loss=0.2973, pruned_loss=0.06026, over 16748.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2805, pruned_loss=0.05106, over 3224437.19 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:38:39,798 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:39:48,615 INFO [train.py:904] (2/8) Epoch 26, batch 4600, loss[loss=0.1742, simple_loss=0.2673, pruned_loss=0.04055, over 16723.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2814, pruned_loss=0.05151, over 3236012.18 frames. ], batch size: 89, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:39:50,256 INFO [optim.py:368] (2/8) 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 04:39:52,480 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5320, 5.4674, 5.3428, 4.9664, 5.0673, 5.3905, 5.2024, 5.0448], device='cuda:2'), covar=tensor([0.0440, 0.0285, 0.0216, 0.0266, 0.0780, 0.0269, 0.0295, 0.0563], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0456, 0.0358, 0.0361, 0.0359, 0.0413, 0.0245, 0.0431], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:40:44,246 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1115, 3.7228, 3.6453, 2.4235, 3.3394, 3.7406, 3.3555, 2.2378], device='cuda:2'), covar=tensor([0.0638, 0.0052, 0.0063, 0.0454, 0.0113, 0.0101, 0.0119, 0.0462], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0088, 0.0090, 0.0136, 0.0102, 0.0114, 0.0098, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 04:40:45,376 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6013, 3.7128, 2.6286, 2.2500, 2.3953, 2.3488, 3.9323, 3.2212], device='cuda:2'), covar=tensor([0.3033, 0.0632, 0.2116, 0.2813, 0.2760, 0.2312, 0.0513, 0.1414], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0274, 0.0311, 0.0322, 0.0305, 0.0271, 0.0302, 0.0349], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 04:41:03,105 INFO [train.py:904] (2/8) Epoch 26, batch 4650, loss[loss=0.1714, simple_loss=0.2568, pruned_loss=0.043, over 16589.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2807, pruned_loss=0.05183, over 3226729.70 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:41:13,427 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 04:41:39,147 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 04:42:14,211 INFO [train.py:904] (2/8) Epoch 26, batch 4700, loss[loss=0.1787, simple_loss=0.2585, pruned_loss=0.04941, over 16614.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2784, pruned_loss=0.0511, over 3232253.03 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:42:16,007 INFO [optim.py:368] (2/8) 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 04:42:16,458 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:26,791 INFO [train.py:904] (2/8) Epoch 26, batch 4750, loss[loss=0.1742, simple_loss=0.2583, pruned_loss=0.0451, over 16842.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2742, pruned_loss=0.04899, over 3229699.55 frames. ], batch size: 39, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:43:38,879 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:44,991 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,082 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:21,153 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258539.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:31,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8922, 2.8218, 2.7021, 2.0032, 2.5750, 2.7709, 2.6185, 1.9224], device='cuda:2'), covar=tensor([0.0498, 0.0087, 0.0085, 0.0374, 0.0142, 0.0135, 0.0141, 0.0455], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0088, 0.0090, 0.0136, 0.0102, 0.0114, 0.0098, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 04:44:41,653 INFO [train.py:904] (2/8) Epoch 26, batch 4800, loss[loss=0.1573, simple_loss=0.2502, pruned_loss=0.03218, over 17255.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.271, pruned_loss=0.04713, over 3222771.93 frames. ], batch size: 52, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:44:43,333 INFO [optim.py:368] (2/8) 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 04:44:56,481 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258563.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:11,433 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258572.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:34,303 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258587.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:57,759 INFO [train.py:904] (2/8) Epoch 26, batch 4850, loss[loss=0.1924, simple_loss=0.2769, pruned_loss=0.054, over 11752.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2718, pruned_loss=0.04641, over 3199223.65 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:46:03,270 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258606.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:46:11,192 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 04:47:14,843 INFO [train.py:904] (2/8) Epoch 26, batch 4900, loss[loss=0.1959, simple_loss=0.2895, pruned_loss=0.0512, over 16727.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2706, pruned_loss=0.04508, over 3193094.73 frames. ], batch size: 134, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:47:16,710 INFO [optim.py:368] (2/8) 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 04:47:17,072 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:48:29,884 INFO [train.py:904] (2/8) Epoch 26, batch 4950, loss[loss=0.1627, simple_loss=0.2502, pruned_loss=0.03763, over 16960.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2699, pruned_loss=0.04439, over 3203049.68 frames. ], batch size: 41, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:49:41,062 INFO [train.py:904] (2/8) Epoch 26, batch 5000, loss[loss=0.1875, simple_loss=0.2861, pruned_loss=0.04446, over 16717.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2719, pruned_loss=0.04454, over 3207200.44 frames. ], batch size: 134, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:49:42,189 INFO [optim.py:368] (2/8) 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 04:50:54,498 INFO [train.py:904] (2/8) Epoch 26, batch 5050, loss[loss=0.1782, simple_loss=0.2814, pruned_loss=0.03752, over 16811.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2734, pruned_loss=0.04478, over 3202237.89 frames. ], batch size: 102, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:51:05,003 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:51:09,364 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-05-02 04:51:59,332 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4459, 3.4285, 3.4769, 3.5490, 3.6093, 3.3532, 3.5935, 3.6657], device='cuda:2'), covar=tensor([0.1250, 0.0896, 0.1025, 0.0616, 0.0591, 0.2129, 0.0893, 0.0716], device='cuda:2'), in_proj_covar=tensor([0.0659, 0.0808, 0.0932, 0.0817, 0.0620, 0.0648, 0.0675, 0.0783], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:52:07,189 INFO [train.py:904] (2/8) Epoch 26, batch 5100, loss[loss=0.1602, simple_loss=0.2544, pruned_loss=0.03298, over 15394.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2719, pruned_loss=0.04421, over 3202614.47 frames. ], batch size: 191, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:52:08,929 INFO [optim.py:368] (2/8) 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 04:52:27,866 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:52:48,206 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9849, 5.3696, 5.5425, 5.2437, 5.3603, 5.9120, 5.3757, 5.0380], device='cuda:2'), covar=tensor([0.0849, 0.1646, 0.1781, 0.1867, 0.2226, 0.0712, 0.1370, 0.2327], device='cuda:2'), in_proj_covar=tensor([0.0421, 0.0617, 0.0677, 0.0505, 0.0667, 0.0705, 0.0529, 0.0674], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 04:53:20,199 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8905, 5.2779, 5.4445, 5.1532, 5.2251, 5.8160, 5.2410, 4.8644], device='cuda:2'), covar=tensor([0.0940, 0.1621, 0.1832, 0.1867, 0.2233, 0.0763, 0.1475, 0.2513], device='cuda:2'), in_proj_covar=tensor([0.0421, 0.0617, 0.0677, 0.0505, 0.0667, 0.0705, 0.0529, 0.0674], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 04:53:21,028 INFO [train.py:904] (2/8) Epoch 26, batch 5150, loss[loss=0.1585, simple_loss=0.2555, pruned_loss=0.03069, over 16720.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2705, pruned_loss=0.04317, over 3195897.46 frames. ], batch size: 76, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:35,100 INFO [train.py:904] (2/8) Epoch 26, batch 5200, loss[loss=0.1654, simple_loss=0.2632, pruned_loss=0.03378, over 16299.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2687, pruned_loss=0.04237, over 3193621.22 frames. ], batch size: 165, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:36,751 INFO [optim.py:368] (2/8) 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 04:54:53,515 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:55:03,415 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2023-05-02 04:55:48,686 INFO [train.py:904] (2/8) Epoch 26, batch 5250, loss[loss=0.1603, simple_loss=0.2543, pruned_loss=0.03309, over 16705.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2661, pruned_loss=0.04207, over 3198975.35 frames. ], batch size: 76, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:55:55,006 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 04:56:22,997 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259026.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:56:51,343 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2591, 3.2159, 3.5504, 1.7670, 3.6549, 3.6957, 2.9130, 2.7367], device='cuda:2'), covar=tensor([0.0887, 0.0272, 0.0192, 0.1305, 0.0086, 0.0163, 0.0455, 0.0539], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0111, 0.0099, 0.0138, 0.0086, 0.0129, 0.0129, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 04:57:03,176 INFO [train.py:904] (2/8) Epoch 26, batch 5300, loss[loss=0.1578, simple_loss=0.2512, pruned_loss=0.03215, over 15371.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2629, pruned_loss=0.04111, over 3191597.48 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:57:04,410 INFO [optim.py:368] (2/8) 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 04:58:18,026 INFO [train.py:904] (2/8) Epoch 26, batch 5350, loss[loss=0.1881, simple_loss=0.2725, pruned_loss=0.05178, over 12393.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2613, pruned_loss=0.04051, over 3189463.27 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:58:27,866 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:04,378 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5374, 5.7945, 5.5116, 5.6614, 5.3158, 5.2363, 5.2141, 5.9283], device='cuda:2'), covar=tensor([0.1232, 0.0821, 0.1100, 0.0739, 0.0765, 0.0628, 0.1162, 0.0812], device='cuda:2'), in_proj_covar=tensor([0.0697, 0.0844, 0.0692, 0.0650, 0.0537, 0.0534, 0.0711, 0.0660], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 04:59:31,424 INFO [train.py:904] (2/8) Epoch 26, batch 5400, loss[loss=0.2001, simple_loss=0.2844, pruned_loss=0.0579, over 11992.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2637, pruned_loss=0.04121, over 3186875.17 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:59:32,572 INFO [optim.py:368] (2/8) 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 04:59:38,343 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:52,966 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:00:36,624 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7207, 4.5894, 4.7826, 4.9497, 5.1121, 4.6031, 5.1271, 5.1023], device='cuda:2'), covar=tensor([0.1897, 0.1277, 0.1724, 0.0757, 0.0620, 0.0929, 0.0621, 0.0700], device='cuda:2'), in_proj_covar=tensor([0.0664, 0.0816, 0.0942, 0.0824, 0.0627, 0.0655, 0.0680, 0.0792], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:00:38,320 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 05:00:48,642 INFO [train.py:904] (2/8) Epoch 26, batch 5450, loss[loss=0.2241, simple_loss=0.3197, pruned_loss=0.06426, over 15466.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2672, pruned_loss=0.04275, over 3192896.92 frames. ], batch size: 191, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:01:07,879 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:01:25,029 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:05,202 INFO [train.py:904] (2/8) Epoch 26, batch 5500, loss[loss=0.2049, simple_loss=0.2994, pruned_loss=0.05517, over 16909.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2743, pruned_loss=0.04716, over 3159732.83 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:02:07,102 INFO [optim.py:368] (2/8) 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 05:02:14,049 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7050, 3.2116, 3.2393, 1.9422, 2.8756, 2.1916, 3.2616, 3.4470], device='cuda:2'), covar=tensor([0.0278, 0.0714, 0.0599, 0.2078, 0.0857, 0.0960, 0.0625, 0.0780], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0170, 0.0171, 0.0157, 0.0149, 0.0132, 0.0147, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 05:02:23,188 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259264.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:58,129 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:22,301 INFO [train.py:904] (2/8) Epoch 26, batch 5550, loss[loss=0.1873, simple_loss=0.2826, pruned_loss=0.04604, over 16430.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2811, pruned_loss=0.05152, over 3153763.37 frames. ], batch size: 75, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:03:50,534 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:57,627 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259325.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:04:02,374 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 05:04:26,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4705, 3.0231, 2.7666, 2.3583, 2.3621, 2.3651, 3.0196, 2.9821], device='cuda:2'), covar=tensor([0.2073, 0.0571, 0.1308, 0.2320, 0.2093, 0.1935, 0.0484, 0.1101], device='cuda:2'), in_proj_covar=tensor([0.0330, 0.0272, 0.0309, 0.0320, 0.0302, 0.0270, 0.0300, 0.0346], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 05:04:40,412 INFO [train.py:904] (2/8) Epoch 26, batch 5600, loss[loss=0.2715, simple_loss=0.3309, pruned_loss=0.106, over 11022.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2865, pruned_loss=0.05628, over 3095849.09 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 16.0 2023-05-02 05:04:41,808 INFO [optim.py:368] (2/8) 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 05:06:04,635 INFO [train.py:904] (2/8) Epoch 26, batch 5650, loss[loss=0.2129, simple_loss=0.2942, pruned_loss=0.06583, over 15324.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2912, pruned_loss=0.0595, over 3083834.21 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:06:15,023 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1571, 2.0660, 1.6757, 1.7303, 2.2913, 1.9584, 1.8960, 2.3692], device='cuda:2'), covar=tensor([0.0268, 0.0413, 0.0568, 0.0522, 0.0285, 0.0416, 0.0204, 0.0277], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0240, 0.0230, 0.0231, 0.0240, 0.0240, 0.0239, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:06:22,307 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259414.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:07:22,196 INFO [train.py:904] (2/8) Epoch 26, batch 5700, loss[loss=0.2074, simple_loss=0.3049, pruned_loss=0.05493, over 16717.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2919, pruned_loss=0.06051, over 3082750.15 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:07:25,089 INFO [optim.py:368] (2/8) 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 05:07:56,940 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259475.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:08:04,016 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5907, 2.5499, 2.4840, 3.8181, 2.9428, 3.6878, 1.2874, 2.9590], device='cuda:2'), covar=tensor([0.1465, 0.0827, 0.1304, 0.0195, 0.0245, 0.0401, 0.1899, 0.0757], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0199, 0.0208, 0.0218, 0.0208, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 05:08:39,969 INFO [train.py:904] (2/8) Epoch 26, batch 5750, loss[loss=0.1955, simple_loss=0.2826, pruned_loss=0.05417, over 16179.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2937, pruned_loss=0.06124, over 3080212.05 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:08:54,442 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7452, 3.7968, 3.9331, 3.6975, 3.8685, 4.2234, 3.8859, 3.6415], device='cuda:2'), covar=tensor([0.2335, 0.2340, 0.2321, 0.2543, 0.2513, 0.1817, 0.1754, 0.2725], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0620, 0.0682, 0.0511, 0.0671, 0.0710, 0.0532, 0.0678], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 05:09:40,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8579, 5.1947, 5.4437, 5.1039, 5.2989, 5.7960, 5.2834, 4.9774], device='cuda:2'), covar=tensor([0.1140, 0.2145, 0.2704, 0.2073, 0.2165, 0.0906, 0.1499, 0.2364], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0621, 0.0683, 0.0511, 0.0672, 0.0710, 0.0533, 0.0679], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 05:10:02,530 INFO [train.py:904] (2/8) Epoch 26, batch 5800, loss[loss=0.1973, simple_loss=0.2903, pruned_loss=0.05218, over 15346.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2935, pruned_loss=0.06048, over 3062904.52 frames. ], batch size: 191, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:05,673 INFO [optim.py:368] (2/8) 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 05:10:11,510 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-02 05:10:47,017 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:10:58,830 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8896, 4.6379, 4.5459, 5.0573, 5.2496, 4.7471, 5.1492, 5.2130], device='cuda:2'), covar=tensor([0.1903, 0.1531, 0.2864, 0.0963, 0.0802, 0.1265, 0.1044, 0.1041], device='cuda:2'), in_proj_covar=tensor([0.0665, 0.0815, 0.0941, 0.0824, 0.0629, 0.0655, 0.0681, 0.0794], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:11:19,029 INFO [train.py:904] (2/8) Epoch 26, batch 5850, loss[loss=0.1809, simple_loss=0.2786, pruned_loss=0.04155, over 16910.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2909, pruned_loss=0.0587, over 3066817.28 frames. ], batch size: 96, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:11:44,961 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259620.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:11:46,396 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:11:50,815 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-05-02 05:12:39,963 INFO [train.py:904] (2/8) Epoch 26, batch 5900, loss[loss=0.1813, simple_loss=0.2708, pruned_loss=0.04586, over 16657.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2901, pruned_loss=0.0584, over 3080692.69 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:12:43,693 INFO [optim.py:368] (2/8) 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 05:13:00,111 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8522, 1.8827, 2.3385, 2.7701, 2.6293, 3.1444, 1.9944, 3.1730], device='cuda:2'), covar=tensor([0.0226, 0.0576, 0.0408, 0.0324, 0.0364, 0.0187, 0.0654, 0.0162], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0197, 0.0185, 0.0190, 0.0205, 0.0163, 0.0202, 0.0163], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:13:08,193 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:14:01,045 INFO [train.py:904] (2/8) Epoch 26, batch 5950, loss[loss=0.2231, simple_loss=0.3056, pruned_loss=0.07029, over 11843.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2913, pruned_loss=0.05759, over 3089178.04 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:15:12,636 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259749.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:15:12,654 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6877, 1.7750, 1.5014, 1.4153, 1.8368, 1.5550, 1.6000, 1.9110], device='cuda:2'), covar=tensor([0.0247, 0.0376, 0.0531, 0.0438, 0.0266, 0.0346, 0.0210, 0.0268], device='cuda:2'), in_proj_covar=tensor([0.0224, 0.0239, 0.0229, 0.0230, 0.0239, 0.0238, 0.0238, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:15:18,035 INFO [train.py:904] (2/8) Epoch 26, batch 6000, loss[loss=0.1702, simple_loss=0.2591, pruned_loss=0.04067, over 16896.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2899, pruned_loss=0.05662, over 3107916.29 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:15:18,035 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 05:15:28,176 INFO [train.py:938] (2/8) Epoch 26, validation: loss=0.1485, simple_loss=0.2607, pruned_loss=0.01818, over 944034.00 frames. 2023-05-02 05:15:28,176 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 05:15:30,551 INFO [optim.py:368] (2/8) 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 05:15:31,534 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6504, 4.8384, 5.0034, 4.7750, 4.8238, 5.3677, 4.8295, 4.5481], device='cuda:2'), covar=tensor([0.1240, 0.1857, 0.2413, 0.1828, 0.2295, 0.0917, 0.1627, 0.2428], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0622, 0.0684, 0.0510, 0.0673, 0.0711, 0.0534, 0.0679], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 05:15:55,089 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:12,015 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0251, 3.1234, 1.9332, 3.2645, 2.4051, 3.3042, 2.1732, 2.5629], device='cuda:2'), covar=tensor([0.0340, 0.0410, 0.1684, 0.0293, 0.0808, 0.0690, 0.1475, 0.0806], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0171, 0.0179, 0.0219, 0.0205, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 05:16:13,706 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:46,425 INFO [train.py:904] (2/8) Epoch 26, batch 6050, loss[loss=0.1831, simple_loss=0.2809, pruned_loss=0.04269, over 16793.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2888, pruned_loss=0.05634, over 3110567.07 frames. ], batch size: 83, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:16:59,706 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259810.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:17:02,682 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2692, 2.9754, 3.2998, 1.7799, 3.4331, 3.4729, 2.8186, 2.6243], device='cuda:2'), covar=tensor([0.0854, 0.0312, 0.0234, 0.1268, 0.0115, 0.0216, 0.0464, 0.0510], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0112, 0.0100, 0.0139, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 05:17:49,540 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 05:17:50,994 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259843.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:17:58,342 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.44 vs. limit=2.0 2023-05-02 05:18:05,787 INFO [train.py:904] (2/8) Epoch 26, batch 6100, loss[loss=0.1987, simple_loss=0.2861, pruned_loss=0.05567, over 15495.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2889, pruned_loss=0.05564, over 3116352.75 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:18:09,302 INFO [optim.py:368] (2/8) 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 05:18:27,855 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:18:38,942 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 05:18:53,014 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259882.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:19:23,802 INFO [train.py:904] (2/8) Epoch 26, batch 6150, loss[loss=0.1844, simple_loss=0.2725, pruned_loss=0.04815, over 16850.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.287, pruned_loss=0.05523, over 3113824.17 frames. ], batch size: 116, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:19:41,866 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5354, 2.4644, 2.2881, 3.5160, 2.4808, 3.7176, 1.4739, 2.6282], device='cuda:2'), covar=tensor([0.1482, 0.0844, 0.1354, 0.0191, 0.0201, 0.0368, 0.1770, 0.0899], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0179, 0.0198, 0.0197, 0.0206, 0.0216, 0.0207, 0.0197], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 05:19:47,896 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1305, 2.3317, 2.3368, 3.8258, 2.1257, 2.6596, 2.3717, 2.4040], device='cuda:2'), covar=tensor([0.1519, 0.3228, 0.3006, 0.0650, 0.4256, 0.2462, 0.3342, 0.3246], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0467, 0.0380, 0.0333, 0.0444, 0.0534, 0.0438, 0.0545], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:19:48,885 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:00,776 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:05,846 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:42,381 INFO [train.py:904] (2/8) Epoch 26, batch 6200, loss[loss=0.2013, simple_loss=0.2841, pruned_loss=0.0592, over 16157.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2859, pruned_loss=0.05546, over 3121487.23 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:20:44,631 INFO [optim.py:368] (2/8) 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 05:21:04,923 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259968.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:22:00,546 INFO [train.py:904] (2/8) Epoch 26, batch 6250, loss[loss=0.2093, simple_loss=0.294, pruned_loss=0.06229, over 17123.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.286, pruned_loss=0.05539, over 3116184.39 frames. ], batch size: 48, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:22:41,261 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:22:54,209 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1549, 4.5278, 3.3069, 2.7395, 3.1773, 2.9187, 4.9897, 4.0269], device='cuda:2'), covar=tensor([0.2575, 0.0539, 0.1646, 0.2449, 0.2513, 0.1848, 0.0310, 0.1103], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0274, 0.0310, 0.0323, 0.0304, 0.0272, 0.0301, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 05:23:15,623 INFO [train.py:904] (2/8) Epoch 26, batch 6300, loss[loss=0.2128, simple_loss=0.2854, pruned_loss=0.07009, over 11612.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2853, pruned_loss=0.05493, over 3100246.59 frames. ], batch size: 246, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:23:19,600 INFO [optim.py:368] (2/8) 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 05:23:44,917 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260070.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:23:49,196 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 05:24:17,409 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:35,598 INFO [train.py:904] (2/8) Epoch 26, batch 6350, loss[loss=0.2756, simple_loss=0.3354, pruned_loss=0.1079, over 11132.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2859, pruned_loss=0.05558, over 3097648.90 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:24:39,493 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260105.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:24:59,054 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260118.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:25:24,174 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1221, 5.0991, 4.9683, 4.5942, 4.6460, 5.0313, 4.9827, 4.7334], device='cuda:2'), covar=tensor([0.0558, 0.0523, 0.0283, 0.0325, 0.0974, 0.0457, 0.0295, 0.0660], device='cuda:2'), in_proj_covar=tensor([0.0304, 0.0456, 0.0353, 0.0358, 0.0356, 0.0412, 0.0243, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:25:29,697 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260138.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:25:52,108 INFO [train.py:904] (2/8) Epoch 26, batch 6400, loss[loss=0.1873, simple_loss=0.2787, pruned_loss=0.04798, over 16864.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2856, pruned_loss=0.05642, over 3089130.69 frames. ], batch size: 116, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:25:54,618 INFO [optim.py:368] (2/8) 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 05:27:08,957 INFO [train.py:904] (2/8) Epoch 26, batch 6450, loss[loss=0.1899, simple_loss=0.2633, pruned_loss=0.05832, over 11780.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2854, pruned_loss=0.05581, over 3092132.48 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:27:39,884 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:27:43,822 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260224.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:28:13,795 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 05:28:28,518 INFO [train.py:904] (2/8) Epoch 26, batch 6500, loss[loss=0.1913, simple_loss=0.276, pruned_loss=0.05325, over 16966.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2836, pruned_loss=0.05524, over 3101739.39 frames. ], batch size: 41, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:28:31,534 INFO [optim.py:368] (2/8) 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 05:29:20,157 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260285.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:29:49,649 INFO [train.py:904] (2/8) Epoch 26, batch 6550, loss[loss=0.1842, simple_loss=0.2942, pruned_loss=0.03704, over 16657.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2864, pruned_loss=0.05618, over 3085719.92 frames. ], batch size: 76, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:07,551 INFO [train.py:904] (2/8) Epoch 26, batch 6600, loss[loss=0.2528, simple_loss=0.3251, pruned_loss=0.0902, over 11597.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.288, pruned_loss=0.05573, over 3101080.96 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:09,947 INFO [optim.py:368] (2/8) 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 05:31:20,515 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 05:31:54,714 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6800, 4.9855, 4.7241, 4.7510, 4.5120, 4.4358, 4.4226, 5.0478], device='cuda:2'), covar=tensor([0.1264, 0.0852, 0.1027, 0.0914, 0.0841, 0.1314, 0.1225, 0.0853], device='cuda:2'), in_proj_covar=tensor([0.0703, 0.0845, 0.0697, 0.0653, 0.0538, 0.0538, 0.0714, 0.0661], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:31:59,467 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:32:26,375 INFO [train.py:904] (2/8) Epoch 26, batch 6650, loss[loss=0.1852, simple_loss=0.2656, pruned_loss=0.05241, over 16468.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2875, pruned_loss=0.05593, over 3113492.36 frames. ], batch size: 35, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:32:30,300 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260405.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:33:21,180 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:33:43,209 INFO [train.py:904] (2/8) Epoch 26, batch 6700, loss[loss=0.1854, simple_loss=0.2666, pruned_loss=0.05207, over 16629.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2861, pruned_loss=0.05616, over 3105242.29 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:33:43,550 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260453.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:33:45,911 INFO [optim.py:368] (2/8) 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 05:34:35,491 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260486.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:34:53,828 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9219, 3.6466, 4.2129, 2.1296, 4.3905, 4.3915, 3.1994, 3.3049], device='cuda:2'), covar=tensor([0.0716, 0.0299, 0.0170, 0.1183, 0.0072, 0.0143, 0.0400, 0.0439], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 05:35:01,132 INFO [train.py:904] (2/8) Epoch 26, batch 6750, loss[loss=0.1794, simple_loss=0.2636, pruned_loss=0.04764, over 16893.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2853, pruned_loss=0.05634, over 3105172.40 frames. ], batch size: 116, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:35:24,251 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 05:35:31,743 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260522.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:36:19,268 INFO [train.py:904] (2/8) Epoch 26, batch 6800, loss[loss=0.1696, simple_loss=0.2681, pruned_loss=0.03553, over 16818.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2857, pruned_loss=0.05624, over 3119096.47 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:36:21,473 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:37:01,723 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260580.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:37:19,395 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-02 05:37:35,979 INFO [train.py:904] (2/8) Epoch 26, batch 6850, loss[loss=0.1892, simple_loss=0.2973, pruned_loss=0.04053, over 17250.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2867, pruned_loss=0.05621, over 3122412.78 frames. ], batch size: 52, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:04,682 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-02 05:38:50,099 INFO [train.py:904] (2/8) Epoch 26, batch 6900, loss[loss=0.2143, simple_loss=0.3014, pruned_loss=0.0636, over 16939.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2891, pruned_loss=0.05604, over 3123091.01 frames. ], batch size: 109, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:53,860 INFO [optim.py:368] (2/8) 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 05:39:09,432 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2054, 4.1901, 4.1210, 3.3506, 4.1708, 1.6674, 3.9514, 3.7977], device='cuda:2'), covar=tensor([0.0155, 0.0136, 0.0217, 0.0376, 0.0121, 0.3148, 0.0170, 0.0293], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0171, 0.0209, 0.0184, 0.0185, 0.0216, 0.0198, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:39:40,166 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260686.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:40:05,851 INFO [train.py:904] (2/8) Epoch 26, batch 6950, loss[loss=0.1879, simple_loss=0.2792, pruned_loss=0.04834, over 16429.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2913, pruned_loss=0.05783, over 3111685.12 frames. ], batch size: 75, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:40:25,175 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0354, 3.0572, 2.0205, 3.2320, 2.4444, 3.3033, 2.1979, 2.5689], device='cuda:2'), covar=tensor([0.0313, 0.0420, 0.1563, 0.0248, 0.0777, 0.0644, 0.1443, 0.0809], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0180, 0.0196, 0.0170, 0.0179, 0.0219, 0.0205, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 05:40:48,850 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:40:54,430 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260734.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:21,298 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:23,880 INFO [train.py:904] (2/8) Epoch 26, batch 7000, loss[loss=0.1767, simple_loss=0.2791, pruned_loss=0.03719, over 17004.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.291, pruned_loss=0.0574, over 3101605.13 frames. ], batch size: 50, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:41:29,415 INFO [optim.py:368] (2/8) 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 05:42:24,288 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:42:42,016 INFO [train.py:904] (2/8) Epoch 26, batch 7050, loss[loss=0.1962, simple_loss=0.2952, pruned_loss=0.04858, over 16679.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2915, pruned_loss=0.05659, over 3118336.26 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:42:55,739 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:43:32,896 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5997, 4.6482, 4.9848, 4.9344, 4.9571, 4.6733, 4.6279, 4.5229], device='cuda:2'), covar=tensor([0.0349, 0.0851, 0.0437, 0.0491, 0.0510, 0.0503, 0.0900, 0.0645], device='cuda:2'), in_proj_covar=tensor([0.0421, 0.0477, 0.0462, 0.0425, 0.0511, 0.0486, 0.0563, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 05:43:59,434 INFO [train.py:904] (2/8) Epoch 26, batch 7100, loss[loss=0.203, simple_loss=0.2983, pruned_loss=0.05384, over 16406.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2908, pruned_loss=0.0572, over 3097037.39 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:44:05,368 INFO [optim.py:368] (2/8) 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 05:44:42,757 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260880.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:44:44,322 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 26, batch 7150, loss[loss=0.1839, simple_loss=0.2732, pruned_loss=0.04733, over 16415.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2883, pruned_loss=0.05675, over 3098467.57 frames. ], batch size: 75, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:45:26,521 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 05:45:39,616 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0437, 5.0301, 4.8157, 4.1916, 4.9624, 1.8038, 4.6926, 4.5725], device='cuda:2'), covar=tensor([0.0085, 0.0096, 0.0232, 0.0450, 0.0095, 0.3113, 0.0132, 0.0277], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0171, 0.0209, 0.0184, 0.0185, 0.0216, 0.0198, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:45:53,426 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260928.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:46:01,361 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0639, 2.2615, 2.2579, 3.8307, 2.1408, 2.6176, 2.3232, 2.4263], device='cuda:2'), covar=tensor([0.1574, 0.3719, 0.3191, 0.0599, 0.4224, 0.2588, 0.3825, 0.3425], device='cuda:2'), in_proj_covar=tensor([0.0415, 0.0465, 0.0379, 0.0332, 0.0443, 0.0532, 0.0438, 0.0545], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:46:07,416 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 05:46:13,540 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260942.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:46:29,363 INFO [train.py:904] (2/8) Epoch 26, batch 7200, loss[loss=0.1879, simple_loss=0.2842, pruned_loss=0.04581, over 16349.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2861, pruned_loss=0.05465, over 3092278.17 frames. ], batch size: 35, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:46:35,554 INFO [optim.py:368] (2/8) 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 05:47:03,328 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9459, 4.7739, 4.5119, 3.1337, 3.9406, 4.5869, 3.9531, 2.5796], device='cuda:2'), covar=tensor([0.0501, 0.0038, 0.0055, 0.0386, 0.0117, 0.0121, 0.0114, 0.0452], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0134, 0.0100, 0.0113, 0.0097, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 05:47:53,028 INFO [train.py:904] (2/8) Epoch 26, batch 7250, loss[loss=0.1664, simple_loss=0.2559, pruned_loss=0.03843, over 16768.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.284, pruned_loss=0.05391, over 3075100.87 frames. ], batch size: 83, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:48:13,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8039, 4.8690, 5.2315, 5.1831, 5.2262, 4.9130, 4.8668, 4.7137], device='cuda:2'), covar=tensor([0.0338, 0.0556, 0.0406, 0.0456, 0.0475, 0.0448, 0.1037, 0.0521], device='cuda:2'), in_proj_covar=tensor([0.0421, 0.0477, 0.0463, 0.0425, 0.0511, 0.0486, 0.0564, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 05:49:09,253 INFO [train.py:904] (2/8) Epoch 26, batch 7300, loss[loss=0.205, simple_loss=0.3035, pruned_loss=0.05326, over 16224.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2835, pruned_loss=0.05373, over 3077026.58 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:49:15,972 INFO [optim.py:368] (2/8) 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 05:49:56,707 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8119, 2.1449, 2.4156, 3.0412, 2.1624, 2.3411, 2.3458, 2.2589], device='cuda:2'), covar=tensor([0.1425, 0.3140, 0.2467, 0.0752, 0.4063, 0.2255, 0.2955, 0.3248], device='cuda:2'), in_proj_covar=tensor([0.0415, 0.0465, 0.0378, 0.0332, 0.0443, 0.0533, 0.0438, 0.0545], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 05:50:00,841 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:50:26,312 INFO [train.py:904] (2/8) Epoch 26, batch 7350, loss[loss=0.1703, simple_loss=0.2683, pruned_loss=0.03617, over 16845.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.285, pruned_loss=0.05568, over 3021073.10 frames. ], batch size: 96, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:50:33,367 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261107.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:51:18,069 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261136.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:51:44,988 INFO [train.py:904] (2/8) Epoch 26, batch 7400, loss[loss=0.2497, simple_loss=0.3209, pruned_loss=0.08925, over 11189.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2863, pruned_loss=0.05612, over 3044929.12 frames. ], batch size: 249, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:51:50,753 INFO [optim.py:368] (2/8) 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 05:52:55,065 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261197.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:53:04,381 INFO [train.py:904] (2/8) Epoch 26, batch 7450, loss[loss=0.2284, simple_loss=0.3143, pruned_loss=0.07121, over 15275.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2871, pruned_loss=0.05723, over 3031321.97 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:54:01,621 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:54:27,095 INFO [train.py:904] (2/8) Epoch 26, batch 7500, loss[loss=0.2245, simple_loss=0.3043, pruned_loss=0.07235, over 15231.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2878, pruned_loss=0.05712, over 3026978.92 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:54:33,503 INFO [optim.py:368] (2/8) 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 05:55:45,643 INFO [train.py:904] (2/8) Epoch 26, batch 7550, loss[loss=0.2321, simple_loss=0.3003, pruned_loss=0.08199, over 11340.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2872, pruned_loss=0.05745, over 3025959.33 frames. ], batch size: 246, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:57:01,728 INFO [train.py:904] (2/8) Epoch 26, batch 7600, loss[loss=0.1871, simple_loss=0.2787, pruned_loss=0.0478, over 16490.00 frames. ], tot_loss[loss=0.2, simple_loss=0.286, pruned_loss=0.05706, over 3056539.76 frames. ], batch size: 75, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 05:57:07,520 INFO [optim.py:368] (2/8) 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 05:57:54,549 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:58:20,640 INFO [train.py:904] (2/8) Epoch 26, batch 7650, loss[loss=0.2301, simple_loss=0.3059, pruned_loss=0.07714, over 11890.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2867, pruned_loss=0.05764, over 3056164.00 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 05:58:26,576 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:05,931 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6374, 3.8607, 2.8348, 2.3087, 2.5872, 2.5002, 4.1444, 3.4731], device='cuda:2'), covar=tensor([0.3029, 0.0673, 0.1853, 0.2838, 0.2757, 0.2096, 0.0445, 0.1270], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0273, 0.0309, 0.0321, 0.0302, 0.0271, 0.0300, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 05:59:08,084 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:36,565 INFO [train.py:904] (2/8) Epoch 26, batch 7700, loss[loss=0.195, simple_loss=0.2826, pruned_loss=0.05371, over 15331.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.287, pruned_loss=0.05792, over 3055178.07 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 05:59:40,110 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261455.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:42,110 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:46,564 INFO [optim.py:368] (2/8) 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 06:00:36,505 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261492.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 06:00:47,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5288, 3.5953, 3.3304, 2.9785, 3.2032, 3.4765, 3.3298, 3.3138], device='cuda:2'), covar=tensor([0.0622, 0.0858, 0.0304, 0.0309, 0.0528, 0.0555, 0.1270, 0.0526], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0448, 0.0348, 0.0350, 0.0348, 0.0404, 0.0239, 0.0421], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 06:00:52,205 INFO [train.py:904] (2/8) Epoch 26, batch 7750, loss[loss=0.1941, simple_loss=0.2818, pruned_loss=0.05317, over 16762.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2871, pruned_loss=0.05773, over 3055662.62 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:01:13,782 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:01:45,016 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:02:09,780 INFO [train.py:904] (2/8) Epoch 26, batch 7800, loss[loss=0.2108, simple_loss=0.3057, pruned_loss=0.05794, over 16927.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2877, pruned_loss=0.05781, over 3077858.86 frames. ], batch size: 109, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:02:19,354 INFO [optim.py:368] (2/8) 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 06:02:19,899 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8482, 2.6786, 2.5613, 1.8646, 2.6044, 2.7381, 2.5812, 1.7399], device='cuda:2'), covar=tensor([0.0489, 0.0129, 0.0122, 0.0448, 0.0172, 0.0188, 0.0155, 0.0571], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0087, 0.0089, 0.0134, 0.0100, 0.0113, 0.0097, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 06:02:27,593 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 06:03:00,429 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261585.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:03:09,918 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 06:03:28,317 INFO [train.py:904] (2/8) Epoch 26, batch 7850, loss[loss=0.2088, simple_loss=0.2906, pruned_loss=0.0635, over 11502.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2879, pruned_loss=0.05752, over 3078015.69 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:03:41,986 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8861, 2.2100, 2.5234, 3.1670, 2.2361, 2.3834, 2.3814, 2.3342], device='cuda:2'), covar=tensor([0.1518, 0.3349, 0.2512, 0.0753, 0.4221, 0.2371, 0.3230, 0.3210], device='cuda:2'), in_proj_covar=tensor([0.0412, 0.0464, 0.0378, 0.0331, 0.0442, 0.0531, 0.0436, 0.0543], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 06:04:46,515 INFO [train.py:904] (2/8) Epoch 26, batch 7900, loss[loss=0.1774, simple_loss=0.2706, pruned_loss=0.04213, over 16544.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2864, pruned_loss=0.05677, over 3069300.46 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:55,564 INFO [optim.py:368] (2/8) 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 06:06:01,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7405, 3.4950, 3.9702, 1.9821, 4.1128, 4.2244, 3.1209, 3.1328], device='cuda:2'), covar=tensor([0.0748, 0.0300, 0.0219, 0.1236, 0.0085, 0.0163, 0.0425, 0.0436], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0138, 0.0086, 0.0129, 0.0129, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 06:06:05,689 INFO [train.py:904] (2/8) Epoch 26, batch 7950, loss[loss=0.1923, simple_loss=0.2718, pruned_loss=0.05638, over 16866.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.287, pruned_loss=0.05715, over 3062860.02 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:06:59,441 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5023, 3.6721, 2.2632, 4.1613, 2.8913, 4.1210, 2.4546, 2.9218], device='cuda:2'), covar=tensor([0.0345, 0.0371, 0.1626, 0.0246, 0.0795, 0.0618, 0.1470, 0.0785], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0181, 0.0198, 0.0172, 0.0180, 0.0221, 0.0207, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 06:06:59,768 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-02 06:07:23,908 INFO [train.py:904] (2/8) Epoch 26, batch 8000, loss[loss=0.2148, simple_loss=0.3001, pruned_loss=0.06469, over 16342.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2879, pruned_loss=0.05798, over 3055086.68 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:07:32,670 INFO [optim.py:368] (2/8) 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 06:07:55,213 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-02 06:08:23,771 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261792.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 06:08:40,305 INFO [train.py:904] (2/8) Epoch 26, batch 8050, loss[loss=0.217, simple_loss=0.3083, pruned_loss=0.06283, over 17184.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2882, pruned_loss=0.05743, over 3077007.74 frames. ], batch size: 46, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:08:54,195 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:13,004 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261824.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:27,153 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:37,845 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261840.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 06:09:57,223 INFO [train.py:904] (2/8) Epoch 26, batch 8100, loss[loss=0.2454, simple_loss=0.3096, pruned_loss=0.09063, over 11639.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2876, pruned_loss=0.05746, over 3060650.79 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:10:06,903 INFO [optim.py:368] (2/8) 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 06:10:46,784 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:00,399 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:14,754 INFO [train.py:904] (2/8) Epoch 26, batch 8150, loss[loss=0.208, simple_loss=0.2801, pruned_loss=0.06796, over 11637.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2853, pruned_loss=0.05607, over 3072086.99 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:11:16,849 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 06:12:33,065 INFO [train.py:904] (2/8) Epoch 26, batch 8200, loss[loss=0.2114, simple_loss=0.2774, pruned_loss=0.07267, over 11952.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2822, pruned_loss=0.05529, over 3095873.36 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:43,219 INFO [optim.py:368] (2/8) 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 06:13:13,003 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261977.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:13:25,171 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 06:13:58,691 INFO [train.py:904] (2/8) Epoch 26, batch 8250, loss[loss=0.1833, simple_loss=0.2818, pruned_loss=0.04241, over 16398.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2812, pruned_loss=0.05244, over 3096457.66 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:14:57,039 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262038.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:15:22,683 INFO [train.py:904] (2/8) Epoch 26, batch 8300, loss[loss=0.1755, simple_loss=0.2795, pruned_loss=0.03577, over 16711.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2789, pruned_loss=0.04995, over 3064287.34 frames. ], batch size: 89, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:15:32,830 INFO [optim.py:368] (2/8) 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 06:16:44,227 INFO [train.py:904] (2/8) Epoch 26, batch 8350, loss[loss=0.176, simple_loss=0.2715, pruned_loss=0.04028, over 16905.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2788, pruned_loss=0.04841, over 3057678.06 frames. ], batch size: 109, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:16:58,992 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:17:14,300 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8940, 3.2502, 3.5807, 2.0284, 3.0715, 2.2636, 3.4476, 3.4130], device='cuda:2'), covar=tensor([0.0281, 0.0844, 0.0472, 0.2181, 0.0754, 0.1015, 0.0599, 0.0996], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0167, 0.0154, 0.0145, 0.0130, 0.0144, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 06:18:04,038 INFO [train.py:904] (2/8) Epoch 26, batch 8400, loss[loss=0.1914, simple_loss=0.2858, pruned_loss=0.04853, over 15242.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2766, pruned_loss=0.04666, over 3056978.03 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:18:13,168 INFO [optim.py:368] (2/8) 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 06:18:14,935 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:18:42,973 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 06:18:47,907 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262180.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:02,234 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:13,094 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-02 06:19:24,257 INFO [train.py:904] (2/8) Epoch 26, batch 8450, loss[loss=0.1629, simple_loss=0.2597, pruned_loss=0.03303, over 15363.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2752, pruned_loss=0.04542, over 3060960.48 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:41,772 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6682, 2.5977, 1.9245, 2.7892, 2.1290, 2.7992, 2.2191, 2.4425], device='cuda:2'), covar=tensor([0.0299, 0.0354, 0.1218, 0.0288, 0.0639, 0.0417, 0.1218, 0.0508], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0176, 0.0193, 0.0167, 0.0176, 0.0214, 0.0202, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 06:20:48,803 INFO [train.py:904] (2/8) Epoch 26, batch 8500, loss[loss=0.1475, simple_loss=0.2338, pruned_loss=0.03056, over 11959.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2713, pruned_loss=0.04327, over 3038805.72 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:57,922 INFO [optim.py:368] (2/8) 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 06:21:03,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9112, 2.7590, 2.9474, 2.2094, 2.7631, 2.1788, 2.8054, 2.9299], device='cuda:2'), covar=tensor([0.0272, 0.0938, 0.0472, 0.1797, 0.0745, 0.0988, 0.0596, 0.0851], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0165, 0.0165, 0.0152, 0.0143, 0.0129, 0.0142, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 06:21:55,775 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0435, 4.1438, 3.9430, 3.6394, 3.6873, 4.0486, 3.7199, 3.8311], device='cuda:2'), covar=tensor([0.0640, 0.0737, 0.0356, 0.0338, 0.0754, 0.0592, 0.1134, 0.0725], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0451, 0.0350, 0.0352, 0.0349, 0.0405, 0.0242, 0.0423], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 06:22:09,819 INFO [train.py:904] (2/8) Epoch 26, batch 8550, loss[loss=0.1702, simple_loss=0.2506, pruned_loss=0.04484, over 12186.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2692, pruned_loss=0.04259, over 3010327.95 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:07,968 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:23:47,973 INFO [train.py:904] (2/8) Epoch 26, batch 8600, loss[loss=0.1658, simple_loss=0.2505, pruned_loss=0.04054, over 12375.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2687, pruned_loss=0.04137, over 2999187.26 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:59,713 INFO [optim.py:368] (2/8) 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 06:25:09,708 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9102, 2.1609, 2.3824, 3.1897, 2.1960, 2.3604, 2.3109, 2.2177], device='cuda:2'), covar=tensor([0.1440, 0.3821, 0.2895, 0.0765, 0.4680, 0.2854, 0.3775, 0.4192], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0459, 0.0376, 0.0326, 0.0437, 0.0524, 0.0431, 0.0537], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 06:25:27,328 INFO [train.py:904] (2/8) Epoch 26, batch 8650, loss[loss=0.1616, simple_loss=0.2702, pruned_loss=0.0265, over 16603.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2669, pruned_loss=0.03982, over 3025426.16 frames. ], batch size: 75, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:26:36,222 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-05-02 06:27:13,545 INFO [train.py:904] (2/8) Epoch 26, batch 8700, loss[loss=0.1523, simple_loss=0.2513, pruned_loss=0.02664, over 16870.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2638, pruned_loss=0.03853, over 3017063.03 frames. ], batch size: 90, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:27:25,316 INFO [optim.py:368] (2/8) 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 06:28:02,805 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:20,327 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:48,924 INFO [train.py:904] (2/8) Epoch 26, batch 8750, loss[loss=0.198, simple_loss=0.3032, pruned_loss=0.04641, over 16658.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2636, pruned_loss=0.03794, over 3035143.76 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:29:48,636 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:09,755 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:41,532 INFO [train.py:904] (2/8) Epoch 26, batch 8800, loss[loss=0.163, simple_loss=0.2592, pruned_loss=0.03338, over 15419.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2622, pruned_loss=0.03675, over 3045654.29 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:30:52,411 INFO [optim.py:368] (2/8) 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 06:31:01,630 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5014, 3.4371, 3.5355, 3.5905, 3.6359, 3.3660, 3.6341, 3.6814], device='cuda:2'), covar=tensor([0.1262, 0.0933, 0.0969, 0.0650, 0.0652, 0.2029, 0.0762, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0630, 0.0781, 0.0896, 0.0791, 0.0603, 0.0627, 0.0653, 0.0768], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 06:31:04,511 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-02 06:31:12,801 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0531, 3.1658, 3.2119, 2.1754, 2.9444, 3.2625, 3.1076, 1.9447], device='cuda:2'), covar=tensor([0.0541, 0.0069, 0.0072, 0.0434, 0.0137, 0.0084, 0.0090, 0.0506], device='cuda:2'), in_proj_covar=tensor([0.0134, 0.0085, 0.0086, 0.0131, 0.0098, 0.0110, 0.0095, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 06:32:25,982 INFO [train.py:904] (2/8) Epoch 26, batch 8850, loss[loss=0.1724, simple_loss=0.2741, pruned_loss=0.03537, over 17044.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2644, pruned_loss=0.03618, over 3042951.61 frames. ], batch size: 55, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:33:32,223 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262633.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:34:13,875 INFO [train.py:904] (2/8) Epoch 26, batch 8900, loss[loss=0.1808, simple_loss=0.2788, pruned_loss=0.0414, over 16448.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.265, pruned_loss=0.03581, over 3047887.57 frames. ], batch size: 147, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:34:26,822 INFO [optim.py:368] (2/8) 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 06:35:20,125 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262681.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:36:18,596 INFO [train.py:904] (2/8) Epoch 26, batch 8950, loss[loss=0.1624, simple_loss=0.2523, pruned_loss=0.03624, over 12850.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2645, pruned_loss=0.03578, over 3072204.42 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:38:00,225 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=262749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:38:08,094 INFO [train.py:904] (2/8) Epoch 26, batch 9000, loss[loss=0.1573, simple_loss=0.2516, pruned_loss=0.03147, over 16284.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2617, pruned_loss=0.03484, over 3076289.63 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:38:08,094 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 06:38:17,363 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1796, 2.6861, 3.1031, 2.0729, 2.8642, 2.1968, 2.9171, 2.8928], device='cuda:2'), covar=tensor([0.0293, 0.1186, 0.0496, 0.2065, 0.0771, 0.0991, 0.0659, 0.1140], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0165, 0.0153, 0.0144, 0.0129, 0.0143, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 06:38:18,508 INFO [train.py:938] (2/8) Epoch 26, validation: loss=0.1435, simple_loss=0.2475, pruned_loss=0.01976, over 944034.00 frames. 2023-05-02 06:38:18,509 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 06:38:30,743 INFO [optim.py:368] (2/8) 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 06:40:03,318 INFO [train.py:904] (2/8) Epoch 26, batch 9050, loss[loss=0.1732, simple_loss=0.2673, pruned_loss=0.03953, over 16369.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2635, pruned_loss=0.03558, over 3093468.06 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:40:20,460 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:41:48,707 INFO [train.py:904] (2/8) Epoch 26, batch 9100, loss[loss=0.1751, simple_loss=0.2787, pruned_loss=0.03579, over 15184.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2629, pruned_loss=0.03607, over 3099291.46 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:42:01,224 INFO [optim.py:368] (2/8) 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 06:42:33,521 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 06:43:45,862 INFO [train.py:904] (2/8) Epoch 26, batch 9150, loss[loss=0.1614, simple_loss=0.2539, pruned_loss=0.03443, over 15414.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2629, pruned_loss=0.03552, over 3088622.98 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:44:01,159 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-05-02 06:45:27,072 INFO [train.py:904] (2/8) Epoch 26, batch 9200, loss[loss=0.1692, simple_loss=0.2667, pruned_loss=0.03586, over 15383.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2591, pruned_loss=0.03473, over 3095833.03 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:45:36,599 INFO [optim.py:368] (2/8) 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 06:45:41,346 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-02 06:46:23,659 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 06:47:01,949 INFO [train.py:904] (2/8) Epoch 26, batch 9250, loss[loss=0.1737, simple_loss=0.2621, pruned_loss=0.04262, over 16911.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2584, pruned_loss=0.03456, over 3089679.71 frames. ], batch size: 116, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:48:49,716 INFO [train.py:904] (2/8) Epoch 26, batch 9300, loss[loss=0.1515, simple_loss=0.2368, pruned_loss=0.03313, over 12233.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2566, pruned_loss=0.03431, over 3066260.52 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:49:02,023 INFO [optim.py:368] (2/8) 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 06:50:20,305 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:33,214 INFO [train.py:904] (2/8) Epoch 26, batch 9350, loss[loss=0.1625, simple_loss=0.2555, pruned_loss=0.03472, over 16983.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2567, pruned_loss=0.03449, over 3074251.94 frames. ], batch size: 109, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:50:38,234 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263105.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:51,066 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:58,702 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9020, 2.7999, 2.7222, 1.9790, 2.5754, 2.8035, 2.6939, 1.8654], device='cuda:2'), covar=tensor([0.0483, 0.0081, 0.0083, 0.0436, 0.0147, 0.0120, 0.0110, 0.0504], device='cuda:2'), in_proj_covar=tensor([0.0133, 0.0085, 0.0086, 0.0131, 0.0098, 0.0109, 0.0094, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 06:51:48,038 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:14,167 INFO [train.py:904] (2/8) Epoch 26, batch 9400, loss[loss=0.1509, simple_loss=0.2599, pruned_loss=0.02095, over 16678.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2566, pruned_loss=0.03405, over 3085059.40 frames. ], batch size: 89, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:52:19,896 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:24,354 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0316, 2.0445, 2.1101, 3.7212, 1.9672, 2.3676, 2.2023, 2.2025], device='cuda:2'), covar=tensor([0.1570, 0.4498, 0.3535, 0.0629, 0.4919, 0.3017, 0.4328, 0.3988], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0457, 0.0376, 0.0324, 0.0436, 0.0521, 0.0429, 0.0534], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 06:52:25,026 INFO [optim.py:368] (2/8) 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 06:52:52,085 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:27,186 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 06:53:29,259 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 06:53:50,914 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:53,495 INFO [train.py:904] (2/8) Epoch 26, batch 9450, loss[loss=0.1568, simple_loss=0.2514, pruned_loss=0.0311, over 16491.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.259, pruned_loss=0.03434, over 3087258.15 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:54:14,258 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:55:19,881 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263246.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:55:33,109 INFO [train.py:904] (2/8) Epoch 26, batch 9500, loss[loss=0.1574, simple_loss=0.2522, pruned_loss=0.0313, over 16912.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2583, pruned_loss=0.03403, over 3079541.77 frames. ], batch size: 42, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:55:47,501 INFO [optim.py:368] (2/8) 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 06:56:19,464 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:57:17,615 INFO [train.py:904] (2/8) Epoch 26, batch 9550, loss[loss=0.1706, simple_loss=0.267, pruned_loss=0.03707, over 16834.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2581, pruned_loss=0.03409, over 3072111.80 frames. ], batch size: 90, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:57:27,260 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:58:22,179 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0197, 2.2949, 2.3652, 3.0516, 1.8215, 3.2912, 1.8145, 2.8342], device='cuda:2'), covar=tensor([0.1202, 0.0677, 0.1023, 0.0162, 0.0086, 0.0369, 0.1469, 0.0642], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0175, 0.0195, 0.0191, 0.0199, 0.0213, 0.0205, 0.0193], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 06:58:58,572 INFO [train.py:904] (2/8) Epoch 26, batch 9600, loss[loss=0.1757, simple_loss=0.2772, pruned_loss=0.03711, over 15240.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2598, pruned_loss=0.03497, over 3066389.23 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:59:09,908 INFO [optim.py:368] (2/8) 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 06:59:49,318 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 07:00:45,647 INFO [train.py:904] (2/8) Epoch 26, batch 9650, loss[loss=0.1597, simple_loss=0.2586, pruned_loss=0.03044, over 16398.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2611, pruned_loss=0.03517, over 3054208.61 frames. ], batch size: 146, lr: 2.56e-03, grad_scale: 8.0 2023-05-02 07:00:51,811 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:00:54,920 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 07:01:50,800 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 07:02:28,447 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263451.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:31,833 INFO [train.py:904] (2/8) Epoch 26, batch 9700, loss[loss=0.1736, simple_loss=0.2683, pruned_loss=0.03942, over 15317.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2598, pruned_loss=0.03511, over 3037959.59 frames. ], batch size: 190, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:02:33,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:42,007 INFO [optim.py:368] (2/8) 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 07:02:58,810 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:04:01,226 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:04:13,678 INFO [train.py:904] (2/8) Epoch 26, batch 9750, loss[loss=0.1734, simple_loss=0.2699, pruned_loss=0.03844, over 16726.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2585, pruned_loss=0.03507, over 3033049.51 frames. ], batch size: 134, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:05:16,197 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9180, 3.2261, 3.5422, 2.0109, 2.9596, 2.2328, 3.3974, 3.4040], device='cuda:2'), covar=tensor([0.0282, 0.0890, 0.0514, 0.2192, 0.0821, 0.1034, 0.0708, 0.1032], device='cuda:2'), in_proj_covar=tensor([0.0154, 0.0161, 0.0163, 0.0151, 0.0142, 0.0127, 0.0141, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:05:51,431 INFO [train.py:904] (2/8) Epoch 26, batch 9800, loss[loss=0.1826, simple_loss=0.289, pruned_loss=0.03812, over 16420.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2583, pruned_loss=0.03398, over 3051700.95 frames. ], batch size: 146, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:06:03,262 INFO [optim.py:368] (2/8) 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 07:06:16,648 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4426, 1.6518, 2.0604, 2.4411, 2.3312, 2.7795, 1.8683, 2.6908], device='cuda:2'), covar=tensor([0.0280, 0.0653, 0.0436, 0.0385, 0.0422, 0.0228, 0.0622, 0.0188], device='cuda:2'), in_proj_covar=tensor([0.0189, 0.0192, 0.0180, 0.0184, 0.0199, 0.0157, 0.0197, 0.0157], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:06:23,918 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:07:33,256 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263602.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:07:34,239 INFO [train.py:904] (2/8) Epoch 26, batch 9850, loss[loss=0.1483, simple_loss=0.2462, pruned_loss=0.02523, over 16415.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2593, pruned_loss=0.03371, over 3035401.77 frames. ], batch size: 68, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:24,088 INFO [train.py:904] (2/8) Epoch 26, batch 9900, loss[loss=0.1559, simple_loss=0.2498, pruned_loss=0.03102, over 12393.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2595, pruned_loss=0.03367, over 3033892.22 frames. ], batch size: 250, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:36,802 INFO [optim.py:368] (2/8) 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 07:11:04,295 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2247, 5.5492, 5.3549, 5.3494, 5.0217, 5.0629, 4.9162, 5.6688], device='cuda:2'), covar=tensor([0.1326, 0.0863, 0.0916, 0.0744, 0.0810, 0.0807, 0.1345, 0.0907], device='cuda:2'), in_proj_covar=tensor([0.0679, 0.0816, 0.0668, 0.0627, 0.0521, 0.0521, 0.0687, 0.0641], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:11:20,738 INFO [train.py:904] (2/8) Epoch 26, batch 9950, loss[loss=0.1836, simple_loss=0.2707, pruned_loss=0.04826, over 16625.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2629, pruned_loss=0.03431, over 3061427.84 frames. ], batch size: 57, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:19,027 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:13:21,140 INFO [train.py:904] (2/8) Epoch 26, batch 10000, loss[loss=0.1568, simple_loss=0.2507, pruned_loss=0.03141, over 13138.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2614, pruned_loss=0.0339, over 3079005.66 frames. ], batch size: 250, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:34,912 INFO [optim.py:368] (2/8) 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 07:13:51,359 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:48,302 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:53,763 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:00,041 INFO [train.py:904] (2/8) Epoch 26, batch 10050, loss[loss=0.1563, simple_loss=0.2493, pruned_loss=0.03162, over 16406.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.262, pruned_loss=0.03388, over 3099951.52 frames. ], batch size: 35, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:15:23,649 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263815.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:57,726 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263833.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:16:17,281 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 26, batch 10100, loss[loss=0.1445, simple_loss=0.2322, pruned_loss=0.02836, over 13010.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2623, pruned_loss=0.03407, over 3107103.90 frames. ], batch size: 248, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:16:39,584 INFO [optim.py:368] (2/8) 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 07:17:02,797 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:17:36,291 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263894.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:18:08,982 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 0, loss[loss=0.2247, simple_loss=0.2876, pruned_loss=0.08091, over 16872.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.2876, pruned_loss=0.08091, over 16872.00 frames. ], batch size: 116, lr: 2.51e-03, grad_scale: 16.0 2023-05-02 07:18:09,862 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 07:18:17,107 INFO [train.py:938] (2/8) Epoch 27, validation: loss=0.1434, simple_loss=0.2467, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 07:18:17,107 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 07:18:30,566 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1144, 4.8272, 5.0926, 5.2697, 5.4915, 4.7419, 5.3942, 5.4599], device='cuda:2'), covar=tensor([0.2121, 0.1480, 0.1956, 0.0926, 0.0718, 0.0898, 0.0698, 0.0775], device='cuda:2'), in_proj_covar=tensor([0.0631, 0.0778, 0.0891, 0.0788, 0.0600, 0.0623, 0.0654, 0.0760], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:18:37,956 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:18:49,616 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1111, 5.1186, 5.5036, 5.4759, 5.5325, 5.1992, 5.1271, 4.8931], device='cuda:2'), covar=tensor([0.0365, 0.0512, 0.0409, 0.0486, 0.0536, 0.0451, 0.1027, 0.0525], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0457, 0.0444, 0.0408, 0.0492, 0.0467, 0.0537, 0.0374], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 07:18:59,720 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 07:19:22,226 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:26,172 INFO [train.py:904] (2/8) Epoch 27, batch 50, loss[loss=0.1962, simple_loss=0.2679, pruned_loss=0.06224, over 16912.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2601, pruned_loss=0.04432, over 749322.57 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:19:36,247 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263960.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:39,886 INFO [optim.py:368] (2/8) 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 07:19:46,797 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2952, 2.3257, 2.4163, 4.0657, 2.2938, 2.6424, 2.4143, 2.4271], device='cuda:2'), covar=tensor([0.1390, 0.3700, 0.3105, 0.0591, 0.4203, 0.2697, 0.3735, 0.3544], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0459, 0.0377, 0.0325, 0.0436, 0.0521, 0.0430, 0.0532], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:20:27,010 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4834, 2.3388, 2.3119, 4.2385, 2.2981, 2.7289, 2.4531, 2.5026], device='cuda:2'), covar=tensor([0.1379, 0.3940, 0.3377, 0.0555, 0.4433, 0.2726, 0.3919, 0.3718], device='cuda:2'), in_proj_covar=tensor([0.0409, 0.0461, 0.0378, 0.0326, 0.0438, 0.0523, 0.0431, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:20:36,408 INFO [train.py:904] (2/8) Epoch 27, batch 100, loss[loss=0.1557, simple_loss=0.2562, pruned_loss=0.02757, over 17125.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2595, pruned_loss=0.0426, over 1315584.94 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:00,955 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:21:23,873 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-02 07:21:44,757 INFO [train.py:904] (2/8) Epoch 27, batch 150, loss[loss=0.1464, simple_loss=0.2295, pruned_loss=0.03163, over 16830.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2587, pruned_loss=0.04245, over 1765659.74 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:55,810 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7281, 2.8550, 3.1841, 2.0782, 2.7356, 2.0748, 3.2734, 3.2804], device='cuda:2'), covar=tensor([0.0261, 0.1072, 0.0630, 0.2079, 0.0956, 0.1159, 0.0605, 0.0995], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0163, 0.0166, 0.0153, 0.0144, 0.0129, 0.0143, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:21:57,549 INFO [optim.py:368] (2/8) 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 07:22:53,632 INFO [train.py:904] (2/8) Epoch 27, batch 200, loss[loss=0.1902, simple_loss=0.2655, pruned_loss=0.0575, over 16770.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2595, pruned_loss=0.04263, over 2116741.04 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:23:34,781 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1649, 5.8753, 5.9864, 5.6468, 5.7916, 6.3178, 5.8472, 5.4935], device='cuda:2'), covar=tensor([0.0974, 0.2196, 0.2842, 0.2139, 0.2510, 0.0950, 0.1675, 0.2397], device='cuda:2'), in_proj_covar=tensor([0.0410, 0.0607, 0.0673, 0.0497, 0.0656, 0.0699, 0.0525, 0.0662], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:24:00,842 INFO [train.py:904] (2/8) Epoch 27, batch 250, loss[loss=0.1783, simple_loss=0.2586, pruned_loss=0.04904, over 16591.00 frames. ], tot_loss[loss=0.171, simple_loss=0.258, pruned_loss=0.04201, over 2384140.33 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:24:09,851 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4813, 5.8629, 5.6434, 5.6589, 5.2599, 5.3340, 5.1996, 5.9933], device='cuda:2'), covar=tensor([0.1522, 0.1053, 0.0980, 0.0895, 0.0948, 0.0726, 0.1398, 0.1035], device='cuda:2'), in_proj_covar=tensor([0.0691, 0.0832, 0.0680, 0.0640, 0.0532, 0.0529, 0.0701, 0.0653], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:24:14,011 INFO [optim.py:368] (2/8) 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 07:24:50,652 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264189.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:25:10,467 INFO [train.py:904] (2/8) Epoch 27, batch 300, loss[loss=0.1671, simple_loss=0.2653, pruned_loss=0.03449, over 17038.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2563, pruned_loss=0.04058, over 2590163.27 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:25:31,448 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8308, 5.1761, 5.3197, 5.0729, 5.1019, 5.7357, 5.2241, 4.9532], device='cuda:2'), covar=tensor([0.1370, 0.2082, 0.2605, 0.2352, 0.2708, 0.1052, 0.1739, 0.2414], device='cuda:2'), in_proj_covar=tensor([0.0412, 0.0610, 0.0677, 0.0499, 0.0661, 0.0701, 0.0528, 0.0665], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:25:52,974 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:26:19,419 INFO [train.py:904] (2/8) Epoch 27, batch 350, loss[loss=0.1672, simple_loss=0.2515, pruned_loss=0.04147, over 12691.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2545, pruned_loss=0.03979, over 2749226.90 frames. ], batch size: 248, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:26:26,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7435, 3.7908, 2.9663, 2.2966, 2.3851, 2.3688, 3.9181, 3.2282], device='cuda:2'), covar=tensor([0.2820, 0.0647, 0.1723, 0.3414, 0.3025, 0.2361, 0.0511, 0.1696], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0272, 0.0310, 0.0322, 0.0298, 0.0272, 0.0301, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:26:34,329 INFO [optim.py:368] (2/8) 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 07:26:41,096 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6455, 3.7417, 2.9015, 2.3057, 2.4484, 2.3820, 3.8253, 3.2457], device='cuda:2'), covar=tensor([0.3147, 0.0732, 0.1872, 0.3482, 0.2974, 0.2387, 0.0684, 0.1793], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0272, 0.0311, 0.0322, 0.0299, 0.0273, 0.0301, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:27:12,075 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6747, 3.6793, 2.3014, 4.0182, 3.0211, 3.9271, 2.4002, 3.0372], device='cuda:2'), covar=tensor([0.0270, 0.0407, 0.1512, 0.0347, 0.0695, 0.0684, 0.1434, 0.0654], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0170, 0.0179, 0.0217, 0.0204, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:27:17,352 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:27:23,623 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 07:27:27,689 INFO [train.py:904] (2/8) Epoch 27, batch 400, loss[loss=0.1923, simple_loss=0.266, pruned_loss=0.05927, over 16722.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2532, pruned_loss=0.04007, over 2871019.95 frames. ], batch size: 89, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:27:44,539 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264316.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:28:28,162 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6242, 3.7536, 2.4826, 4.3970, 2.9767, 4.3045, 2.5897, 3.1402], device='cuda:2'), covar=tensor([0.0378, 0.0455, 0.1644, 0.0359, 0.0909, 0.0588, 0.1567, 0.0845], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0170, 0.0179, 0.0217, 0.0205, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:28:33,908 INFO [train.py:904] (2/8) Epoch 27, batch 450, loss[loss=0.145, simple_loss=0.2352, pruned_loss=0.02742, over 16774.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2526, pruned_loss=0.03949, over 2974050.46 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:28:34,543 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3957, 2.2995, 2.2430, 4.0745, 2.3032, 2.6296, 2.3839, 2.4619], device='cuda:2'), covar=tensor([0.1453, 0.3958, 0.3618, 0.0641, 0.4513, 0.2854, 0.3876, 0.3963], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0469, 0.0385, 0.0333, 0.0444, 0.0533, 0.0439, 0.0546], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:28:48,386 INFO [optim.py:368] (2/8) 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 07:29:44,911 INFO [train.py:904] (2/8) Epoch 27, batch 500, loss[loss=0.1637, simple_loss=0.264, pruned_loss=0.03167, over 17101.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2505, pruned_loss=0.0384, over 3052393.50 frames. ], batch size: 47, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:30:24,725 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6551, 3.8786, 2.5221, 4.4224, 3.1685, 4.3390, 2.5822, 3.2591], device='cuda:2'), covar=tensor([0.0407, 0.0432, 0.1611, 0.0379, 0.0811, 0.0629, 0.1496, 0.0746], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0179, 0.0196, 0.0171, 0.0179, 0.0218, 0.0205, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:30:51,106 INFO [train.py:904] (2/8) Epoch 27, batch 550, loss[loss=0.1623, simple_loss=0.2513, pruned_loss=0.0366, over 16500.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2497, pruned_loss=0.03835, over 3113766.90 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:31:04,229 INFO [optim.py:368] (2/8) 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 07:31:18,820 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2364, 5.2132, 4.9790, 4.5562, 5.0446, 2.0648, 4.7836, 4.8571], device='cuda:2'), covar=tensor([0.0090, 0.0091, 0.0246, 0.0388, 0.0114, 0.2652, 0.0158, 0.0250], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0180, 0.0184, 0.0216, 0.0197, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:31:26,695 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264479.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:31:40,790 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264489.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:31:55,563 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264500.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:31:59,347 INFO [train.py:904] (2/8) Epoch 27, batch 600, loss[loss=0.1749, simple_loss=0.2517, pruned_loss=0.04899, over 16948.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2498, pruned_loss=0.03921, over 3159021.10 frames. ], batch size: 116, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:32:04,414 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 07:32:46,462 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264537.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:32:50,458 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:32:58,171 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2861, 3.7502, 4.2794, 2.4105, 4.5021, 4.6213, 3.4044, 3.4681], device='cuda:2'), covar=tensor([0.0636, 0.0317, 0.0255, 0.1107, 0.0093, 0.0174, 0.0432, 0.0447], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0111, 0.0099, 0.0139, 0.0085, 0.0129, 0.0130, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:33:08,795 INFO [train.py:904] (2/8) Epoch 27, batch 650, loss[loss=0.1503, simple_loss=0.2229, pruned_loss=0.03886, over 16884.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.248, pruned_loss=0.03845, over 3186786.91 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:33:17,737 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7615, 4.1457, 4.1521, 2.9614, 3.5984, 4.1804, 3.8319, 2.4751], device='cuda:2'), covar=tensor([0.0513, 0.0102, 0.0064, 0.0388, 0.0142, 0.0126, 0.0110, 0.0513], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0134, 0.0100, 0.0112, 0.0096, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 07:33:18,868 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264561.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:33:20,704 INFO [optim.py:368] (2/8) 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 07:33:58,321 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:34:15,989 INFO [train.py:904] (2/8) Epoch 27, batch 700, loss[loss=0.177, simple_loss=0.2546, pruned_loss=0.04966, over 16701.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2478, pruned_loss=0.0381, over 3222032.79 frames. ], batch size: 134, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:34:34,287 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264616.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:35:22,117 INFO [train.py:904] (2/8) Epoch 27, batch 750, loss[loss=0.1984, simple_loss=0.2748, pruned_loss=0.061, over 16958.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.248, pruned_loss=0.03827, over 3243937.75 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:35:35,422 INFO [optim.py:368] (2/8) 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 07:35:37,454 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264664.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:36:28,312 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6287, 4.5729, 4.5519, 4.0400, 4.6012, 1.8835, 4.3884, 4.1468], device='cuda:2'), covar=tensor([0.0154, 0.0131, 0.0181, 0.0333, 0.0113, 0.2796, 0.0153, 0.0264], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0169, 0.0205, 0.0179, 0.0183, 0.0214, 0.0195, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:36:29,180 INFO [train.py:904] (2/8) Epoch 27, batch 800, loss[loss=0.1432, simple_loss=0.2235, pruned_loss=0.03147, over 16827.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2473, pruned_loss=0.03812, over 3256859.57 frames. ], batch size: 102, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:36:51,846 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2390, 5.2387, 5.1249, 4.5281, 4.7200, 5.1165, 5.1020, 4.7441], device='cuda:2'), covar=tensor([0.0640, 0.0612, 0.0376, 0.0474, 0.1308, 0.0554, 0.0341, 0.0923], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0463, 0.0361, 0.0364, 0.0361, 0.0420, 0.0249, 0.0435], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:37:36,918 INFO [train.py:904] (2/8) Epoch 27, batch 850, loss[loss=0.1743, simple_loss=0.2622, pruned_loss=0.04315, over 16701.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2473, pruned_loss=0.03791, over 3279872.13 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:37:51,806 INFO [optim.py:368] (2/8) 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 07:38:09,470 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3250, 5.2971, 4.9743, 4.5303, 5.1689, 2.0479, 4.9225, 4.8321], device='cuda:2'), covar=tensor([0.0093, 0.0083, 0.0235, 0.0380, 0.0103, 0.2830, 0.0132, 0.0241], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0169, 0.0206, 0.0179, 0.0183, 0.0214, 0.0195, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:38:09,700 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 07:38:21,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8358, 2.9886, 2.5877, 2.8682, 3.2221, 2.9457, 3.4238, 3.4101], device='cuda:2'), covar=tensor([0.0178, 0.0474, 0.0578, 0.0455, 0.0334, 0.0424, 0.0341, 0.0321], device='cuda:2'), in_proj_covar=tensor([0.0228, 0.0245, 0.0234, 0.0234, 0.0245, 0.0244, 0.0243, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:38:41,358 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4434, 3.9456, 4.1896, 2.5127, 3.2154, 3.0408, 3.7889, 4.0543], device='cuda:2'), covar=tensor([0.0411, 0.0910, 0.0538, 0.1989, 0.0975, 0.0876, 0.0919, 0.1099], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0168, 0.0169, 0.0157, 0.0147, 0.0132, 0.0146, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:38:44,873 INFO [train.py:904] (2/8) Epoch 27, batch 900, loss[loss=0.1682, simple_loss=0.2663, pruned_loss=0.03501, over 17041.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2464, pruned_loss=0.03721, over 3286475.12 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:38:47,120 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:28,077 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:51,349 INFO [train.py:904] (2/8) Epoch 27, batch 950, loss[loss=0.1449, simple_loss=0.2359, pruned_loss=0.02688, over 17205.00 frames. ], tot_loss[loss=0.161, simple_loss=0.247, pruned_loss=0.03748, over 3300065.01 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:39:55,587 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264856.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:40:04,148 INFO [optim.py:368] (2/8) 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 07:40:07,724 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264865.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:15,592 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 07:40:41,671 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264890.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:45,357 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0139, 4.1359, 2.9907, 4.8526, 3.4868, 4.7646, 3.0331, 3.5676], device='cuda:2'), covar=tensor([0.0337, 0.0384, 0.1359, 0.0290, 0.0716, 0.0427, 0.1359, 0.0715], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0182, 0.0198, 0.0174, 0.0182, 0.0222, 0.0208, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:40:57,191 INFO [train.py:904] (2/8) Epoch 27, batch 1000, loss[loss=0.16, simple_loss=0.24, pruned_loss=0.03999, over 16559.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2464, pruned_loss=0.03786, over 3300646.20 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:41:43,579 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0245, 4.2385, 4.4612, 2.4861, 4.7296, 4.9219, 3.5758, 3.5589], device='cuda:2'), covar=tensor([0.1062, 0.0208, 0.0251, 0.1270, 0.0099, 0.0182, 0.0436, 0.0563], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0140, 0.0086, 0.0131, 0.0131, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:41:46,840 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:42:07,409 INFO [train.py:904] (2/8) Epoch 27, batch 1050, loss[loss=0.1491, simple_loss=0.2322, pruned_loss=0.03304, over 16672.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2453, pruned_loss=0.03751, over 3311810.69 frames. ], batch size: 37, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:42:19,712 INFO [optim.py:368] (2/8) 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 07:42:38,841 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 07:43:16,444 INFO [train.py:904] (2/8) Epoch 27, batch 1100, loss[loss=0.158, simple_loss=0.2421, pruned_loss=0.03696, over 16512.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2452, pruned_loss=0.03701, over 3316916.82 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:43:43,156 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 07:44:10,287 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6095, 2.7018, 2.4108, 2.4979, 2.9096, 2.6474, 3.1914, 3.1297], device='cuda:2'), covar=tensor([0.0205, 0.0485, 0.0568, 0.0500, 0.0374, 0.0465, 0.0268, 0.0340], device='cuda:2'), in_proj_covar=tensor([0.0230, 0.0247, 0.0236, 0.0235, 0.0247, 0.0246, 0.0245, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:44:25,474 INFO [train.py:904] (2/8) Epoch 27, batch 1150, loss[loss=0.1425, simple_loss=0.2317, pruned_loss=0.0267, over 15994.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2449, pruned_loss=0.03671, over 3312923.55 frames. ], batch size: 35, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:44:39,251 INFO [optim.py:368] (2/8) 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 07:45:34,345 INFO [train.py:904] (2/8) Epoch 27, batch 1200, loss[loss=0.1363, simple_loss=0.2271, pruned_loss=0.02273, over 17216.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2445, pruned_loss=0.03631, over 3313321.79 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:45:52,583 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2550, 3.2936, 3.7350, 2.1680, 3.1327, 2.4430, 3.6593, 3.5853], device='cuda:2'), covar=tensor([0.0258, 0.1101, 0.0562, 0.2127, 0.0848, 0.1038, 0.0648, 0.1172], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0157, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 07:46:00,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3483, 5.3156, 5.2414, 4.6913, 4.8757, 5.1781, 5.1866, 4.8431], device='cuda:2'), covar=tensor([0.0641, 0.0497, 0.0317, 0.0402, 0.1122, 0.0562, 0.0338, 0.0824], device='cuda:2'), in_proj_covar=tensor([0.0319, 0.0474, 0.0369, 0.0372, 0.0368, 0.0430, 0.0254, 0.0444], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:46:04,276 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0806, 2.1497, 2.2956, 3.6957, 2.2213, 2.4306, 2.2518, 2.2632], device='cuda:2'), covar=tensor([0.1619, 0.3977, 0.3409, 0.0817, 0.4158, 0.2913, 0.4090, 0.3617], device='cuda:2'), in_proj_covar=tensor([0.0420, 0.0472, 0.0387, 0.0336, 0.0447, 0.0539, 0.0443, 0.0552], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:46:19,345 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265135.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:46:42,961 INFO [train.py:904] (2/8) Epoch 27, batch 1250, loss[loss=0.1424, simple_loss=0.2371, pruned_loss=0.02391, over 17019.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.2443, pruned_loss=0.03619, over 3316427.40 frames. ], batch size: 41, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:46:48,316 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265156.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:46:53,475 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:46:57,310 INFO [optim.py:368] (2/8) 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 07:47:25,262 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 1300, loss[loss=0.1612, simple_loss=0.2407, pruned_loss=0.04084, over 16910.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.245, pruned_loss=0.03695, over 3318089.08 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:47:54,314 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265204.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:48:11,285 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1210, 5.6423, 5.7918, 5.4856, 5.5665, 6.1642, 5.6432, 5.3379], device='cuda:2'), covar=tensor([0.1021, 0.2193, 0.2611, 0.2163, 0.2717, 0.0973, 0.1618, 0.2361], device='cuda:2'), in_proj_covar=tensor([0.0427, 0.0633, 0.0703, 0.0521, 0.0689, 0.0727, 0.0547, 0.0691], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:49:00,603 INFO [train.py:904] (2/8) Epoch 27, batch 1350, loss[loss=0.1599, simple_loss=0.244, pruned_loss=0.03789, over 16513.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2449, pruned_loss=0.03697, over 3317242.52 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:49:14,442 INFO [optim.py:368] (2/8) 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 07:49:35,225 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 1400, loss[loss=0.1592, simple_loss=0.2375, pruned_loss=0.04048, over 16876.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2452, pruned_loss=0.03713, over 3313684.10 frames. ], batch size: 96, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:50:23,468 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 07:50:56,842 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:51:15,559 INFO [train.py:904] (2/8) Epoch 27, batch 1450, loss[loss=0.1586, simple_loss=0.257, pruned_loss=0.03004, over 17259.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2442, pruned_loss=0.03708, over 3315341.62 frames. ], batch size: 52, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:51:22,608 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 07:51:29,668 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 27, batch 1500, loss[loss=0.1643, simple_loss=0.2552, pruned_loss=0.03668, over 16626.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2452, pruned_loss=0.03671, over 3323233.26 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:52:42,509 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5461, 3.8520, 4.1578, 2.3032, 3.3374, 2.7911, 3.9018, 4.0039], device='cuda:2'), covar=tensor([0.0299, 0.0938, 0.0474, 0.2072, 0.0799, 0.0981, 0.0609, 0.0987], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 07:53:36,975 INFO [train.py:904] (2/8) Epoch 27, batch 1550, loss[loss=0.1714, simple_loss=0.2529, pruned_loss=0.04495, over 16260.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2465, pruned_loss=0.03744, over 3319178.11 frames. ], batch size: 165, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:53:45,902 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265460.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:53:49,832 INFO [optim.py:368] (2/8) 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 07:54:41,907 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1159, 4.8665, 5.1136, 5.2903, 5.5182, 4.8925, 5.4867, 5.5119], device='cuda:2'), covar=tensor([0.1904, 0.1413, 0.1742, 0.0788, 0.0571, 0.0900, 0.0497, 0.0628], device='cuda:2'), in_proj_covar=tensor([0.0690, 0.0849, 0.0981, 0.0860, 0.0652, 0.0678, 0.0714, 0.0830], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 07:54:43,792 INFO [train.py:904] (2/8) Epoch 27, batch 1600, loss[loss=0.1849, simple_loss=0.2611, pruned_loss=0.05432, over 16504.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2479, pruned_loss=0.03823, over 3323591.17 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:54:51,418 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:17,533 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 07:55:19,080 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5389, 5.4807, 5.3996, 4.8045, 5.0208, 5.3940, 5.4006, 5.0221], device='cuda:2'), covar=tensor([0.0580, 0.0462, 0.0328, 0.0370, 0.1141, 0.0452, 0.0239, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0321, 0.0478, 0.0372, 0.0375, 0.0371, 0.0432, 0.0255, 0.0448], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:55:21,305 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:30,769 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2242, 3.3662, 3.6788, 2.1713, 3.0112, 2.4341, 3.6553, 3.6460], device='cuda:2'), covar=tensor([0.0280, 0.0980, 0.0570, 0.2130, 0.0895, 0.1003, 0.0582, 0.1161], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0169, 0.0169, 0.0156, 0.0147, 0.0132, 0.0145, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 07:55:51,085 INFO [train.py:904] (2/8) Epoch 27, batch 1650, loss[loss=0.1631, simple_loss=0.2588, pruned_loss=0.03367, over 17239.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2494, pruned_loss=0.03895, over 3312702.46 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:56:00,241 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-05-02 07:56:04,414 INFO [optim.py:368] (2/8) 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 07:56:43,834 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:56:52,744 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-02 07:57:00,225 INFO [train.py:904] (2/8) Epoch 27, batch 1700, loss[loss=0.1959, simple_loss=0.288, pruned_loss=0.05188, over 16715.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2509, pruned_loss=0.03952, over 3314623.54 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:57:16,018 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:57:39,645 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0965, 5.7062, 5.9073, 5.5317, 5.7075, 6.2565, 5.7351, 5.4554], device='cuda:2'), covar=tensor([0.0944, 0.2099, 0.2510, 0.2139, 0.2451, 0.0961, 0.1622, 0.2188], device='cuda:2'), in_proj_covar=tensor([0.0430, 0.0638, 0.0706, 0.0524, 0.0693, 0.0731, 0.0550, 0.0696], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 07:57:42,663 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265634.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:58:07,539 INFO [train.py:904] (2/8) Epoch 27, batch 1750, loss[loss=0.1491, simple_loss=0.2384, pruned_loss=0.02983, over 15865.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2502, pruned_loss=0.03847, over 3324781.16 frames. ], batch size: 35, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:58:14,387 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 07:58:22,125 INFO [optim.py:368] (2/8) 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 07:58:38,415 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265675.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:59:15,654 INFO [train.py:904] (2/8) Epoch 27, batch 1800, loss[loss=0.2235, simple_loss=0.3005, pruned_loss=0.07322, over 12432.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2516, pruned_loss=0.03879, over 3323992.18 frames. ], batch size: 247, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:59:32,419 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:00:07,669 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 08:00:23,589 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 08:00:23,856 INFO [train.py:904] (2/8) Epoch 27, batch 1850, loss[loss=0.1606, simple_loss=0.2521, pruned_loss=0.03452, over 17236.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2534, pruned_loss=0.03918, over 3321518.26 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:00:37,843 INFO [optim.py:368] (2/8) 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 08:00:55,759 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:01:33,280 INFO [train.py:904] (2/8) Epoch 27, batch 1900, loss[loss=0.1538, simple_loss=0.2335, pruned_loss=0.03708, over 16765.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2535, pruned_loss=0.03881, over 3316411.01 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:42,098 INFO [train.py:904] (2/8) Epoch 27, batch 1950, loss[loss=0.1479, simple_loss=0.2392, pruned_loss=0.02831, over 17226.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2532, pruned_loss=0.03837, over 3317210.55 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:54,907 INFO [optim.py:368] (2/8) 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 08:02:58,412 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2379, 5.9067, 6.0323, 5.6873, 5.8188, 6.3545, 5.8755, 5.5457], device='cuda:2'), covar=tensor([0.0925, 0.1988, 0.2477, 0.2256, 0.2537, 0.0940, 0.1778, 0.2561], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0640, 0.0707, 0.0525, 0.0696, 0.0734, 0.0551, 0.0699], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:03:08,230 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265872.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:26,217 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:50,712 INFO [train.py:904] (2/8) Epoch 27, batch 2000, loss[loss=0.1661, simple_loss=0.2462, pruned_loss=0.04298, over 16828.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2526, pruned_loss=0.03724, over 3322437.07 frames. ], batch size: 116, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:04:10,492 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:33,613 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265933.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:34,865 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:01,212 INFO [train.py:904] (2/8) Epoch 27, batch 2050, loss[loss=0.1498, simple_loss=0.2367, pruned_loss=0.03143, over 17044.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.252, pruned_loss=0.03769, over 3325185.00 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:05:14,204 INFO [optim.py:368] (2/8) 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 08:05:25,438 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265970.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 08:05:36,800 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265982.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:06:14,060 INFO [train.py:904] (2/8) Epoch 27, batch 2100, loss[loss=0.1802, simple_loss=0.2705, pruned_loss=0.04498, over 17126.00 frames. ], tot_loss[loss=0.165, simple_loss=0.253, pruned_loss=0.0385, over 3319806.83 frames. ], batch size: 49, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:07:16,618 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3584, 3.5596, 3.9040, 2.2134, 3.1057, 2.3677, 3.6783, 3.7269], device='cuda:2'), covar=tensor([0.0296, 0.0933, 0.0510, 0.2113, 0.0877, 0.1078, 0.0667, 0.1146], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0169, 0.0169, 0.0156, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 08:07:22,091 INFO [train.py:904] (2/8) Epoch 27, batch 2150, loss[loss=0.1637, simple_loss=0.2566, pruned_loss=0.03536, over 17256.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2539, pruned_loss=0.03888, over 3326263.90 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:07:37,031 INFO [optim.py:368] (2/8) 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 08:07:46,308 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:08:32,965 INFO [train.py:904] (2/8) Epoch 27, batch 2200, loss[loss=0.1795, simple_loss=0.255, pruned_loss=0.05202, over 16857.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.254, pruned_loss=0.03937, over 3320074.64 frames. ], batch size: 102, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:41,323 INFO [train.py:904] (2/8) Epoch 27, batch 2250, loss[loss=0.1686, simple_loss=0.2479, pruned_loss=0.04469, over 16723.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2548, pruned_loss=0.04006, over 3323911.30 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:54,304 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-02 08:09:56,594 INFO [optim.py:368] (2/8) 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 08:10:27,379 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:10:27,774 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 08:10:51,415 INFO [train.py:904] (2/8) Epoch 27, batch 2300, loss[loss=0.1482, simple_loss=0.2487, pruned_loss=0.02387, over 17107.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2543, pruned_loss=0.03965, over 3334317.27 frames. ], batch size: 47, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:11:06,828 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:23,593 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266226.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:11:25,874 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266228.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:34,294 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:59,649 INFO [train.py:904] (2/8) Epoch 27, batch 2350, loss[loss=0.1689, simple_loss=0.2641, pruned_loss=0.03684, over 17138.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2555, pruned_loss=0.04047, over 3331980.20 frames. ], batch size: 49, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:12:14,159 INFO [optim.py:368] (2/8) 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 08:12:18,089 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.24 vs. limit=5.0 2023-05-02 08:12:22,381 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266270.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:12:27,949 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:30,358 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:45,701 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-02 08:12:46,316 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266287.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:13:07,763 INFO [train.py:904] (2/8) Epoch 27, batch 2400, loss[loss=0.1838, simple_loss=0.2639, pruned_loss=0.05188, over 15714.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2562, pruned_loss=0.0406, over 3330709.32 frames. ], batch size: 190, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:13:28,959 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:13:54,417 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6708, 3.7362, 2.4963, 3.9487, 3.0029, 3.9344, 2.5294, 3.1368], device='cuda:2'), covar=tensor([0.0283, 0.0434, 0.1426, 0.0426, 0.0712, 0.0714, 0.1363, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0183, 0.0199, 0.0176, 0.0183, 0.0224, 0.0207, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:14:07,083 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266346.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:14:14,353 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1135, 5.0747, 4.7873, 3.9884, 4.9552, 1.6848, 4.6387, 4.5590], device='cuda:2'), covar=tensor([0.0122, 0.0107, 0.0293, 0.0578, 0.0141, 0.3529, 0.0197, 0.0386], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0175, 0.0211, 0.0185, 0.0188, 0.0219, 0.0202, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:14:16,216 INFO [train.py:904] (2/8) Epoch 27, batch 2450, loss[loss=0.14, simple_loss=0.2325, pruned_loss=0.02381, over 17246.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2564, pruned_loss=0.03993, over 3337001.66 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:14:31,086 INFO [optim.py:368] (2/8) 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 08:14:36,278 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1166, 5.0718, 5.0008, 4.5227, 4.6454, 5.0166, 4.9727, 4.6930], device='cuda:2'), covar=tensor([0.0596, 0.0593, 0.0332, 0.0393, 0.1131, 0.0527, 0.0377, 0.0803], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0478, 0.0372, 0.0377, 0.0372, 0.0431, 0.0255, 0.0448], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:14:41,998 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:15:26,443 INFO [train.py:904] (2/8) Epoch 27, batch 2500, loss[loss=0.1537, simple_loss=0.2422, pruned_loss=0.03263, over 17194.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2562, pruned_loss=0.03956, over 3334488.83 frames. ], batch size: 46, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:15:31,573 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:15:48,870 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:16:01,912 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0371, 3.2206, 3.4137, 2.1287, 2.8464, 2.3168, 3.5217, 3.5339], device='cuda:2'), covar=tensor([0.0317, 0.1017, 0.0716, 0.2110, 0.1023, 0.1152, 0.0651, 0.0986], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0171, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 08:16:35,712 INFO [train.py:904] (2/8) Epoch 27, batch 2550, loss[loss=0.1749, simple_loss=0.267, pruned_loss=0.04136, over 16023.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2558, pruned_loss=0.0391, over 3332277.92 frames. ], batch size: 35, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:16:50,695 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 08:16:51,124 INFO [optim.py:368] (2/8) 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 08:16:51,505 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266464.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:16:58,858 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-05-02 08:17:02,179 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3568, 2.3991, 2.3406, 4.0904, 2.3238, 2.7717, 2.5025, 2.5503], device='cuda:2'), covar=tensor([0.1394, 0.3642, 0.3259, 0.0604, 0.4142, 0.2652, 0.3668, 0.3726], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0475, 0.0388, 0.0339, 0.0447, 0.0544, 0.0446, 0.0555], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:17:41,126 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-02 08:17:45,250 INFO [train.py:904] (2/8) Epoch 27, batch 2600, loss[loss=0.1662, simple_loss=0.2635, pruned_loss=0.03451, over 16426.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2557, pruned_loss=0.03858, over 3331863.13 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:18:17,173 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266525.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:19,481 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8080, 3.3957, 3.8521, 1.9661, 3.9227, 3.9487, 3.1867, 2.9483], device='cuda:2'), covar=tensor([0.0707, 0.0287, 0.0192, 0.1278, 0.0111, 0.0218, 0.0412, 0.0477], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0138, 0.0086, 0.0131, 0.0130, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:18:21,013 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:55,437 INFO [train.py:904] (2/8) Epoch 27, batch 2650, loss[loss=0.1455, simple_loss=0.2433, pruned_loss=0.02388, over 17044.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2556, pruned_loss=0.03801, over 3332354.04 frames. ], batch size: 50, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:19:11,568 INFO [optim.py:368] (2/8) 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 08:19:20,871 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:24,558 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:29,418 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:37,433 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266582.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:20:05,798 INFO [train.py:904] (2/8) Epoch 27, batch 2700, loss[loss=0.1761, simple_loss=0.2536, pruned_loss=0.04935, over 16655.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2559, pruned_loss=0.038, over 3331344.23 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:20:31,907 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:20:44,175 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8544, 2.2057, 2.3847, 3.1536, 2.2172, 2.3394, 2.3843, 2.2905], device='cuda:2'), covar=tensor([0.1549, 0.3427, 0.2784, 0.0805, 0.4155, 0.2538, 0.3172, 0.3484], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0474, 0.0388, 0.0338, 0.0447, 0.0544, 0.0445, 0.0555], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:21:15,403 INFO [train.py:904] (2/8) Epoch 27, batch 2750, loss[loss=0.1561, simple_loss=0.2429, pruned_loss=0.03459, over 16855.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2568, pruned_loss=0.03797, over 3320588.17 frames. ], batch size: 42, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:21:20,099 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.08 vs. limit=2.0 2023-05-02 08:21:29,196 INFO [optim.py:368] (2/8) 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 08:21:51,205 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:21:59,600 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 08:22:14,777 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0221, 4.1197, 2.9349, 4.8993, 3.3147, 4.7737, 2.9157, 3.4865], device='cuda:2'), covar=tensor([0.0326, 0.0380, 0.1520, 0.0249, 0.0815, 0.0463, 0.1518, 0.0767], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0183, 0.0198, 0.0176, 0.0183, 0.0224, 0.0207, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:22:22,656 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:23,526 INFO [train.py:904] (2/8) Epoch 27, batch 2800, loss[loss=0.147, simple_loss=0.2399, pruned_loss=0.02705, over 16989.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2567, pruned_loss=0.03841, over 3315757.79 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:22:25,087 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266704.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:26,239 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1210, 5.5745, 5.7910, 5.4804, 5.6266, 6.1170, 5.5886, 5.3399], device='cuda:2'), covar=tensor([0.0957, 0.1909, 0.2389, 0.2074, 0.2525, 0.0994, 0.1530, 0.2395], device='cuda:2'), in_proj_covar=tensor([0.0438, 0.0648, 0.0716, 0.0532, 0.0705, 0.0743, 0.0556, 0.0707], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:23:15,452 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266740.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:23:28,165 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9594, 2.9274, 2.9221, 5.1094, 4.1296, 4.4119, 1.8949, 3.3335], device='cuda:2'), covar=tensor([0.1296, 0.0786, 0.1082, 0.0175, 0.0191, 0.0377, 0.1515, 0.0699], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0200, 0.0206, 0.0219, 0.0208, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:23:30,545 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2949, 5.6423, 5.4092, 5.4849, 5.1156, 5.1326, 5.0498, 5.8043], device='cuda:2'), covar=tensor([0.1522, 0.1011, 0.1207, 0.0940, 0.1000, 0.0784, 0.1367, 0.0865], device='cuda:2'), in_proj_covar=tensor([0.0727, 0.0880, 0.0718, 0.0678, 0.0558, 0.0552, 0.0744, 0.0689], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:23:31,358 INFO [train.py:904] (2/8) Epoch 27, batch 2850, loss[loss=0.1683, simple_loss=0.2653, pruned_loss=0.0356, over 16554.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2559, pruned_loss=0.03841, over 3311785.93 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:23:48,196 INFO [optim.py:368] (2/8) 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 08:23:49,863 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:23:54,474 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2705, 5.2166, 5.1640, 4.6234, 4.7596, 5.1657, 5.1638, 4.8025], device='cuda:2'), covar=tensor([0.0597, 0.0548, 0.0306, 0.0368, 0.1143, 0.0460, 0.0342, 0.0800], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0479, 0.0373, 0.0378, 0.0374, 0.0433, 0.0255, 0.0449], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:23:58,595 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3738, 2.3458, 2.5538, 4.0891, 2.3498, 2.6948, 2.4394, 2.5392], device='cuda:2'), covar=tensor([0.1608, 0.3906, 0.3083, 0.0664, 0.4073, 0.2655, 0.4029, 0.3205], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0475, 0.0388, 0.0339, 0.0447, 0.0543, 0.0445, 0.0555], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:24:41,244 INFO [train.py:904] (2/8) Epoch 27, batch 2900, loss[loss=0.1699, simple_loss=0.2565, pruned_loss=0.04159, over 16232.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2552, pruned_loss=0.03872, over 3320041.90 frames. ], batch size: 35, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:25:04,249 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:25:49,402 INFO [train.py:904] (2/8) Epoch 27, batch 2950, loss[loss=0.1613, simple_loss=0.2429, pruned_loss=0.03984, over 16864.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2542, pruned_loss=0.03883, over 3315717.36 frames. ], batch size: 96, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:04,411 INFO [optim.py:368] (2/8) 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 08:26:14,851 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:26:29,627 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266882.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:26:58,265 INFO [train.py:904] (2/8) Epoch 27, batch 3000, loss[loss=0.1753, simple_loss=0.2546, pruned_loss=0.04805, over 16694.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2545, pruned_loss=0.03986, over 3304184.18 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:58,265 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 08:27:07,039 INFO [train.py:938] (2/8) Epoch 27, validation: loss=0.1336, simple_loss=0.2386, pruned_loss=0.01429, over 944034.00 frames. 2023-05-02 08:27:07,040 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 08:27:14,893 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 08:27:27,845 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:30,386 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:40,698 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 08:27:45,279 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266930.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 08:28:16,682 INFO [train.py:904] (2/8) Epoch 27, batch 3050, loss[loss=0.155, simple_loss=0.2388, pruned_loss=0.03567, over 16807.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2541, pruned_loss=0.0398, over 3313607.28 frames. ], batch size: 83, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:28:30,545 INFO [optim.py:368] (2/8) 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 08:28:55,840 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266981.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:28:57,091 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7675, 3.8937, 2.9853, 2.3696, 2.5274, 2.5586, 4.0918, 3.3765], device='cuda:2'), covar=tensor([0.2918, 0.0574, 0.1796, 0.3231, 0.2830, 0.2169, 0.0538, 0.1682], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0277, 0.0314, 0.0327, 0.0306, 0.0276, 0.0305, 0.0354], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:29:24,970 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267002.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:29:25,704 INFO [train.py:904] (2/8) Epoch 27, batch 3100, loss[loss=0.1583, simple_loss=0.2492, pruned_loss=0.03375, over 17213.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2532, pruned_loss=0.03966, over 3308653.52 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:30:11,851 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:33,234 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267050.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:36,918 INFO [train.py:904] (2/8) Epoch 27, batch 3150, loss[loss=0.1522, simple_loss=0.2313, pruned_loss=0.03658, over 16713.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2527, pruned_loss=0.03967, over 3304242.28 frames. ], batch size: 89, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:30:45,763 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267060.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:50,834 INFO [optim.py:368] (2/8) 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 08:30:51,195 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4385, 3.4135, 3.4928, 3.5559, 3.6214, 3.3460, 3.5812, 3.6759], device='cuda:2'), covar=tensor([0.1362, 0.1000, 0.1126, 0.0642, 0.0630, 0.2538, 0.1191, 0.0777], device='cuda:2'), in_proj_covar=tensor([0.0699, 0.0859, 0.0997, 0.0868, 0.0659, 0.0691, 0.0720, 0.0841], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:31:07,969 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6111, 5.6313, 5.4828, 5.0211, 5.1372, 5.5357, 5.3896, 5.1785], device='cuda:2'), covar=tensor([0.0560, 0.0463, 0.0289, 0.0321, 0.0953, 0.0433, 0.0325, 0.0772], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0480, 0.0374, 0.0380, 0.0374, 0.0435, 0.0257, 0.0451], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:31:15,347 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:37,647 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267098.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:43,509 INFO [train.py:904] (2/8) Epoch 27, batch 3200, loss[loss=0.1678, simple_loss=0.2511, pruned_loss=0.04226, over 15513.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2522, pruned_loss=0.03919, over 3293750.51 frames. ], batch size: 190, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:31:44,083 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7903, 2.4343, 2.4472, 3.6479, 2.9309, 3.8287, 1.6636, 2.7859], device='cuda:2'), covar=tensor([0.1427, 0.0809, 0.1224, 0.0218, 0.0178, 0.0381, 0.1656, 0.0881], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0180, 0.0199, 0.0201, 0.0206, 0.0219, 0.0208, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:32:08,063 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267120.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:35,276 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 08:32:39,107 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:50,800 INFO [train.py:904] (2/8) Epoch 27, batch 3250, loss[loss=0.168, simple_loss=0.2643, pruned_loss=0.03587, over 17063.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2523, pruned_loss=0.03893, over 3294591.17 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:33:01,396 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:07,327 INFO [optim.py:368] (2/8) 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 08:33:12,829 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267168.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:28,370 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-05-02 08:34:00,064 INFO [train.py:904] (2/8) Epoch 27, batch 3300, loss[loss=0.1689, simple_loss=0.2649, pruned_loss=0.03641, over 17051.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2538, pruned_loss=0.03961, over 3301402.88 frames. ], batch size: 53, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:34:43,934 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-05-02 08:35:07,437 INFO [train.py:904] (2/8) Epoch 27, batch 3350, loss[loss=0.1552, simple_loss=0.2548, pruned_loss=0.02779, over 17115.00 frames. ], tot_loss[loss=0.167, simple_loss=0.255, pruned_loss=0.03948, over 3302978.65 frames. ], batch size: 48, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:35:22,678 INFO [optim.py:368] (2/8) 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 08:35:31,672 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4242, 4.4197, 4.5732, 4.3897, 4.4963, 5.0138, 4.5285, 4.2377], device='cuda:2'), covar=tensor([0.1826, 0.2182, 0.2386, 0.2263, 0.2577, 0.1211, 0.1607, 0.2546], device='cuda:2'), in_proj_covar=tensor([0.0437, 0.0647, 0.0714, 0.0528, 0.0706, 0.0742, 0.0554, 0.0704], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:35:38,147 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:35:58,699 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9861, 3.8556, 4.3838, 2.4046, 4.4546, 4.6815, 3.3832, 3.6491], device='cuda:2'), covar=tensor([0.0778, 0.0257, 0.0263, 0.1092, 0.0092, 0.0167, 0.0454, 0.0380], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0140, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:36:14,882 INFO [train.py:904] (2/8) Epoch 27, batch 3400, loss[loss=0.1635, simple_loss=0.262, pruned_loss=0.03251, over 16659.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2549, pruned_loss=0.03934, over 3310090.87 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:36:44,217 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 08:36:58,128 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267335.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:02,336 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7891, 4.7027, 4.6385, 4.0484, 4.6922, 1.9165, 4.3981, 4.2834], device='cuda:2'), covar=tensor([0.0159, 0.0129, 0.0210, 0.0401, 0.0128, 0.2988, 0.0182, 0.0254], device='cuda:2'), in_proj_covar=tensor([0.0183, 0.0177, 0.0215, 0.0188, 0.0192, 0.0221, 0.0205, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:37:14,519 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 08:37:21,859 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0144, 5.3847, 5.1564, 5.1793, 4.8745, 4.8453, 4.8170, 5.4589], device='cuda:2'), covar=tensor([0.1417, 0.0860, 0.1011, 0.0888, 0.0891, 0.1039, 0.1304, 0.0943], device='cuda:2'), in_proj_covar=tensor([0.0736, 0.0890, 0.0726, 0.0685, 0.0564, 0.0558, 0.0751, 0.0696], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:37:22,625 INFO [train.py:904] (2/8) Epoch 27, batch 3450, loss[loss=0.1642, simple_loss=0.2445, pruned_loss=0.04191, over 16837.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2538, pruned_loss=0.03894, over 3307773.36 frames. ], batch size: 90, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:37:31,757 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267360.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:36,673 INFO [optim.py:368] (2/8) 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 08:38:03,487 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:38:20,978 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-05-02 08:38:30,913 INFO [train.py:904] (2/8) Epoch 27, batch 3500, loss[loss=0.1481, simple_loss=0.2364, pruned_loss=0.0299, over 17197.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.252, pruned_loss=0.03833, over 3306519.14 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:38:39,646 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267408.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:19,824 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 3550, loss[loss=0.1512, simple_loss=0.2465, pruned_loss=0.02795, over 17184.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.251, pruned_loss=0.03781, over 3310857.34 frames. ], batch size: 46, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:39:41,926 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:53,961 INFO [optim.py:368] (2/8) 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 08:40:36,667 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:40:48,162 INFO [train.py:904] (2/8) Epoch 27, batch 3600, loss[loss=0.1715, simple_loss=0.2533, pruned_loss=0.04483, over 11730.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2501, pruned_loss=0.03782, over 3291271.88 frames. ], batch size: 248, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:00,665 INFO [train.py:904] (2/8) Epoch 27, batch 3650, loss[loss=0.1627, simple_loss=0.2332, pruned_loss=0.04608, over 16910.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2487, pruned_loss=0.03798, over 3293378.73 frames. ], batch size: 109, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:03,416 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:42:05,393 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6458, 2.5755, 2.0135, 2.7348, 2.2020, 2.7968, 2.2111, 2.3981], device='cuda:2'), covar=tensor([0.0317, 0.0379, 0.1325, 0.0274, 0.0687, 0.0396, 0.1218, 0.0691], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0184, 0.0199, 0.0177, 0.0182, 0.0224, 0.0206, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:42:07,885 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1015, 2.3817, 2.5414, 1.9674, 2.6678, 2.6940, 2.4272, 2.4260], device='cuda:2'), covar=tensor([0.0793, 0.0326, 0.0267, 0.0963, 0.0142, 0.0301, 0.0572, 0.0438], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0112, 0.0102, 0.0139, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:42:16,686 INFO [optim.py:368] (2/8) 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 08:42:30,719 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0792, 2.2430, 2.3492, 3.6472, 2.2698, 2.5106, 2.3418, 2.3403], device='cuda:2'), covar=tensor([0.1545, 0.3543, 0.3112, 0.0737, 0.3880, 0.2582, 0.3707, 0.3370], device='cuda:2'), in_proj_covar=tensor([0.0424, 0.0476, 0.0389, 0.0341, 0.0449, 0.0545, 0.0447, 0.0557], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:42:36,626 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:42:56,920 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9177, 3.6685, 3.9939, 2.3131, 4.1388, 4.1724, 3.4134, 3.2734], device='cuda:2'), covar=tensor([0.0761, 0.0280, 0.0239, 0.1128, 0.0111, 0.0202, 0.0366, 0.0448], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:43:14,274 INFO [train.py:904] (2/8) Epoch 27, batch 3700, loss[loss=0.1647, simple_loss=0.2379, pruned_loss=0.04569, over 16890.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2473, pruned_loss=0.03909, over 3269593.97 frames. ], batch size: 96, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:43:45,314 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267624.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:44:27,013 INFO [train.py:904] (2/8) Epoch 27, batch 3750, loss[loss=0.1706, simple_loss=0.2456, pruned_loss=0.04777, over 16677.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.248, pruned_loss=0.04087, over 3262328.38 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:44:42,769 INFO [optim.py:368] (2/8) 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 08:44:45,006 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0951, 5.4667, 5.2392, 5.2146, 4.9692, 4.8803, 4.8801, 5.5470], device='cuda:2'), covar=tensor([0.1385, 0.0885, 0.1083, 0.0944, 0.0802, 0.0968, 0.1282, 0.0921], device='cuda:2'), in_proj_covar=tensor([0.0733, 0.0886, 0.0721, 0.0683, 0.0562, 0.0557, 0.0749, 0.0696], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:44:48,600 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 08:44:49,703 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0923, 3.2859, 3.1462, 1.9068, 2.6942, 2.1521, 3.5524, 3.7197], device='cuda:2'), covar=tensor([0.0225, 0.0793, 0.0731, 0.2356, 0.1055, 0.1160, 0.0528, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0158, 0.0149, 0.0133, 0.0148, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 08:45:12,593 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0351, 5.0385, 4.9596, 4.5795, 4.6394, 4.9943, 4.7749, 4.7352], device='cuda:2'), covar=tensor([0.0680, 0.0666, 0.0325, 0.0326, 0.0928, 0.0438, 0.0475, 0.0644], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0482, 0.0374, 0.0380, 0.0375, 0.0436, 0.0256, 0.0451], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:45:40,198 INFO [train.py:904] (2/8) Epoch 27, batch 3800, loss[loss=0.1777, simple_loss=0.2619, pruned_loss=0.04678, over 17016.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2492, pruned_loss=0.04198, over 3267533.46 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:31,883 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:46:41,549 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-05-02 08:46:52,904 INFO [train.py:904] (2/8) Epoch 27, batch 3850, loss[loss=0.1695, simple_loss=0.2425, pruned_loss=0.04823, over 16784.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2495, pruned_loss=0.04257, over 3268806.41 frames. ], batch size: 83, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:54,250 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:47:08,875 INFO [optim.py:368] (2/8) 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 08:47:39,702 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:48:02,755 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:48:03,541 INFO [train.py:904] (2/8) Epoch 27, batch 3900, loss[loss=0.1913, simple_loss=0.2606, pruned_loss=0.06101, over 16915.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2495, pruned_loss=0.04345, over 3273422.58 frames. ], batch size: 109, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:48:10,168 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6687, 3.8069, 3.9489, 2.8218, 3.6162, 4.0311, 3.6550, 2.4077], device='cuda:2'), covar=tensor([0.0463, 0.0189, 0.0063, 0.0361, 0.0113, 0.0117, 0.0112, 0.0469], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0090, 0.0091, 0.0137, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 08:49:11,681 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267850.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:49:14,859 INFO [train.py:904] (2/8) Epoch 27, batch 3950, loss[loss=0.1468, simple_loss=0.2319, pruned_loss=0.03082, over 17065.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2486, pruned_loss=0.04366, over 3272679.52 frames. ], batch size: 50, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:49:32,325 INFO [optim.py:368] (2/8) 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 08:50:28,770 INFO [train.py:904] (2/8) Epoch 27, batch 4000, loss[loss=0.1892, simple_loss=0.2648, pruned_loss=0.0568, over 16822.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2492, pruned_loss=0.04432, over 3275219.96 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:50:58,067 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3141, 5.9783, 6.1802, 5.7661, 5.9203, 6.4066, 5.9118, 5.6236], device='cuda:2'), covar=tensor([0.0823, 0.1598, 0.1651, 0.1740, 0.2071, 0.0790, 0.1501, 0.2238], device='cuda:2'), in_proj_covar=tensor([0.0435, 0.0644, 0.0707, 0.0525, 0.0701, 0.0734, 0.0551, 0.0700], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:51:42,546 INFO [train.py:904] (2/8) Epoch 27, batch 4050, loss[loss=0.1727, simple_loss=0.2565, pruned_loss=0.0445, over 16701.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2499, pruned_loss=0.04358, over 3273850.67 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:58,298 INFO [optim.py:368] (2/8) 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 08:52:59,882 INFO [train.py:904] (2/8) Epoch 27, batch 4100, loss[loss=0.1834, simple_loss=0.2693, pruned_loss=0.04879, over 16463.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.252, pruned_loss=0.04342, over 3257553.64 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:54:16,165 INFO [train.py:904] (2/8) Epoch 27, batch 4150, loss[loss=0.1915, simple_loss=0.2738, pruned_loss=0.05458, over 16862.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2584, pruned_loss=0.0449, over 3244128.04 frames. ], batch size: 42, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:54:33,170 INFO [optim.py:368] (2/8) 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 08:55:02,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8130, 5.1547, 5.3521, 4.9745, 5.0637, 5.6946, 5.1276, 4.7940], device='cuda:2'), covar=tensor([0.1029, 0.1754, 0.1707, 0.1940, 0.2430, 0.0862, 0.1474, 0.2283], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0639, 0.0702, 0.0521, 0.0696, 0.0730, 0.0546, 0.0696], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 08:55:16,389 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7226, 5.9702, 5.7107, 5.8631, 5.5118, 5.1630, 5.4782, 6.1329], device='cuda:2'), covar=tensor([0.1201, 0.0850, 0.1226, 0.0821, 0.0823, 0.0730, 0.1219, 0.0872], device='cuda:2'), in_proj_covar=tensor([0.0724, 0.0878, 0.0714, 0.0675, 0.0558, 0.0552, 0.0741, 0.0689], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:55:32,119 INFO [train.py:904] (2/8) Epoch 27, batch 4200, loss[loss=0.1966, simple_loss=0.297, pruned_loss=0.04806, over 16673.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2653, pruned_loss=0.04635, over 3236622.71 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:55:47,129 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 08:56:40,947 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268150.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:56:44,662 INFO [train.py:904] (2/8) Epoch 27, batch 4250, loss[loss=0.1736, simple_loss=0.2661, pruned_loss=0.0405, over 16864.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2688, pruned_loss=0.04617, over 3220162.00 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:57:00,873 INFO [optim.py:368] (2/8) 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 08:57:05,051 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8720, 2.2082, 2.4201, 3.1607, 2.2174, 2.3439, 2.3543, 2.3314], device='cuda:2'), covar=tensor([0.1523, 0.3644, 0.2716, 0.0784, 0.4362, 0.2803, 0.3540, 0.3580], device='cuda:2'), in_proj_covar=tensor([0.0421, 0.0474, 0.0386, 0.0338, 0.0446, 0.0542, 0.0444, 0.0555], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:57:38,455 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3935, 4.2883, 4.4357, 4.5751, 4.7338, 4.2793, 4.6765, 4.7718], device='cuda:2'), covar=tensor([0.1739, 0.1270, 0.1518, 0.0724, 0.0520, 0.1137, 0.0776, 0.0665], device='cuda:2'), in_proj_covar=tensor([0.0680, 0.0837, 0.0972, 0.0849, 0.0645, 0.0676, 0.0703, 0.0819], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 08:57:50,362 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:57:57,372 INFO [train.py:904] (2/8) Epoch 27, batch 4300, loss[loss=0.1922, simple_loss=0.3009, pruned_loss=0.0418, over 16820.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2699, pruned_loss=0.04535, over 3218104.04 frames. ], batch size: 102, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:58:10,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1649, 2.8315, 3.0504, 1.7758, 3.1495, 3.1766, 2.6533, 2.5727], device='cuda:2'), covar=tensor([0.0900, 0.0336, 0.0254, 0.1242, 0.0126, 0.0218, 0.0537, 0.0524], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0101, 0.0140, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 08:58:10,906 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268212.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:59:00,262 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8573, 2.7649, 2.6583, 1.9003, 2.6023, 2.6946, 2.5826, 1.9492], device='cuda:2'), covar=tensor([0.0473, 0.0087, 0.0089, 0.0383, 0.0145, 0.0137, 0.0148, 0.0402], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0090, 0.0092, 0.0137, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 08:59:12,217 INFO [train.py:904] (2/8) Epoch 27, batch 4350, loss[loss=0.1893, simple_loss=0.2834, pruned_loss=0.04763, over 16314.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2732, pruned_loss=0.04644, over 3223216.49 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:59:27,930 INFO [optim.py:368] (2/8) 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 08:59:43,169 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268273.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 08:59:45,551 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:00:26,994 INFO [train.py:904] (2/8) Epoch 27, batch 4400, loss[loss=0.1792, simple_loss=0.2729, pruned_loss=0.0428, over 15382.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2751, pruned_loss=0.0477, over 3198895.44 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:02,264 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8718, 2.1803, 2.2426, 3.3758, 2.0518, 2.4141, 2.2759, 2.2467], device='cuda:2'), covar=tensor([0.1531, 0.3339, 0.2930, 0.0667, 0.4257, 0.2394, 0.3396, 0.3483], device='cuda:2'), in_proj_covar=tensor([0.0421, 0.0475, 0.0386, 0.0337, 0.0447, 0.0543, 0.0444, 0.0555], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:01:11,124 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6909, 6.0520, 5.6093, 6.0830, 5.5267, 5.0770, 5.6001, 6.2025], device='cuda:2'), covar=tensor([0.2275, 0.1214, 0.2126, 0.1209, 0.1535, 0.1156, 0.2160, 0.1520], device='cuda:2'), in_proj_covar=tensor([0.0717, 0.0870, 0.0707, 0.0670, 0.0552, 0.0547, 0.0731, 0.0683], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:01:14,951 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268336.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:01:37,962 INFO [train.py:904] (2/8) Epoch 27, batch 4450, loss[loss=0.1883, simple_loss=0.279, pruned_loss=0.04885, over 16695.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2787, pruned_loss=0.04938, over 3199292.16 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:40,366 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3090, 4.2917, 4.1169, 3.3333, 4.1807, 1.7028, 3.9822, 3.5483], device='cuda:2'), covar=tensor([0.0077, 0.0074, 0.0173, 0.0310, 0.0065, 0.3300, 0.0117, 0.0311], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0175, 0.0212, 0.0187, 0.0189, 0.0219, 0.0202, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:01:55,123 INFO [optim.py:368] (2/8) 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 09:02:50,819 INFO [train.py:904] (2/8) Epoch 27, batch 4500, loss[loss=0.1926, simple_loss=0.2774, pruned_loss=0.05388, over 16758.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2794, pruned_loss=0.05022, over 3213795.15 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:03:07,145 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 09:03:13,451 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7420, 3.4174, 3.8978, 2.0217, 4.0545, 4.0657, 3.1128, 3.2152], device='cuda:2'), covar=tensor([0.0748, 0.0324, 0.0215, 0.1122, 0.0086, 0.0131, 0.0463, 0.0397], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0111, 0.0101, 0.0140, 0.0087, 0.0131, 0.0131, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 09:03:36,836 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 09:04:03,668 INFO [train.py:904] (2/8) Epoch 27, batch 4550, loss[loss=0.2152, simple_loss=0.2955, pruned_loss=0.06744, over 16780.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2803, pruned_loss=0.0514, over 3202907.67 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:04:15,858 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 09:04:20,725 INFO [optim.py:368] (2/8) 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 09:04:55,487 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9446, 3.5768, 4.1143, 2.1015, 4.3030, 4.2963, 3.2184, 3.3313], device='cuda:2'), covar=tensor([0.0800, 0.0344, 0.0224, 0.1234, 0.0088, 0.0138, 0.0467, 0.0457], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0140, 0.0087, 0.0132, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 09:05:15,491 INFO [train.py:904] (2/8) Epoch 27, batch 4600, loss[loss=0.1957, simple_loss=0.2815, pruned_loss=0.05494, over 16720.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2814, pruned_loss=0.05127, over 3213966.60 frames. ], batch size: 89, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:05:40,405 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 09:06:01,309 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0845, 3.7438, 3.6800, 2.2954, 3.3871, 3.7061, 3.3794, 2.1074], device='cuda:2'), covar=tensor([0.0638, 0.0047, 0.0059, 0.0485, 0.0105, 0.0100, 0.0114, 0.0509], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0089, 0.0091, 0.0136, 0.0102, 0.0115, 0.0098, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 09:06:23,484 INFO [train.py:904] (2/8) Epoch 27, batch 4650, loss[loss=0.1724, simple_loss=0.2605, pruned_loss=0.04212, over 16648.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2809, pruned_loss=0.05177, over 3229761.24 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:06:40,830 INFO [optim.py:368] (2/8) 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 09:06:45,257 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268568.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:07:32,917 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:07:34,884 INFO [train.py:904] (2/8) Epoch 27, batch 4700, loss[loss=0.1759, simple_loss=0.2632, pruned_loss=0.04434, over 16436.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2781, pruned_loss=0.05051, over 3218871.37 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:08:15,777 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268631.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:08:34,300 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-05-02 09:08:45,864 INFO [train.py:904] (2/8) Epoch 27, batch 4750, loss[loss=0.1576, simple_loss=0.2472, pruned_loss=0.03397, over 17002.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2741, pruned_loss=0.04843, over 3219774.38 frames. ], batch size: 50, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:08:59,557 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 09:09:59,314 INFO [train.py:904] (2/8) Epoch 27, batch 4800, loss[loss=0.1823, simple_loss=0.2794, pruned_loss=0.04259, over 16638.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2701, pruned_loss=0.04636, over 3214866.71 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:14,169 INFO [train.py:904] (2/8) Epoch 27, batch 4850, loss[loss=0.1914, simple_loss=0.2779, pruned_loss=0.05246, over 12309.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2706, pruned_loss=0.04559, over 3193992.65 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:31,500 INFO [optim.py:368] (2/8) 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 09:11:50,078 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5917, 3.6719, 3.4239, 3.1156, 3.2817, 3.5200, 3.3669, 3.3909], device='cuda:2'), covar=tensor([0.0570, 0.0559, 0.0280, 0.0259, 0.0517, 0.0447, 0.1313, 0.0429], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0462, 0.0359, 0.0365, 0.0360, 0.0418, 0.0247, 0.0430], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:12:27,703 INFO [train.py:904] (2/8) Epoch 27, batch 4900, loss[loss=0.1528, simple_loss=0.2433, pruned_loss=0.03122, over 16540.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2704, pruned_loss=0.04518, over 3164831.85 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:12,339 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 09:13:17,735 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5316, 3.6709, 2.2326, 4.1901, 2.7690, 4.1017, 2.4133, 2.9217], device='cuda:2'), covar=tensor([0.0327, 0.0408, 0.1676, 0.0177, 0.0917, 0.0637, 0.1530, 0.0908], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0180, 0.0195, 0.0171, 0.0179, 0.0219, 0.0202, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 09:13:37,966 INFO [train.py:904] (2/8) Epoch 27, batch 4950, loss[loss=0.1919, simple_loss=0.2817, pruned_loss=0.05106, over 11938.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2694, pruned_loss=0.04428, over 3174345.40 frames. ], batch size: 248, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:54,419 INFO [optim.py:368] (2/8) 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 09:13:58,982 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268868.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:14:36,335 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:14:50,663 INFO [train.py:904] (2/8) Epoch 27, batch 5000, loss[loss=0.1675, simple_loss=0.2635, pruned_loss=0.03573, over 16782.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2707, pruned_loss=0.044, over 3197378.21 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:15:09,313 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268916.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:15:29,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0354, 5.4266, 5.6368, 5.3327, 5.4589, 5.9781, 5.4041, 5.0788], device='cuda:2'), covar=tensor([0.0899, 0.1602, 0.1983, 0.1811, 0.2321, 0.0925, 0.1377, 0.2178], device='cuda:2'), in_proj_covar=tensor([0.0426, 0.0629, 0.0688, 0.0514, 0.0686, 0.0719, 0.0538, 0.0684], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 09:15:29,480 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:00,154 INFO [train.py:904] (2/8) Epoch 27, batch 5050, loss[loss=0.1838, simple_loss=0.2757, pruned_loss=0.04588, over 16916.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2715, pruned_loss=0.04376, over 3204802.71 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:16:03,776 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:05,781 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 09:16:37,071 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:54,197 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268991.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:17:11,192 INFO [train.py:904] (2/8) Epoch 27, batch 5100, loss[loss=0.1591, simple_loss=0.2481, pruned_loss=0.03511, over 16711.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2696, pruned_loss=0.04306, over 3222914.93 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:17:18,479 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4329, 4.6999, 4.5044, 4.5661, 4.2824, 4.1916, 4.2135, 4.7340], device='cuda:2'), covar=tensor([0.1257, 0.0846, 0.0952, 0.0758, 0.0816, 0.1606, 0.1110, 0.0888], device='cuda:2'), in_proj_covar=tensor([0.0708, 0.0860, 0.0701, 0.0660, 0.0545, 0.0542, 0.0724, 0.0675], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:17:31,881 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 09:18:22,289 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:18:23,595 INFO [train.py:904] (2/8) Epoch 27, batch 5150, loss[loss=0.1743, simple_loss=0.2756, pruned_loss=0.03654, over 16385.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2697, pruned_loss=0.0428, over 3198966.93 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:18:41,483 INFO [optim.py:368] (2/8) 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 09:19:36,079 INFO [train.py:904] (2/8) Epoch 27, batch 5200, loss[loss=0.1563, simple_loss=0.2504, pruned_loss=0.03111, over 16896.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.269, pruned_loss=0.0424, over 3187713.82 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:20:47,210 INFO [train.py:904] (2/8) Epoch 27, batch 5250, loss[loss=0.1765, simple_loss=0.2693, pruned_loss=0.04184, over 16903.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2664, pruned_loss=0.04198, over 3205181.44 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:21:04,386 INFO [optim.py:368] (2/8) 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 09:21:19,799 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4369, 4.7041, 4.5536, 4.5554, 4.3210, 4.2559, 4.2391, 4.7559], device='cuda:2'), covar=tensor([0.1217, 0.0840, 0.0841, 0.0685, 0.0723, 0.1426, 0.1058, 0.0850], device='cuda:2'), in_proj_covar=tensor([0.0711, 0.0862, 0.0702, 0.0661, 0.0547, 0.0543, 0.0726, 0.0677], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:21:28,264 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0979, 3.8780, 3.8263, 2.3449, 3.3787, 3.8375, 3.4011, 2.0543], device='cuda:2'), covar=tensor([0.0665, 0.0049, 0.0054, 0.0458, 0.0114, 0.0113, 0.0124, 0.0552], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0089, 0.0090, 0.0136, 0.0101, 0.0114, 0.0098, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 09:22:00,558 INFO [train.py:904] (2/8) Epoch 27, batch 5300, loss[loss=0.1589, simple_loss=0.2507, pruned_loss=0.03358, over 16813.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2626, pruned_loss=0.04073, over 3214827.11 frames. ], batch size: 102, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:22:23,019 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6300, 3.8362, 2.8180, 2.3800, 2.4644, 2.5307, 4.0654, 3.3422], device='cuda:2'), covar=tensor([0.2973, 0.0624, 0.1991, 0.2747, 0.2717, 0.2073, 0.0446, 0.1324], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0326, 0.0305, 0.0274, 0.0304, 0.0351], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 09:22:55,457 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:08,330 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269250.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:12,961 INFO [train.py:904] (2/8) Epoch 27, batch 5350, loss[loss=0.1716, simple_loss=0.2656, pruned_loss=0.03883, over 16828.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.261, pruned_loss=0.03988, over 3216643.20 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:23:18,806 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269257.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:29,777 INFO [optim.py:368] (2/8) 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 09:23:33,788 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5565, 4.6160, 4.4491, 4.1210, 4.1455, 4.5340, 4.2901, 4.3166], device='cuda:2'), covar=tensor([0.0633, 0.0663, 0.0309, 0.0313, 0.0869, 0.0578, 0.0558, 0.0596], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0464, 0.0360, 0.0365, 0.0360, 0.0420, 0.0246, 0.0430], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:23:43,207 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8658, 3.4179, 3.4413, 2.1208, 3.0858, 3.3745, 3.1338, 1.9619], device='cuda:2'), covar=tensor([0.0686, 0.0061, 0.0063, 0.0480, 0.0121, 0.0121, 0.0133, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0135, 0.0101, 0.0114, 0.0098, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 09:24:22,297 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 09:24:24,489 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:24:25,135 INFO [train.py:904] (2/8) Epoch 27, batch 5400, loss[loss=0.1886, simple_loss=0.2786, pruned_loss=0.04932, over 16682.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2638, pruned_loss=0.04043, over 3217512.37 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:24:28,351 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:05,529 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-05-02 09:25:31,308 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:40,816 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:42,376 INFO [train.py:904] (2/8) Epoch 27, batch 5450, loss[loss=0.2059, simple_loss=0.2954, pruned_loss=0.0582, over 16506.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2663, pruned_loss=0.04141, over 3211236.15 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:26:01,151 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 27, batch 5500, loss[loss=0.2033, simple_loss=0.293, pruned_loss=0.05678, over 16376.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2731, pruned_loss=0.04503, over 3209481.82 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:27:06,091 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0012, 3.8933, 4.0481, 4.1663, 4.2774, 3.8935, 4.2248, 4.2873], device='cuda:2'), covar=tensor([0.1665, 0.1124, 0.1351, 0.0676, 0.0601, 0.1495, 0.0886, 0.0764], device='cuda:2'), in_proj_covar=tensor([0.0665, 0.0814, 0.0947, 0.0827, 0.0627, 0.0659, 0.0685, 0.0800], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:27:16,660 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:00,079 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:19,422 INFO [train.py:904] (2/8) Epoch 27, batch 5550, loss[loss=0.2546, simple_loss=0.3179, pruned_loss=0.09566, over 11381.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2804, pruned_loss=0.05043, over 3172126.90 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:28:26,893 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 09:28:38,497 INFO [optim.py:368] (2/8) 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 09:29:37,448 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269502.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:29:38,097 INFO [train.py:904] (2/8) Epoch 27, batch 5600, loss[loss=0.2067, simple_loss=0.2962, pruned_loss=0.05866, over 16831.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.286, pruned_loss=0.05529, over 3098447.20 frames. ], batch size: 90, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:30:22,495 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2737, 2.0235, 1.6884, 1.7535, 2.3109, 1.9584, 2.0226, 2.4252], device='cuda:2'), covar=tensor([0.0230, 0.0417, 0.0549, 0.0476, 0.0267, 0.0383, 0.0213, 0.0261], device='cuda:2'), in_proj_covar=tensor([0.0226, 0.0241, 0.0232, 0.0232, 0.0243, 0.0241, 0.0240, 0.0241], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:30:56,806 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 5650, loss[loss=0.2137, simple_loss=0.3015, pruned_loss=0.06297, over 16911.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2892, pruned_loss=0.05757, over 3088652.34 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:31:20,024 INFO [optim.py:368] (2/8) 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 09:32:11,860 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:13,054 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:18,709 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 09:32:20,977 INFO [train.py:904] (2/8) Epoch 27, batch 5700, loss[loss=0.1987, simple_loss=0.291, pruned_loss=0.05321, over 16774.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2908, pruned_loss=0.05941, over 3078362.26 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:11,507 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 09:33:31,907 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269647.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:33:41,282 INFO [train.py:904] (2/8) Epoch 27, batch 5750, loss[loss=0.1957, simple_loss=0.2822, pruned_loss=0.05458, over 16636.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2933, pruned_loss=0.06023, over 3086018.15 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:59,181 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:35:02,094 INFO [train.py:904] (2/8) Epoch 27, batch 5800, loss[loss=0.1874, simple_loss=0.2817, pruned_loss=0.0465, over 16566.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2937, pruned_loss=0.06015, over 3049905.79 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:35:11,563 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269708.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:36:01,582 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1266, 2.2692, 2.3386, 3.7635, 2.1537, 2.5994, 2.3382, 2.4252], device='cuda:2'), covar=tensor([0.1439, 0.3565, 0.3000, 0.0621, 0.4176, 0.2496, 0.3606, 0.3422], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0468, 0.0379, 0.0332, 0.0441, 0.0535, 0.0438, 0.0547], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:36:21,269 INFO [train.py:904] (2/8) Epoch 27, batch 5850, loss[loss=0.2128, simple_loss=0.3058, pruned_loss=0.05985, over 16860.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2914, pruned_loss=0.05848, over 3062986.58 frames. ], batch size: 102, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:36:40,940 INFO [optim.py:368] (2/8) 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 09:36:52,889 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:37:34,889 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269797.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:37:45,794 INFO [train.py:904] (2/8) Epoch 27, batch 5900, loss[loss=0.1878, simple_loss=0.2818, pruned_loss=0.04692, over 16721.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2911, pruned_loss=0.05837, over 3069613.61 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:38:38,102 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269834.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:38:40,704 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:38:44,266 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6435, 2.6145, 1.8662, 2.7554, 2.0787, 2.7981, 2.1120, 2.3686], device='cuda:2'), covar=tensor([0.0333, 0.0384, 0.1344, 0.0353, 0.0789, 0.0522, 0.1272, 0.0699], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0171, 0.0180, 0.0219, 0.0203, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 09:39:06,822 INFO [train.py:904] (2/8) Epoch 27, batch 5950, loss[loss=0.2073, simple_loss=0.298, pruned_loss=0.05824, over 17064.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2918, pruned_loss=0.05755, over 3063201.38 frames. ], batch size: 55, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:39:08,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7435, 1.8068, 1.6385, 1.4883, 1.9672, 1.5557, 1.5598, 1.9210], device='cuda:2'), covar=tensor([0.0244, 0.0324, 0.0463, 0.0396, 0.0248, 0.0329, 0.0201, 0.0248], device='cuda:2'), in_proj_covar=tensor([0.0226, 0.0240, 0.0231, 0.0231, 0.0242, 0.0241, 0.0239, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:39:27,598 INFO [optim.py:368] (2/8) 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 09:40:17,419 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:17,576 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:24,914 INFO [train.py:904] (2/8) Epoch 27, batch 6000, loss[loss=0.1653, simple_loss=0.2481, pruned_loss=0.04126, over 16644.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2908, pruned_loss=0.05728, over 3050705.55 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:40:24,914 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 09:40:35,098 INFO [train.py:938] (2/8) Epoch 27, validation: loss=0.148, simple_loss=0.2603, pruned_loss=0.01783, over 944034.00 frames. 2023-05-02 09:40:35,099 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 09:40:35,651 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3599, 2.9703, 2.6858, 2.3289, 2.2909, 2.3343, 2.9780, 2.8856], device='cuda:2'), covar=tensor([0.2533, 0.0720, 0.1656, 0.2470, 0.2347, 0.2181, 0.0531, 0.1389], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0327, 0.0305, 0.0275, 0.0304, 0.0351], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 09:41:37,530 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:41:52,126 INFO [train.py:904] (2/8) Epoch 27, batch 6050, loss[loss=0.1985, simple_loss=0.3046, pruned_loss=0.04624, over 16607.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2902, pruned_loss=0.057, over 3072258.78 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:42:12,241 INFO [optim.py:368] (2/8) 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 09:42:13,590 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 09:42:26,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5393, 4.6196, 4.9240, 4.8975, 4.9029, 4.6354, 4.5894, 4.5224], device='cuda:2'), covar=tensor([0.0381, 0.0667, 0.0456, 0.0420, 0.0509, 0.0504, 0.0956, 0.0566], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0481, 0.0467, 0.0428, 0.0515, 0.0492, 0.0568, 0.0395], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 09:43:13,684 INFO [train.py:904] (2/8) Epoch 27, batch 6100, loss[loss=0.1933, simple_loss=0.2882, pruned_loss=0.04923, over 16904.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2893, pruned_loss=0.05564, over 3087347.17 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:43:22,435 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270008.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:16,078 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5874, 2.5365, 1.9047, 2.6385, 2.1449, 2.7498, 2.1661, 2.3576], device='cuda:2'), covar=tensor([0.0327, 0.0406, 0.1308, 0.0269, 0.0652, 0.0540, 0.1149, 0.0654], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0171, 0.0180, 0.0219, 0.0203, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 09:44:30,335 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7110, 4.9528, 5.1189, 4.9029, 4.9862, 5.5008, 4.9360, 4.7040], device='cuda:2'), covar=tensor([0.1106, 0.1812, 0.2162, 0.1881, 0.2243, 0.0897, 0.1734, 0.2484], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0625, 0.0686, 0.0508, 0.0680, 0.0713, 0.0536, 0.0684], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 09:44:33,190 INFO [train.py:904] (2/8) Epoch 27, batch 6150, loss[loss=0.2141, simple_loss=0.2991, pruned_loss=0.06454, over 15247.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2872, pruned_loss=0.05538, over 3099496.02 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:44:36,351 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4402, 4.5210, 4.8485, 4.7930, 4.8270, 4.5214, 4.5160, 4.3860], device='cuda:2'), covar=tensor([0.0349, 0.0543, 0.0345, 0.0399, 0.0437, 0.0421, 0.0896, 0.0536], device='cuda:2'), in_proj_covar=tensor([0.0430, 0.0482, 0.0468, 0.0429, 0.0516, 0.0493, 0.0569, 0.0396], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 09:44:37,804 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270056.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:53,892 INFO [optim.py:368] (2/8) 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 09:45:16,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7178, 1.8010, 1.6484, 1.5000, 1.9484, 1.5689, 1.6276, 1.9595], device='cuda:2'), covar=tensor([0.0219, 0.0318, 0.0430, 0.0373, 0.0227, 0.0302, 0.0182, 0.0236], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0240, 0.0230, 0.0231, 0.0242, 0.0240, 0.0238, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:45:42,278 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270097.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:45:51,123 INFO [train.py:904] (2/8) Epoch 27, batch 6200, loss[loss=0.2138, simple_loss=0.302, pruned_loss=0.06278, over 15421.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.285, pruned_loss=0.05481, over 3090967.39 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:46:33,640 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270129.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:46:57,435 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:47:09,692 INFO [train.py:904] (2/8) Epoch 27, batch 6250, loss[loss=0.173, simple_loss=0.2646, pruned_loss=0.04074, over 16846.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2845, pruned_loss=0.05407, over 3111667.64 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:47:29,202 INFO [optim.py:368] (2/8) 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 09:48:07,951 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 09:48:08,657 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:48:25,154 INFO [train.py:904] (2/8) Epoch 27, batch 6300, loss[loss=0.1916, simple_loss=0.2835, pruned_loss=0.04987, over 16844.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2844, pruned_loss=0.05391, over 3100816.05 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:49:45,986 INFO [train.py:904] (2/8) Epoch 27, batch 6350, loss[loss=0.1615, simple_loss=0.2584, pruned_loss=0.03228, over 16866.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2852, pruned_loss=0.05508, over 3088610.17 frames. ], batch size: 102, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:49:47,938 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 09:50:05,608 INFO [optim.py:368] (2/8) 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 09:50:28,059 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270280.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:51:02,482 INFO [train.py:904] (2/8) Epoch 27, batch 6400, loss[loss=0.2762, simple_loss=0.3398, pruned_loss=0.1063, over 11047.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2852, pruned_loss=0.05618, over 3090709.82 frames. ], batch size: 250, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:51:04,776 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-02 09:52:01,121 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270341.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:52:18,784 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-02 09:52:19,089 INFO [train.py:904] (2/8) Epoch 27, batch 6450, loss[loss=0.1938, simple_loss=0.288, pruned_loss=0.04985, over 16323.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2846, pruned_loss=0.05475, over 3106665.22 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:39,047 INFO [optim.py:368] (2/8) 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 09:53:13,264 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-05-02 09:53:25,919 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6684, 2.7973, 2.5884, 4.2867, 3.0980, 4.0386, 1.6114, 2.8779], device='cuda:2'), covar=tensor([0.1471, 0.0773, 0.1270, 0.0196, 0.0275, 0.0409, 0.1768, 0.0878], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0180, 0.0201, 0.0202, 0.0209, 0.0220, 0.0210, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 09:53:27,547 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:53:37,245 INFO [train.py:904] (2/8) Epoch 27, batch 6500, loss[loss=0.2098, simple_loss=0.2742, pruned_loss=0.07271, over 11500.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.283, pruned_loss=0.05481, over 3078062.30 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:54:17,020 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270429.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:54:33,580 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270439.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:54:58,115 INFO [train.py:904] (2/8) Epoch 27, batch 6550, loss[loss=0.1989, simple_loss=0.3022, pruned_loss=0.04783, over 17105.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2851, pruned_loss=0.0554, over 3074779.72 frames. ], batch size: 49, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:55:04,780 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:17,872 INFO [optim.py:368] (2/8) 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 09:55:34,671 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:57,781 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:59,542 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5662, 5.8898, 5.5939, 5.6394, 5.3394, 5.2519, 5.2109, 6.0169], device='cuda:2'), covar=tensor([0.1412, 0.0878, 0.1065, 0.0918, 0.0895, 0.0740, 0.1369, 0.0858], device='cuda:2'), in_proj_covar=tensor([0.0705, 0.0853, 0.0696, 0.0657, 0.0538, 0.0537, 0.0717, 0.0666], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:56:10,015 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 6600, loss[loss=0.193, simple_loss=0.2845, pruned_loss=0.05078, over 16723.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2876, pruned_loss=0.056, over 3089638.39 frames. ], batch size: 57, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:09,702 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:57:30,476 INFO [train.py:904] (2/8) Epoch 27, batch 6650, loss[loss=0.1874, simple_loss=0.2797, pruned_loss=0.04759, over 16915.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2872, pruned_loss=0.05605, over 3110554.25 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:34,280 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5559, 2.2305, 1.8616, 2.1349, 2.5714, 2.2660, 2.3314, 2.7342], device='cuda:2'), covar=tensor([0.0248, 0.0487, 0.0629, 0.0512, 0.0294, 0.0448, 0.0257, 0.0305], device='cuda:2'), in_proj_covar=tensor([0.0223, 0.0239, 0.0229, 0.0230, 0.0240, 0.0239, 0.0237, 0.0238], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 09:57:50,326 INFO [optim.py:368] (2/8) 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 09:58:46,096 INFO [train.py:904] (2/8) Epoch 27, batch 6700, loss[loss=0.2243, simple_loss=0.2939, pruned_loss=0.07737, over 11134.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2866, pruned_loss=0.05691, over 3088721.69 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:59:01,553 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:59:04,620 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8734, 4.9549, 4.7075, 3.1425, 4.1400, 4.7217, 4.0478, 2.9106], device='cuda:2'), covar=tensor([0.0530, 0.0033, 0.0044, 0.0406, 0.0107, 0.0109, 0.0094, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0089, 0.0090, 0.0136, 0.0101, 0.0115, 0.0098, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 09:59:35,301 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270636.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:00:01,086 INFO [train.py:904] (2/8) Epoch 27, batch 6750, loss[loss=0.1864, simple_loss=0.2746, pruned_loss=0.04907, over 15233.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2864, pruned_loss=0.05728, over 3083495.12 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:00:20,167 INFO [optim.py:368] (2/8) 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 10:00:32,908 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:01:15,210 INFO [train.py:904] (2/8) Epoch 27, batch 6800, loss[loss=0.1922, simple_loss=0.2887, pruned_loss=0.04783, over 16469.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2864, pruned_loss=0.05733, over 3072912.23 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:01:34,671 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 10:01:38,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1037, 3.6076, 3.6781, 2.2634, 3.0392, 2.5481, 3.5033, 3.8237], device='cuda:2'), covar=tensor([0.0406, 0.0850, 0.0646, 0.2112, 0.0998, 0.0976, 0.0899, 0.1013], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 10:01:53,306 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270726.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:00,734 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3841, 5.6389, 5.3757, 5.3514, 5.0930, 5.0169, 5.0281, 5.7506], device='cuda:2'), covar=tensor([0.1260, 0.0847, 0.0967, 0.0916, 0.0804, 0.0849, 0.1316, 0.0879], device='cuda:2'), in_proj_covar=tensor([0.0706, 0.0854, 0.0698, 0.0658, 0.0539, 0.0538, 0.0718, 0.0667], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:02:15,324 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0615, 4.1525, 4.4536, 4.4044, 4.4262, 4.1856, 4.1319, 4.1655], device='cuda:2'), covar=tensor([0.0384, 0.0669, 0.0415, 0.0428, 0.0502, 0.0465, 0.1049, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0428, 0.0479, 0.0465, 0.0426, 0.0512, 0.0491, 0.0563, 0.0392], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 10:02:33,308 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:34,157 INFO [train.py:904] (2/8) Epoch 27, batch 6850, loss[loss=0.1889, simple_loss=0.292, pruned_loss=0.04289, over 16643.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2874, pruned_loss=0.05743, over 3090478.74 frames. ], batch size: 57, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:02:53,221 INFO [optim.py:368] (2/8) 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 10:03:16,041 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0526, 4.1394, 3.9456, 3.6364, 3.7077, 4.0581, 3.7589, 3.8682], device='cuda:2'), covar=tensor([0.0740, 0.1013, 0.0333, 0.0356, 0.0764, 0.0637, 0.1174, 0.0653], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0460, 0.0357, 0.0361, 0.0357, 0.0414, 0.0245, 0.0429], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:03:25,249 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:37,647 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270795.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:46,618 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4113, 4.4671, 4.7724, 4.7361, 4.7610, 4.5139, 4.4349, 4.4188], device='cuda:2'), covar=tensor([0.0428, 0.0753, 0.0583, 0.0529, 0.0626, 0.0644, 0.1024, 0.0599], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0480, 0.0466, 0.0427, 0.0513, 0.0492, 0.0565, 0.0393], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 10:03:49,486 INFO [train.py:904] (2/8) Epoch 27, batch 6900, loss[loss=0.2116, simple_loss=0.295, pruned_loss=0.06406, over 15362.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2896, pruned_loss=0.05743, over 3087488.56 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:05:07,709 INFO [train.py:904] (2/8) Epoch 27, batch 6950, loss[loss=0.1751, simple_loss=0.2696, pruned_loss=0.04028, over 17242.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2916, pruned_loss=0.05916, over 3066501.25 frames. ], batch size: 43, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:05:28,446 INFO [optim.py:368] (2/8) 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 10:06:23,846 INFO [train.py:904] (2/8) Epoch 27, batch 7000, loss[loss=0.2095, simple_loss=0.3011, pruned_loss=0.05889, over 16884.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2909, pruned_loss=0.05749, over 3091327.56 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:06:46,519 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:07:15,572 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270936.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 10:07:40,705 INFO [train.py:904] (2/8) Epoch 27, batch 7050, loss[loss=0.2164, simple_loss=0.2887, pruned_loss=0.07205, over 11092.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2914, pruned_loss=0.05738, over 3084297.41 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:08:01,040 INFO [optim.py:368] (2/8) 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 10:08:05,105 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:18,489 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:27,533 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270984.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:08:58,235 INFO [train.py:904] (2/8) Epoch 27, batch 7100, loss[loss=0.2117, simple_loss=0.3033, pruned_loss=0.06009, over 16699.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2901, pruned_loss=0.05677, over 3099467.44 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:09:22,210 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 10:10:15,260 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:10:16,497 INFO [train.py:904] (2/8) Epoch 27, batch 7150, loss[loss=0.1681, simple_loss=0.273, pruned_loss=0.03158, over 16899.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.289, pruned_loss=0.05701, over 3093311.74 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:10:37,966 INFO [optim.py:368] (2/8) 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 10:10:59,671 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:18,928 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:28,068 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:31,638 INFO [train.py:904] (2/8) Epoch 27, batch 7200, loss[loss=0.1826, simple_loss=0.2744, pruned_loss=0.04539, over 15450.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2869, pruned_loss=0.05575, over 3081995.80 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:11:53,027 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1048, 2.3067, 2.2680, 3.8230, 2.2039, 2.6410, 2.3850, 2.4548], device='cuda:2'), covar=tensor([0.1517, 0.3584, 0.3226, 0.0609, 0.4229, 0.2490, 0.3624, 0.3467], device='cuda:2'), in_proj_covar=tensor([0.0417, 0.0470, 0.0381, 0.0333, 0.0444, 0.0537, 0.0441, 0.0549], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:12:37,364 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:12:53,489 INFO [train.py:904] (2/8) Epoch 27, batch 7250, loss[loss=0.2214, simple_loss=0.2902, pruned_loss=0.07635, over 11283.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2845, pruned_loss=0.0545, over 3080650.41 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:13:16,038 INFO [optim.py:368] (2/8) 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 10:14:10,772 INFO [train.py:904] (2/8) Epoch 27, batch 7300, loss[loss=0.1777, simple_loss=0.2715, pruned_loss=0.04201, over 17206.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.284, pruned_loss=0.05439, over 3090655.04 frames. ], batch size: 44, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:14:58,404 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 10:15:29,743 INFO [train.py:904] (2/8) Epoch 27, batch 7350, loss[loss=0.2031, simple_loss=0.2882, pruned_loss=0.05898, over 16438.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2845, pruned_loss=0.05506, over 3077964.30 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:30,538 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-05-02 10:15:42,311 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:15:50,537 INFO [optim.py:368] (2/8) 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 10:15:51,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9609, 2.1582, 2.2926, 3.3239, 2.1058, 2.4341, 2.2916, 2.2833], device='cuda:2'), covar=tensor([0.1524, 0.3707, 0.2992, 0.0752, 0.4405, 0.2592, 0.3627, 0.3666], device='cuda:2'), in_proj_covar=tensor([0.0418, 0.0471, 0.0382, 0.0334, 0.0445, 0.0539, 0.0442, 0.0550], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:15:54,194 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:15:59,451 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5565, 4.6160, 4.4455, 4.1400, 4.1471, 4.5429, 4.2999, 4.2618], device='cuda:2'), covar=tensor([0.0584, 0.0499, 0.0298, 0.0312, 0.0823, 0.0452, 0.0611, 0.0616], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0456, 0.0354, 0.0356, 0.0353, 0.0409, 0.0243, 0.0424], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:16:01,221 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:48,724 INFO [train.py:904] (2/8) Epoch 27, batch 7400, loss[loss=0.2149, simple_loss=0.2897, pruned_loss=0.07002, over 11436.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2862, pruned_loss=0.05632, over 3068170.07 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:17:10,655 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:17:18,294 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:06,482 INFO [train.py:904] (2/8) Epoch 27, batch 7450, loss[loss=0.1832, simple_loss=0.2832, pruned_loss=0.04161, over 16787.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.287, pruned_loss=0.05675, over 3080397.14 frames. ], batch size: 89, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:18:30,891 INFO [optim.py:368] (2/8) 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 10:18:45,266 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:55,945 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271382.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:19:19,484 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0271, 3.5630, 3.6255, 2.2717, 3.3163, 3.6327, 3.3362, 1.9949], device='cuda:2'), covar=tensor([0.0657, 0.0079, 0.0079, 0.0497, 0.0129, 0.0139, 0.0124, 0.0553], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0089, 0.0089, 0.0135, 0.0100, 0.0114, 0.0097, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 10:19:30,466 INFO [train.py:904] (2/8) Epoch 27, batch 7500, loss[loss=0.2038, simple_loss=0.2885, pruned_loss=0.05953, over 16193.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2878, pruned_loss=0.05692, over 3062066.20 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:19:47,669 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-02 10:20:08,065 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:14,058 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:22,935 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271436.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:49,673 INFO [train.py:904] (2/8) Epoch 27, batch 7550, loss[loss=0.1853, simple_loss=0.2692, pruned_loss=0.05068, over 16409.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2866, pruned_loss=0.05713, over 3058090.58 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:21:11,190 INFO [optim.py:368] (2/8) 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 10:21:29,614 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8435, 1.4178, 1.7480, 1.7139, 1.8000, 1.9146, 1.6417, 1.8495], device='cuda:2'), covar=tensor([0.0288, 0.0459, 0.0249, 0.0343, 0.0311, 0.0201, 0.0498, 0.0159], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0188, 0.0203, 0.0162, 0.0200, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:21:41,025 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:21:57,533 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 10:22:05,495 INFO [train.py:904] (2/8) Epoch 27, batch 7600, loss[loss=0.235, simple_loss=0.3017, pruned_loss=0.08412, over 11329.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.286, pruned_loss=0.05771, over 3031777.77 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:22,937 INFO [train.py:904] (2/8) Epoch 27, batch 7650, loss[loss=0.2541, simple_loss=0.3161, pruned_loss=0.09609, over 11480.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.286, pruned_loss=0.05787, over 3038713.88 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:45,488 INFO [optim.py:368] (2/8) 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 10:23:55,620 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:24:02,645 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6486, 4.7078, 4.5345, 4.2104, 4.1968, 4.6195, 4.3920, 4.3440], device='cuda:2'), covar=tensor([0.0652, 0.0569, 0.0346, 0.0346, 0.0955, 0.0547, 0.0506, 0.0693], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0456, 0.0354, 0.0356, 0.0352, 0.0410, 0.0243, 0.0423], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:24:29,309 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8506, 2.6911, 2.5565, 1.9069, 2.5441, 2.6636, 2.5337, 1.9504], device='cuda:2'), covar=tensor([0.0464, 0.0108, 0.0104, 0.0405, 0.0161, 0.0156, 0.0137, 0.0447], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0134, 0.0099, 0.0113, 0.0097, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 10:24:34,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6831, 1.7952, 1.6202, 1.4796, 1.9231, 1.5643, 1.5791, 1.9130], device='cuda:2'), covar=tensor([0.0207, 0.0295, 0.0464, 0.0398, 0.0227, 0.0286, 0.0176, 0.0221], device='cuda:2'), in_proj_covar=tensor([0.0224, 0.0241, 0.0230, 0.0232, 0.0242, 0.0239, 0.0239, 0.0239], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:24:43,692 INFO [train.py:904] (2/8) Epoch 27, batch 7700, loss[loss=0.2428, simple_loss=0.3092, pruned_loss=0.08817, over 11916.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2859, pruned_loss=0.05799, over 3036009.87 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:25:01,709 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 10:25:06,163 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:26:02,027 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 10:26:02,414 INFO [train.py:904] (2/8) Epoch 27, batch 7750, loss[loss=0.197, simple_loss=0.2872, pruned_loss=0.05342, over 17178.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.286, pruned_loss=0.05768, over 3046923.07 frames. ], batch size: 46, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:26:24,502 INFO [optim.py:368] (2/8) 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 10:26:35,168 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3469, 2.6060, 2.1215, 2.3196, 2.9302, 2.5475, 2.8954, 3.1163], device='cuda:2'), covar=tensor([0.0183, 0.0443, 0.0601, 0.0504, 0.0276, 0.0410, 0.0250, 0.0285], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0243, 0.0231, 0.0233, 0.0244, 0.0241, 0.0240, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:27:19,377 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3298, 4.4013, 4.2179, 3.9372, 3.9348, 4.3098, 4.0162, 4.0516], device='cuda:2'), covar=tensor([0.0635, 0.0764, 0.0331, 0.0336, 0.0835, 0.0564, 0.0841, 0.0712], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0455, 0.0353, 0.0356, 0.0352, 0.0409, 0.0243, 0.0423], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:27:20,090 INFO [train.py:904] (2/8) Epoch 27, batch 7800, loss[loss=0.1836, simple_loss=0.2767, pruned_loss=0.04521, over 16446.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2865, pruned_loss=0.05801, over 3063460.45 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:27:42,990 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.70 vs. limit=5.0 2023-05-02 10:28:04,234 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 7850, loss[loss=0.2132, simple_loss=0.3025, pruned_loss=0.06197, over 15466.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.288, pruned_loss=0.05812, over 3073098.17 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:28:57,992 INFO [optim.py:368] (2/8) 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 10:29:20,882 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:29:24,347 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0885, 2.4196, 2.5837, 1.9542, 2.6850, 2.7553, 2.4113, 2.3928], device='cuda:2'), covar=tensor([0.0694, 0.0312, 0.0247, 0.0956, 0.0146, 0.0312, 0.0483, 0.0480], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0110, 0.0101, 0.0138, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 10:29:45,391 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7540, 4.7997, 4.6378, 4.3009, 4.3060, 4.6968, 4.5507, 4.3976], device='cuda:2'), covar=tensor([0.0665, 0.0603, 0.0327, 0.0358, 0.0990, 0.0566, 0.0397, 0.0740], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0454, 0.0353, 0.0355, 0.0352, 0.0409, 0.0243, 0.0423], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:29:52,537 INFO [train.py:904] (2/8) Epoch 27, batch 7900, loss[loss=0.198, simple_loss=0.294, pruned_loss=0.05099, over 16909.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2869, pruned_loss=0.05681, over 3103813.68 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:31:12,162 INFO [train.py:904] (2/8) Epoch 27, batch 7950, loss[loss=0.1891, simple_loss=0.277, pruned_loss=0.05059, over 16155.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.287, pruned_loss=0.05718, over 3097913.91 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:31:15,339 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:31:36,730 INFO [optim.py:368] (2/8) 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 10:31:48,134 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 10:32:29,919 INFO [train.py:904] (2/8) Epoch 27, batch 8000, loss[loss=0.1815, simple_loss=0.2789, pruned_loss=0.04206, over 16767.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2877, pruned_loss=0.05801, over 3078347.69 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:32:49,713 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271915.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:32:52,880 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:19,443 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:44,067 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 10:33:47,184 INFO [train.py:904] (2/8) Epoch 27, batch 8050, loss[loss=0.2053, simple_loss=0.2945, pruned_loss=0.05806, over 16896.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2877, pruned_loss=0.05765, over 3065393.80 frames. ], batch size: 109, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:34:05,655 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:34:09,130 INFO [optim.py:368] (2/8) 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 10:34:49,585 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:35:04,671 INFO [train.py:904] (2/8) Epoch 27, batch 8100, loss[loss=0.2076, simple_loss=0.2869, pruned_loss=0.06413, over 16400.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2868, pruned_loss=0.05665, over 3074875.79 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:35:45,687 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:36:17,765 INFO [train.py:904] (2/8) Epoch 27, batch 8150, loss[loss=0.1771, simple_loss=0.2691, pruned_loss=0.04255, over 16891.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.284, pruned_loss=0.05541, over 3085524.28 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:36:39,765 INFO [optim.py:368] (2/8) 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 10:36:57,158 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:01,573 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:32,132 INFO [train.py:904] (2/8) Epoch 27, batch 8200, loss[loss=0.1965, simple_loss=0.2696, pruned_loss=0.06168, over 11492.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2819, pruned_loss=0.055, over 3083509.76 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:38:12,530 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:16,439 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:53,591 INFO [train.py:904] (2/8) Epoch 27, batch 8250, loss[loss=0.1585, simple_loss=0.2615, pruned_loss=0.02775, over 16449.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2804, pruned_loss=0.05252, over 3073373.99 frames. ], batch size: 75, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:39:19,036 INFO [optim.py:368] (2/8) 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 10:39:40,163 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272181.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:39:50,768 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272188.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:39:59,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3487, 3.4621, 3.6623, 3.6452, 3.6450, 3.4819, 3.4880, 3.5527], device='cuda:2'), covar=tensor([0.0520, 0.0993, 0.0549, 0.0516, 0.0523, 0.0741, 0.0889, 0.0650], device='cuda:2'), in_proj_covar=tensor([0.0430, 0.0484, 0.0467, 0.0429, 0.0514, 0.0493, 0.0566, 0.0395], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 10:40:04,144 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 10:40:14,329 INFO [train.py:904] (2/8) Epoch 27, batch 8300, loss[loss=0.182, simple_loss=0.2768, pruned_loss=0.04364, over 15381.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.278, pruned_loss=0.0494, over 3082282.67 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:40:26,348 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:35,503 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0495, 1.8538, 1.7096, 1.4872, 1.9739, 1.6362, 1.6104, 1.9672], device='cuda:2'), covar=tensor([0.0233, 0.0351, 0.0460, 0.0427, 0.0269, 0.0307, 0.0204, 0.0240], device='cuda:2'), in_proj_covar=tensor([0.0222, 0.0240, 0.0228, 0.0230, 0.0241, 0.0239, 0.0237, 0.0237], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:40:46,066 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:15,147 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272242.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:41:22,920 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272247.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:31,410 INFO [train.py:904] (2/8) Epoch 27, batch 8350, loss[loss=0.2146, simple_loss=0.2932, pruned_loss=0.06805, over 12161.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2773, pruned_loss=0.04744, over 3080196.28 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:41:54,866 INFO [optim.py:368] (2/8) 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 10:42:20,364 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272284.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:30,778 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272290.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:50,436 INFO [train.py:904] (2/8) Epoch 27, batch 8400, loss[loss=0.1392, simple_loss=0.241, pruned_loss=0.01871, over 16746.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2748, pruned_loss=0.0459, over 3037006.42 frames. ], batch size: 89, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:42:59,232 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:43:31,260 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4525, 2.8865, 3.2056, 2.0737, 2.8643, 2.1426, 3.0952, 3.1417], device='cuda:2'), covar=tensor([0.0314, 0.1003, 0.0505, 0.2159, 0.0833, 0.1094, 0.0630, 0.0986], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0168, 0.0167, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 10:44:09,889 INFO [train.py:904] (2/8) Epoch 27, batch 8450, loss[loss=0.1539, simple_loss=0.2578, pruned_loss=0.02496, over 16814.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2733, pruned_loss=0.04435, over 3035441.79 frames. ], batch size: 83, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:44:34,140 INFO [optim.py:368] (2/8) 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 10:44:42,851 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272373.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:45:32,400 INFO [train.py:904] (2/8) Epoch 27, batch 8500, loss[loss=0.1535, simple_loss=0.2496, pruned_loss=0.02866, over 16714.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2695, pruned_loss=0.04219, over 3044375.62 frames. ], batch size: 83, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:45:58,315 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6660, 2.9719, 3.3708, 2.1313, 2.9449, 2.1848, 3.2996, 3.1806], device='cuda:2'), covar=tensor([0.0282, 0.0988, 0.0492, 0.2117, 0.0805, 0.1060, 0.0633, 0.0983], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0154, 0.0146, 0.0130, 0.0143, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 10:46:24,075 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 27, batch 8550, loss[loss=0.1591, simple_loss=0.2577, pruned_loss=0.03024, over 15177.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2672, pruned_loss=0.04126, over 3024737.62 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:47:25,000 INFO [optim.py:368] (2/8) 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 10:47:38,246 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-05-02 10:47:55,107 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272483.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:48:35,726 INFO [train.py:904] (2/8) Epoch 27, batch 8600, loss[loss=0.188, simple_loss=0.2926, pruned_loss=0.04172, over 15359.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2681, pruned_loss=0.04087, over 3026130.13 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:48:50,801 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272510.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:49:43,699 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272537.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 10:50:13,935 INFO [train.py:904] (2/8) Epoch 27, batch 8650, loss[loss=0.1508, simple_loss=0.2571, pruned_loss=0.02226, over 16894.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2669, pruned_loss=0.03936, over 3051477.23 frames. ], batch size: 102, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:50:26,983 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:50:48,990 INFO [optim.py:368] (2/8) 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 10:50:49,866 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4208, 2.0887, 1.8318, 1.8626, 2.3089, 1.9903, 1.7988, 2.3705], device='cuda:2'), covar=tensor([0.0222, 0.0465, 0.0575, 0.0576, 0.0318, 0.0433, 0.0213, 0.0297], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0239, 0.0228, 0.0230, 0.0240, 0.0237, 0.0236, 0.0235], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:51:15,218 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:51:36,623 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:00,241 INFO [train.py:904] (2/8) Epoch 27, batch 8700, loss[loss=0.1551, simple_loss=0.2426, pruned_loss=0.03381, over 12358.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2637, pruned_loss=0.03795, over 3037957.46 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:52:01,200 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:03,095 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2277, 4.3575, 4.4730, 4.2341, 4.3884, 4.8389, 4.4172, 4.1295], device='cuda:2'), covar=tensor([0.1792, 0.2066, 0.2329, 0.2293, 0.2529, 0.1029, 0.1527, 0.2353], device='cuda:2'), in_proj_covar=tensor([0.0416, 0.0617, 0.0677, 0.0503, 0.0672, 0.0709, 0.0533, 0.0672], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 10:52:17,102 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3972, 3.4777, 3.6641, 3.6643, 3.6528, 3.4734, 3.5232, 3.5603], device='cuda:2'), covar=tensor([0.0489, 0.0941, 0.0573, 0.0555, 0.0631, 0.0755, 0.0805, 0.0559], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0475, 0.0460, 0.0423, 0.0507, 0.0485, 0.0556, 0.0389], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 10:52:21,607 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:53:06,582 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272638.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:53:35,410 INFO [train.py:904] (2/8) Epoch 27, batch 8750, loss[loss=0.1767, simple_loss=0.2754, pruned_loss=0.03901, over 15434.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2638, pruned_loss=0.03747, over 3038344.32 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 10:54:15,877 INFO [optim.py:368] (2/8) 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 10:54:28,892 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:55:13,540 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0514, 2.1360, 2.1338, 3.6457, 2.0356, 2.4262, 2.2694, 2.2085], device='cuda:2'), covar=tensor([0.1517, 0.3971, 0.3584, 0.0654, 0.4702, 0.2765, 0.3822, 0.3966], device='cuda:2'), in_proj_covar=tensor([0.0409, 0.0462, 0.0376, 0.0326, 0.0437, 0.0527, 0.0434, 0.0539], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:55:27,076 INFO [train.py:904] (2/8) Epoch 27, batch 8800, loss[loss=0.1761, simple_loss=0.2695, pruned_loss=0.04138, over 16927.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2618, pruned_loss=0.03639, over 3037903.25 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:56:23,109 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:57:01,433 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8032, 5.0851, 4.9167, 4.9308, 4.6329, 4.5666, 4.4854, 5.1764], device='cuda:2'), covar=tensor([0.1061, 0.0900, 0.0813, 0.0693, 0.0754, 0.1069, 0.1307, 0.0812], device='cuda:2'), in_proj_covar=tensor([0.0691, 0.0838, 0.0689, 0.0644, 0.0527, 0.0529, 0.0701, 0.0655], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 10:57:12,472 INFO [train.py:904] (2/8) Epoch 27, batch 8850, loss[loss=0.1744, simple_loss=0.2791, pruned_loss=0.03489, over 16676.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2644, pruned_loss=0.03595, over 3035112.76 frames. ], batch size: 134, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:57:46,548 INFO [optim.py:368] (2/8) 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 10:58:17,295 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272783.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:58:52,771 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-05-02 10:58:57,770 INFO [train.py:904] (2/8) Epoch 27, batch 8900, loss[loss=0.1533, simple_loss=0.2435, pruned_loss=0.03153, over 12757.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.265, pruned_loss=0.03539, over 3035056.37 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:59:10,732 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 11:00:01,531 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:00:20,796 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272837.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 11:00:33,021 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 11:01:00,397 INFO [train.py:904] (2/8) Epoch 27, batch 8950, loss[loss=0.1614, simple_loss=0.2521, pruned_loss=0.03533, over 12892.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2641, pruned_loss=0.0356, over 3027876.51 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:01:35,337 INFO [optim.py:368] (2/8) 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 11:01:57,008 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272879.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:02:12,416 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272885.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 11:02:14,627 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6173, 3.7293, 2.3699, 4.2399, 2.8541, 4.1140, 2.4656, 2.9948], device='cuda:2'), covar=tensor([0.0304, 0.0365, 0.1543, 0.0176, 0.0800, 0.0537, 0.1555, 0.0770], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0175, 0.0191, 0.0164, 0.0175, 0.0211, 0.0199, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:02:27,868 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2755, 4.3205, 4.4788, 4.2296, 4.4125, 4.8517, 4.4573, 4.1578], device='cuda:2'), covar=tensor([0.1671, 0.2075, 0.2369, 0.2146, 0.2285, 0.1031, 0.1607, 0.2537], device='cuda:2'), in_proj_covar=tensor([0.0414, 0.0613, 0.0672, 0.0500, 0.0665, 0.0706, 0.0530, 0.0669], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 11:02:44,623 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1690, 2.1219, 2.6539, 3.1566, 2.9428, 3.5083, 1.9421, 3.5381], device='cuda:2'), covar=tensor([0.0202, 0.0614, 0.0356, 0.0259, 0.0330, 0.0165, 0.0826, 0.0144], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0187, 0.0203, 0.0162, 0.0199, 0.0162], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:02:49,720 INFO [train.py:904] (2/8) Epoch 27, batch 9000, loss[loss=0.1601, simple_loss=0.2492, pruned_loss=0.03548, over 12238.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2611, pruned_loss=0.03447, over 3040651.29 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:02:49,720 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 11:02:59,733 INFO [train.py:938] (2/8) Epoch 27, validation: loss=0.1436, simple_loss=0.2474, pruned_loss=0.01989, over 944034.00 frames. 2023-05-02 11:02:59,733 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 11:03:00,824 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:03:24,647 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6249, 2.5866, 2.3994, 4.2772, 2.6382, 4.0159, 1.4803, 2.8980], device='cuda:2'), covar=tensor([0.1459, 0.0830, 0.1259, 0.0158, 0.0144, 0.0346, 0.1777, 0.0794], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0177, 0.0197, 0.0195, 0.0202, 0.0214, 0.0207, 0.0195], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:03:51,479 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:42,534 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272951.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:44,600 INFO [train.py:904] (2/8) Epoch 27, batch 9050, loss[loss=0.1685, simple_loss=0.268, pruned_loss=0.03453, over 16713.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2625, pruned_loss=0.03504, over 3063778.00 frames. ], batch size: 76, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:05:16,160 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 11:05:18,775 INFO [optim.py:368] (2/8) 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 11:05:19,390 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:05:57,066 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0233, 1.9862, 2.1398, 3.5934, 1.9762, 2.2096, 2.1467, 2.0931], device='cuda:2'), covar=tensor([0.1554, 0.4434, 0.3454, 0.0661, 0.5306, 0.3196, 0.4095, 0.4149], device='cuda:2'), in_proj_covar=tensor([0.0407, 0.0460, 0.0375, 0.0324, 0.0434, 0.0524, 0.0433, 0.0536], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:06:32,242 INFO [train.py:904] (2/8) Epoch 27, batch 9100, loss[loss=0.1607, simple_loss=0.265, pruned_loss=0.0282, over 15123.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2618, pruned_loss=0.03515, over 3076701.64 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:06:33,566 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5885, 3.5468, 3.5132, 2.7004, 3.4279, 2.0179, 3.2364, 2.8242], device='cuda:2'), covar=tensor([0.0153, 0.0136, 0.0201, 0.0211, 0.0109, 0.2637, 0.0138, 0.0248], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0169, 0.0205, 0.0179, 0.0182, 0.0212, 0.0194, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:07:36,038 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273029.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:08:32,208 INFO [train.py:904] (2/8) Epoch 27, batch 9150, loss[loss=0.1664, simple_loss=0.2557, pruned_loss=0.03858, over 11898.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2619, pruned_loss=0.0346, over 3065495.34 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:09:08,126 INFO [optim.py:368] (2/8) 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 11:09:27,708 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273077.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:10:18,352 INFO [train.py:904] (2/8) Epoch 27, batch 9200, loss[loss=0.1893, simple_loss=0.2807, pruned_loss=0.04893, over 15146.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2577, pruned_loss=0.0337, over 3068075.56 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:11:05,807 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6476, 3.5951, 3.9858, 1.9981, 4.1223, 4.2052, 3.1403, 3.2028], device='cuda:2'), covar=tensor([0.0807, 0.0255, 0.0201, 0.1265, 0.0078, 0.0150, 0.0434, 0.0439], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0106, 0.0097, 0.0134, 0.0083, 0.0125, 0.0125, 0.0126], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 11:11:53,369 INFO [train.py:904] (2/8) Epoch 27, batch 9250, loss[loss=0.1597, simple_loss=0.2568, pruned_loss=0.03127, over 15432.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2573, pruned_loss=0.03365, over 3059950.40 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:11:56,734 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6775, 4.6288, 4.3961, 3.7868, 4.5578, 1.9129, 4.3040, 4.1343], device='cuda:2'), covar=tensor([0.0095, 0.0101, 0.0212, 0.0259, 0.0092, 0.2685, 0.0133, 0.0240], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0168, 0.0204, 0.0178, 0.0181, 0.0211, 0.0193, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:12:14,614 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 11:12:25,531 INFO [optim.py:368] (2/8) 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 11:13:35,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2215, 4.3315, 4.1472, 3.8821, 3.8975, 4.2542, 3.9392, 4.0136], device='cuda:2'), covar=tensor([0.0625, 0.0707, 0.0327, 0.0301, 0.0693, 0.0651, 0.0858, 0.0646], device='cuda:2'), in_proj_covar=tensor([0.0299, 0.0448, 0.0348, 0.0351, 0.0345, 0.0404, 0.0241, 0.0418], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:13:43,821 INFO [train.py:904] (2/8) Epoch 27, batch 9300, loss[loss=0.1605, simple_loss=0.2539, pruned_loss=0.03356, over 16215.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2558, pruned_loss=0.03346, over 3050057.79 frames. ], batch size: 35, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:13:59,779 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7667, 6.2077, 5.9035, 5.9986, 5.5529, 5.5520, 5.6245, 6.3049], device='cuda:2'), covar=tensor([0.1234, 0.0859, 0.0978, 0.0797, 0.0837, 0.0560, 0.1261, 0.0807], device='cuda:2'), in_proj_covar=tensor([0.0690, 0.0837, 0.0685, 0.0643, 0.0527, 0.0529, 0.0700, 0.0655], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:15:07,504 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2683, 5.8303, 5.9743, 5.6360, 5.8166, 6.2712, 5.7810, 5.4672], device='cuda:2'), covar=tensor([0.0716, 0.1713, 0.1744, 0.1999, 0.2028, 0.0776, 0.1368, 0.2230], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0606, 0.0665, 0.0494, 0.0656, 0.0698, 0.0524, 0.0657], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 11:15:28,949 INFO [train.py:904] (2/8) Epoch 27, batch 9350, loss[loss=0.1529, simple_loss=0.2415, pruned_loss=0.03213, over 12413.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2557, pruned_loss=0.03312, over 3069097.36 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:16:02,980 INFO [optim.py:368] (2/8) 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 11:16:03,978 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:17:10,295 INFO [train.py:904] (2/8) Epoch 27, batch 9400, loss[loss=0.1505, simple_loss=0.2412, pruned_loss=0.02989, over 12360.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2561, pruned_loss=0.03336, over 3062097.47 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:17:39,075 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:18:03,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9442, 2.7465, 2.9225, 2.1959, 2.7261, 2.1617, 2.7595, 2.9255], device='cuda:2'), covar=tensor([0.0323, 0.0970, 0.0539, 0.1854, 0.0804, 0.1030, 0.0606, 0.0882], device='cuda:2'), in_proj_covar=tensor([0.0155, 0.0164, 0.0164, 0.0152, 0.0144, 0.0129, 0.0141, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:18:52,064 INFO [train.py:904] (2/8) Epoch 27, batch 9450, loss[loss=0.1584, simple_loss=0.2537, pruned_loss=0.03152, over 16593.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.257, pruned_loss=0.03357, over 3032253.46 frames. ], batch size: 62, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:19:21,969 INFO [optim.py:368] (2/8) 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 11:19:40,434 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1773, 4.2364, 4.5569, 4.5246, 4.5256, 4.2977, 4.2444, 4.2694], device='cuda:2'), covar=tensor([0.0567, 0.1227, 0.0755, 0.0905, 0.0921, 0.0852, 0.1179, 0.0653], device='cuda:2'), in_proj_covar=tensor([0.0411, 0.0461, 0.0448, 0.0413, 0.0494, 0.0473, 0.0541, 0.0379], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 11:20:32,362 INFO [train.py:904] (2/8) Epoch 27, batch 9500, loss[loss=0.1433, simple_loss=0.231, pruned_loss=0.02776, over 13023.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.257, pruned_loss=0.03358, over 3041419.35 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:20:49,220 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273409.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:21:38,354 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:22:18,627 INFO [train.py:904] (2/8) Epoch 27, batch 9550, loss[loss=0.1704, simple_loss=0.2706, pruned_loss=0.03511, over 16237.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2569, pruned_loss=0.0336, over 3063500.45 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:22:51,234 INFO [optim.py:368] (2/8) 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 11:22:53,767 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:32,979 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2379, 4.3088, 4.1233, 3.8129, 3.8781, 4.2233, 3.9313, 3.9705], device='cuda:2'), covar=tensor([0.0589, 0.0607, 0.0320, 0.0318, 0.0720, 0.0596, 0.0854, 0.0637], device='cuda:2'), in_proj_covar=tensor([0.0296, 0.0444, 0.0346, 0.0348, 0.0343, 0.0400, 0.0239, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:23:46,461 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:57,853 INFO [train.py:904] (2/8) Epoch 27, batch 9600, loss[loss=0.1805, simple_loss=0.2822, pruned_loss=0.03937, over 16394.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2586, pruned_loss=0.03436, over 3076056.05 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:24:16,811 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6386, 2.5589, 2.3188, 4.1101, 2.1430, 3.8884, 1.4642, 2.8253], device='cuda:2'), covar=tensor([0.1531, 0.0892, 0.1374, 0.0149, 0.0129, 0.0375, 0.1862, 0.0857], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0193, 0.0199, 0.0212, 0.0205, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:25:10,438 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-02 11:25:44,217 INFO [train.py:904] (2/8) Epoch 27, batch 9650, loss[loss=0.171, simple_loss=0.2672, pruned_loss=0.03745, over 16340.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2606, pruned_loss=0.03481, over 3077971.52 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:26:23,245 INFO [optim.py:368] (2/8) 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 11:27:30,800 INFO [train.py:904] (2/8) Epoch 27, batch 9700, loss[loss=0.1854, simple_loss=0.2757, pruned_loss=0.04757, over 15179.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2599, pruned_loss=0.03462, over 3069111.13 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:12,471 INFO [train.py:904] (2/8) Epoch 27, batch 9750, loss[loss=0.1456, simple_loss=0.2463, pruned_loss=0.02247, over 16934.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2587, pruned_loss=0.03465, over 3067679.75 frames. ], batch size: 102, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:38,184 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8868, 2.7093, 2.9705, 2.1187, 2.7431, 2.1981, 2.7465, 2.8745], device='cuda:2'), covar=tensor([0.0289, 0.1007, 0.0449, 0.1899, 0.0733, 0.0940, 0.0601, 0.0864], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0165, 0.0153, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:29:42,115 INFO [optim.py:368] (2/8) 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 11:30:51,288 INFO [train.py:904] (2/8) Epoch 27, batch 9800, loss[loss=0.1674, simple_loss=0.2689, pruned_loss=0.03293, over 16990.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2589, pruned_loss=0.03393, over 3063799.82 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:31:16,481 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9800, 5.2648, 5.3952, 5.1881, 5.2577, 5.7771, 5.2377, 4.9394], device='cuda:2'), covar=tensor([0.0871, 0.1845, 0.1841, 0.1819, 0.2182, 0.0750, 0.1457, 0.2301], device='cuda:2'), in_proj_covar=tensor([0.0404, 0.0604, 0.0661, 0.0490, 0.0652, 0.0695, 0.0521, 0.0652], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 11:32:36,071 INFO [train.py:904] (2/8) Epoch 27, batch 9850, loss[loss=0.1476, simple_loss=0.2355, pruned_loss=0.02985, over 12319.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.26, pruned_loss=0.03351, over 3065992.35 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:32:45,142 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4801, 3.5725, 2.6895, 2.1401, 2.1748, 2.3300, 3.7861, 3.0442], device='cuda:2'), covar=tensor([0.3254, 0.0630, 0.2109, 0.3359, 0.3200, 0.2402, 0.0444, 0.1644], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0269, 0.0307, 0.0319, 0.0295, 0.0270, 0.0297, 0.0341], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 11:33:00,286 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:33:11,081 INFO [optim.py:368] (2/8) 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 11:34:02,563 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:34:25,789 INFO [train.py:904] (2/8) Epoch 27, batch 9900, loss[loss=0.1664, simple_loss=0.2535, pruned_loss=0.03967, over 12572.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.26, pruned_loss=0.03343, over 3047651.67 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:35:10,190 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9467, 2.1377, 2.1968, 3.4021, 2.0463, 2.3693, 2.2507, 2.2219], device='cuda:2'), covar=tensor([0.1442, 0.3705, 0.3165, 0.0696, 0.4439, 0.2655, 0.3560, 0.3674], device='cuda:2'), in_proj_covar=tensor([0.0408, 0.0460, 0.0376, 0.0325, 0.0435, 0.0523, 0.0433, 0.0535], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:36:22,572 INFO [train.py:904] (2/8) Epoch 27, batch 9950, loss[loss=0.1537, simple_loss=0.2545, pruned_loss=0.02644, over 16612.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.262, pruned_loss=0.03375, over 3058163.53 frames. ], batch size: 57, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:36:23,918 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0614, 4.1398, 3.9687, 3.6865, 3.7227, 4.0554, 3.7410, 3.8528], device='cuda:2'), covar=tensor([0.0560, 0.0638, 0.0345, 0.0305, 0.0715, 0.0524, 0.1131, 0.0575], device='cuda:2'), in_proj_covar=tensor([0.0294, 0.0440, 0.0343, 0.0346, 0.0340, 0.0397, 0.0237, 0.0410], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:37:03,258 INFO [optim.py:368] (2/8) 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 11:38:25,115 INFO [train.py:904] (2/8) Epoch 27, batch 10000, loss[loss=0.169, simple_loss=0.2653, pruned_loss=0.03641, over 15249.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2609, pruned_loss=0.03298, over 3077918.94 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:39:11,780 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7816, 2.6977, 2.6368, 4.4420, 2.8163, 4.0618, 1.5707, 3.1010], device='cuda:2'), covar=tensor([0.1384, 0.0844, 0.1152, 0.0158, 0.0168, 0.0367, 0.1724, 0.0687], device='cuda:2'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0192, 0.0198, 0.0212, 0.0205, 0.0194], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:39:45,873 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8596, 3.8799, 3.9995, 3.7654, 3.9800, 4.3410, 4.0123, 3.7061], device='cuda:2'), covar=tensor([0.2129, 0.2242, 0.2382, 0.2633, 0.2443, 0.1406, 0.1579, 0.2764], device='cuda:2'), in_proj_covar=tensor([0.0403, 0.0602, 0.0661, 0.0490, 0.0653, 0.0694, 0.0521, 0.0651], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 11:40:06,628 INFO [train.py:904] (2/8) Epoch 27, batch 10050, loss[loss=0.1689, simple_loss=0.2667, pruned_loss=0.03555, over 16404.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2606, pruned_loss=0.0331, over 3083292.18 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:40:22,693 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 11:40:39,173 INFO [optim.py:368] (2/8) 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 11:41:40,636 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274001.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:41:43,716 INFO [train.py:904] (2/8) Epoch 27, batch 10100, loss[loss=0.1663, simple_loss=0.2575, pruned_loss=0.03756, over 16886.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.261, pruned_loss=0.03316, over 3089172.46 frames. ], batch size: 116, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:42:47,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2896, 4.4504, 4.1806, 3.8537, 3.7870, 4.3652, 4.0919, 3.9900], device='cuda:2'), covar=tensor([0.0699, 0.0588, 0.0456, 0.0465, 0.1153, 0.0651, 0.0751, 0.0767], device='cuda:2'), in_proj_covar=tensor([0.0292, 0.0435, 0.0341, 0.0342, 0.0337, 0.0393, 0.0235, 0.0406], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:43:00,240 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8642, 2.8041, 2.6551, 2.0549, 2.5264, 2.7962, 2.6209, 1.9812], device='cuda:2'), covar=tensor([0.0452, 0.0082, 0.0088, 0.0357, 0.0167, 0.0112, 0.0119, 0.0459], device='cuda:2'), in_proj_covar=tensor([0.0135, 0.0087, 0.0088, 0.0133, 0.0099, 0.0111, 0.0095, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 11:43:27,167 INFO [train.py:904] (2/8) Epoch 28, batch 0, loss[loss=0.1578, simple_loss=0.2549, pruned_loss=0.0303, over 17124.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2549, pruned_loss=0.0303, over 17124.00 frames. ], batch size: 47, lr: 2.42e-03, grad_scale: 8.0 2023-05-02 11:43:27,167 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 11:43:34,590 INFO [train.py:938] (2/8) Epoch 28, validation: loss=0.1434, simple_loss=0.2464, pruned_loss=0.02022, over 944034.00 frames. 2023-05-02 11:43:34,590 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 11:43:48,092 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274062.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:43:52,122 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:01,201 INFO [optim.py:368] (2/8) 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 11:44:27,471 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:43,867 INFO [train.py:904] (2/8) Epoch 28, batch 50, loss[loss=0.1579, simple_loss=0.2506, pruned_loss=0.03262, over 17116.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2648, pruned_loss=0.04482, over 758301.49 frames. ], batch size: 47, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:44:59,063 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274113.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:24,125 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:34,057 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274139.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:53,997 INFO [train.py:904] (2/8) Epoch 28, batch 100, loss[loss=0.1696, simple_loss=0.2521, pruned_loss=0.04349, over 16696.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2633, pruned_loss=0.04507, over 1316620.37 frames. ], batch size: 89, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:45:59,762 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-05-02 11:46:14,591 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1225, 5.6620, 5.7483, 5.4915, 5.5950, 6.1375, 5.5756, 5.3246], device='cuda:2'), covar=tensor([0.0974, 0.1861, 0.2575, 0.2104, 0.2519, 0.0941, 0.1608, 0.2410], device='cuda:2'), in_proj_covar=tensor([0.0410, 0.0613, 0.0674, 0.0498, 0.0663, 0.0705, 0.0528, 0.0661], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 11:46:20,296 INFO [optim.py:368] (2/8) 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 11:46:48,024 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:47:02,022 INFO [train.py:904] (2/8) Epoch 28, batch 150, loss[loss=0.1867, simple_loss=0.2798, pruned_loss=0.04681, over 16626.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2606, pruned_loss=0.04264, over 1765510.55 frames. ], batch size: 62, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:48:08,592 INFO [train.py:904] (2/8) Epoch 28, batch 200, loss[loss=0.2065, simple_loss=0.2759, pruned_loss=0.06859, over 16835.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2603, pruned_loss=0.04274, over 2110361.76 frames. ], batch size: 96, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:48:15,377 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 11:48:29,688 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:48:33,859 INFO [optim.py:368] (2/8) 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 11:49:16,127 INFO [train.py:904] (2/8) Epoch 28, batch 250, loss[loss=0.1412, simple_loss=0.2359, pruned_loss=0.02323, over 17212.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2588, pruned_loss=0.04223, over 2361496.35 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:49:16,619 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274303.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:49:52,003 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274330.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:49:58,224 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.65 vs. limit=5.0 2023-05-02 11:50:23,018 INFO [train.py:904] (2/8) Epoch 28, batch 300, loss[loss=0.1901, simple_loss=0.2673, pruned_loss=0.05643, over 12180.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2575, pruned_loss=0.04167, over 2554628.67 frames. ], batch size: 248, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:50:29,898 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:39,297 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:49,959 INFO [optim.py:368] (2/8) 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 11:51:31,008 INFO [train.py:904] (2/8) Epoch 28, batch 350, loss[loss=0.1719, simple_loss=0.2533, pruned_loss=0.04527, over 16644.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2541, pruned_loss=0.04003, over 2724098.94 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:52:37,185 INFO [train.py:904] (2/8) Epoch 28, batch 400, loss[loss=0.1697, simple_loss=0.2615, pruned_loss=0.03896, over 16508.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2531, pruned_loss=0.04004, over 2861590.52 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:53:03,892 INFO [optim.py:368] (2/8) 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 11:53:24,167 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:53:29,268 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-02 11:53:44,149 INFO [train.py:904] (2/8) Epoch 28, batch 450, loss[loss=0.1763, simple_loss=0.2537, pruned_loss=0.04945, over 12413.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.252, pruned_loss=0.03962, over 2959997.34 frames. ], batch size: 246, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:54:35,055 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:54:52,695 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4935, 2.3375, 1.8442, 2.0562, 2.6460, 2.4115, 2.5389, 2.7083], device='cuda:2'), covar=tensor([0.0321, 0.0479, 0.0698, 0.0569, 0.0302, 0.0442, 0.0285, 0.0386], device='cuda:2'), in_proj_covar=tensor([0.0229, 0.0246, 0.0235, 0.0235, 0.0247, 0.0245, 0.0242, 0.0242], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 11:54:53,415 INFO [train.py:904] (2/8) Epoch 28, batch 500, loss[loss=0.178, simple_loss=0.2519, pruned_loss=0.05209, over 16474.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2503, pruned_loss=0.03878, over 3034203.33 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:55:01,574 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3600, 3.3409, 2.1362, 3.5627, 2.6963, 3.5395, 2.3141, 2.8469], device='cuda:2'), covar=tensor([0.0323, 0.0514, 0.1587, 0.0383, 0.0789, 0.0815, 0.1340, 0.0686], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0170, 0.0178, 0.0217, 0.0204, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:55:21,739 INFO [optim.py:368] (2/8) 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 11:55:58,905 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:56:00,838 INFO [train.py:904] (2/8) Epoch 28, batch 550, loss[loss=0.1584, simple_loss=0.2439, pruned_loss=0.03644, over 16811.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2485, pruned_loss=0.03774, over 3107211.26 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:56:21,046 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 11:56:30,884 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274625.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:56:41,956 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7690, 3.9874, 2.6070, 4.5962, 3.1620, 4.4957, 2.7021, 3.3491], device='cuda:2'), covar=tensor([0.0347, 0.0413, 0.1637, 0.0292, 0.0888, 0.0550, 0.1515, 0.0744], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0172, 0.0179, 0.0219, 0.0206, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 11:56:56,831 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 11:57:09,874 INFO [train.py:904] (2/8) Epoch 28, batch 600, loss[loss=0.1635, simple_loss=0.2407, pruned_loss=0.04315, over 16705.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2482, pruned_loss=0.03753, over 3160441.25 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:57:15,322 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:17,484 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:36,633 INFO [optim.py:368] (2/8) 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 11:58:16,066 INFO [train.py:904] (2/8) Epoch 28, batch 650, loss[loss=0.1497, simple_loss=0.2555, pruned_loss=0.02196, over 17016.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2479, pruned_loss=0.03719, over 3201776.77 frames. ], batch size: 50, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:58:18,639 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:58:57,834 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0701, 2.2143, 2.7594, 3.0840, 2.9175, 3.5255, 2.5542, 3.4750], device='cuda:2'), covar=tensor([0.0292, 0.0598, 0.0354, 0.0387, 0.0382, 0.0233, 0.0542, 0.0223], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0193, 0.0208, 0.0167, 0.0206, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 11:59:25,786 INFO [train.py:904] (2/8) Epoch 28, batch 700, loss[loss=0.164, simple_loss=0.2461, pruned_loss=0.04093, over 16435.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2468, pruned_loss=0.03706, over 3221917.23 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:59:54,343 INFO [optim.py:368] (2/8) 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 12:00:13,339 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:00:34,708 INFO [train.py:904] (2/8) Epoch 28, batch 750, loss[loss=0.158, simple_loss=0.2518, pruned_loss=0.03212, over 17120.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2475, pruned_loss=0.03737, over 3243060.87 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 12:01:19,361 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:01:22,579 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0007, 2.8220, 2.9058, 5.1118, 4.0701, 4.4388, 1.9814, 3.3078], device='cuda:2'), covar=tensor([0.1307, 0.0817, 0.1114, 0.0156, 0.0224, 0.0416, 0.1489, 0.0740], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0179, 0.0200, 0.0200, 0.0204, 0.0218, 0.0209, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 12:01:45,403 INFO [train.py:904] (2/8) Epoch 28, batch 800, loss[loss=0.1677, simple_loss=0.2555, pruned_loss=0.03997, over 17021.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2477, pruned_loss=0.03696, over 3264195.01 frames. ], batch size: 50, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:02:15,002 INFO [optim.py:368] (2/8) 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 12:02:43,469 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:46,262 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274896.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:57,015 INFO [train.py:904] (2/8) Epoch 28, batch 850, loss[loss=0.145, simple_loss=0.2238, pruned_loss=0.0331, over 16328.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2478, pruned_loss=0.0371, over 3273460.03 frames. ], batch size: 165, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:03:26,697 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274925.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:05,739 INFO [train.py:904] (2/8) Epoch 28, batch 900, loss[loss=0.1538, simple_loss=0.2335, pruned_loss=0.03705, over 15462.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2468, pruned_loss=0.03676, over 3282607.02 frames. ], batch size: 190, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:04:09,193 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:13,707 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274959.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:19,895 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7482, 4.9490, 5.0181, 4.8343, 4.8920, 5.5231, 4.9897, 4.7113], device='cuda:2'), covar=tensor([0.1442, 0.2264, 0.2951, 0.2409, 0.2908, 0.1088, 0.2080, 0.2708], device='cuda:2'), in_proj_covar=tensor([0.0425, 0.0638, 0.0700, 0.0517, 0.0689, 0.0727, 0.0549, 0.0686], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:04:33,593 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:34,535 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 28, batch 950, loss[loss=0.1529, simple_loss=0.2382, pruned_loss=0.03379, over 15430.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2473, pruned_loss=0.03719, over 3285433.15 frames. ], batch size: 190, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:05:20,985 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275007.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:05:56,244 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4832, 4.5780, 4.8795, 4.8746, 4.8958, 4.6175, 4.5984, 4.4523], device='cuda:2'), covar=tensor([0.0497, 0.1137, 0.0578, 0.0550, 0.0581, 0.0571, 0.1040, 0.0813], device='cuda:2'), in_proj_covar=tensor([0.0439, 0.0493, 0.0477, 0.0438, 0.0524, 0.0502, 0.0575, 0.0402], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 12:06:23,178 INFO [train.py:904] (2/8) Epoch 28, batch 1000, loss[loss=0.1573, simple_loss=0.252, pruned_loss=0.0313, over 17036.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2465, pruned_loss=0.03696, over 3286150.96 frames. ], batch size: 50, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:06:52,342 INFO [optim.py:368] (2/8) 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 12:07:31,780 INFO [train.py:904] (2/8) Epoch 28, batch 1050, loss[loss=0.1564, simple_loss=0.2442, pruned_loss=0.03431, over 17207.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2464, pruned_loss=0.03684, over 3299511.56 frames. ], batch size: 44, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:09,134 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5707, 3.5824, 4.2108, 2.2194, 3.3145, 2.6324, 3.9881, 3.8151], device='cuda:2'), covar=tensor([0.0262, 0.1020, 0.0461, 0.2187, 0.0801, 0.0983, 0.0559, 0.1153], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0157, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 12:08:39,619 INFO [train.py:904] (2/8) Epoch 28, batch 1100, loss[loss=0.1677, simple_loss=0.2482, pruned_loss=0.04362, over 16737.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2459, pruned_loss=0.03659, over 3302361.77 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:47,404 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:07,965 INFO [optim.py:368] (2/8) 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 12:09:39,576 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:49,907 INFO [train.py:904] (2/8) Epoch 28, batch 1150, loss[loss=0.1455, simple_loss=0.2346, pruned_loss=0.02823, over 17193.00 frames. ], tot_loss[loss=0.1584, simple_loss=0.245, pruned_loss=0.03588, over 3312821.99 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:09:52,240 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 12:10:10,259 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2373, 5.2334, 4.9227, 4.3649, 5.0622, 1.8163, 4.8101, 4.7180], device='cuda:2'), covar=tensor([0.0105, 0.0089, 0.0231, 0.0422, 0.0107, 0.3061, 0.0148, 0.0261], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0174, 0.0210, 0.0182, 0.0187, 0.0217, 0.0200, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:10:11,462 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:30,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8006, 2.4749, 2.0656, 2.3089, 2.8173, 2.5588, 2.7587, 2.9348], device='cuda:2'), covar=tensor([0.0305, 0.0551, 0.0617, 0.0556, 0.0307, 0.0441, 0.0258, 0.0329], device='cuda:2'), in_proj_covar=tensor([0.0234, 0.0250, 0.0238, 0.0238, 0.0251, 0.0248, 0.0247, 0.0247], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:10:44,955 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:53,440 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275250.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:57,366 INFO [train.py:904] (2/8) Epoch 28, batch 1200, loss[loss=0.1768, simple_loss=0.2495, pruned_loss=0.05201, over 16459.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2438, pruned_loss=0.03589, over 3310489.82 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:11:20,268 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3774, 3.5413, 3.7197, 2.6219, 3.3540, 3.8378, 3.5539, 2.1978], device='cuda:2'), covar=tensor([0.0553, 0.0188, 0.0069, 0.0411, 0.0143, 0.0101, 0.0105, 0.0536], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0102, 0.0114, 0.0098, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 12:11:21,640 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 12:11:26,948 INFO [optim.py:368] (2/8) 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 12:12:06,865 INFO [train.py:904] (2/8) Epoch 28, batch 1250, loss[loss=0.1803, simple_loss=0.2814, pruned_loss=0.03959, over 17024.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.2436, pruned_loss=0.03652, over 3309519.79 frames. ], batch size: 53, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:12:12,676 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:08,405 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275347.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:13:17,209 INFO [train.py:904] (2/8) Epoch 28, batch 1300, loss[loss=0.1425, simple_loss=0.228, pruned_loss=0.0285, over 16840.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.2436, pruned_loss=0.03629, over 3304557.59 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:13:39,138 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275368.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:46,872 INFO [optim.py:368] (2/8) 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 12:14:02,897 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275385.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:14:26,876 INFO [train.py:904] (2/8) Epoch 28, batch 1350, loss[loss=0.1637, simple_loss=0.2452, pruned_loss=0.04107, over 16460.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.244, pruned_loss=0.03653, over 3295894.62 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:14:35,357 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275408.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:15:28,637 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:15:37,727 INFO [train.py:904] (2/8) Epoch 28, batch 1400, loss[loss=0.1563, simple_loss=0.2364, pruned_loss=0.03815, over 16847.00 frames. ], tot_loss[loss=0.1584, simple_loss=0.2442, pruned_loss=0.03628, over 3298647.45 frames. ], batch size: 96, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:16:06,674 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 28, batch 1450, loss[loss=0.1503, simple_loss=0.2387, pruned_loss=0.03096, over 17239.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2439, pruned_loss=0.03663, over 3310068.41 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:17:03,271 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:05,692 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9921, 2.2497, 2.5714, 2.8803, 2.8155, 3.1646, 2.2497, 3.2347], device='cuda:2'), covar=tensor([0.0281, 0.0509, 0.0371, 0.0379, 0.0384, 0.0260, 0.0574, 0.0220], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0200, 0.0189, 0.0194, 0.0209, 0.0168, 0.0205, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 12:17:53,972 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:57,211 INFO [train.py:904] (2/8) Epoch 28, batch 1500, loss[loss=0.1693, simple_loss=0.2447, pruned_loss=0.04701, over 16895.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2439, pruned_loss=0.03653, over 3315835.86 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:18:26,332 INFO [optim.py:368] (2/8) 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 12:19:01,050 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:19:07,363 INFO [train.py:904] (2/8) Epoch 28, batch 1550, loss[loss=0.1763, simple_loss=0.2647, pruned_loss=0.04396, over 15519.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2446, pruned_loss=0.03698, over 3302764.36 frames. ], batch size: 191, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:19:20,787 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9636, 2.2399, 2.5682, 2.8716, 2.7997, 3.0693, 2.2898, 3.2589], device='cuda:2'), covar=tensor([0.0245, 0.0460, 0.0335, 0.0316, 0.0360, 0.0255, 0.0549, 0.0177], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0193, 0.0209, 0.0167, 0.0205, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:20:18,408 INFO [train.py:904] (2/8) Epoch 28, batch 1600, loss[loss=0.1385, simple_loss=0.2224, pruned_loss=0.02728, over 17007.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2468, pruned_loss=0.03746, over 3300263.82 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:20:33,168 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275663.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:20:48,274 INFO [optim.py:368] (2/8) 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 12:21:04,022 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4676, 5.4169, 5.3329, 4.7850, 4.9043, 5.3761, 5.2867, 4.9656], device='cuda:2'), covar=tensor([0.0677, 0.0624, 0.0359, 0.0377, 0.1221, 0.0565, 0.0309, 0.0867], device='cuda:2'), in_proj_covar=tensor([0.0318, 0.0477, 0.0372, 0.0374, 0.0369, 0.0428, 0.0255, 0.0445], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:21:28,991 INFO [train.py:904] (2/8) Epoch 28, batch 1650, loss[loss=0.1475, simple_loss=0.2433, pruned_loss=0.02581, over 17095.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2486, pruned_loss=0.0381, over 3297634.49 frames. ], batch size: 47, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:21:29,293 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275703.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 12:21:52,568 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6682, 4.6184, 4.5388, 3.7521, 4.6142, 1.6935, 4.3224, 4.2010], device='cuda:2'), covar=tensor([0.0208, 0.0164, 0.0256, 0.0569, 0.0154, 0.3304, 0.0241, 0.0357], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0175, 0.0212, 0.0185, 0.0189, 0.0219, 0.0202, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:22:00,692 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3071, 2.3401, 2.3475, 4.0039, 2.2825, 2.7426, 2.4110, 2.4690], device='cuda:2'), covar=tensor([0.1580, 0.4064, 0.3382, 0.0667, 0.4359, 0.2803, 0.3777, 0.3796], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0474, 0.0388, 0.0338, 0.0446, 0.0543, 0.0446, 0.0554], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:22:02,060 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 12:22:21,191 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275741.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:22:37,918 INFO [train.py:904] (2/8) Epoch 28, batch 1700, loss[loss=0.1536, simple_loss=0.2518, pruned_loss=0.02768, over 17216.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2498, pruned_loss=0.03813, over 3307717.71 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:22:49,421 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2869, 4.2950, 4.6480, 4.6186, 4.6764, 4.3995, 4.3847, 4.2811], device='cuda:2'), covar=tensor([0.0386, 0.0727, 0.0388, 0.0450, 0.0500, 0.0458, 0.0869, 0.0684], device='cuda:2'), in_proj_covar=tensor([0.0439, 0.0492, 0.0475, 0.0439, 0.0523, 0.0502, 0.0576, 0.0402], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 12:23:01,396 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1594, 4.1823, 4.5009, 4.4832, 4.5319, 4.2669, 4.2557, 4.1765], device='cuda:2'), covar=tensor([0.0396, 0.0715, 0.0423, 0.0432, 0.0544, 0.0447, 0.0892, 0.0675], device='cuda:2'), in_proj_covar=tensor([0.0439, 0.0493, 0.0475, 0.0439, 0.0524, 0.0503, 0.0577, 0.0402], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 12:23:08,710 INFO [optim.py:368] (2/8) 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 12:23:14,351 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 12:23:49,162 INFO [train.py:904] (2/8) Epoch 28, batch 1750, loss[loss=0.1555, simple_loss=0.2461, pruned_loss=0.03242, over 17211.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2506, pruned_loss=0.0381, over 3321490.23 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:24:05,546 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:24:59,129 INFO [train.py:904] (2/8) Epoch 28, batch 1800, loss[loss=0.1608, simple_loss=0.2557, pruned_loss=0.03295, over 17182.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2504, pruned_loss=0.03767, over 3315740.35 frames. ], batch size: 46, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:25:08,985 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275862.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:19,200 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4440, 4.2529, 4.4327, 4.6580, 4.7594, 4.3626, 4.6136, 4.7616], device='cuda:2'), covar=tensor([0.1966, 0.1596, 0.1958, 0.0939, 0.0788, 0.1312, 0.2564, 0.1200], device='cuda:2'), in_proj_covar=tensor([0.0691, 0.0842, 0.0980, 0.0856, 0.0652, 0.0683, 0.0718, 0.0831], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:25:30,537 INFO [optim.py:368] (2/8) 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 12:25:59,919 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3722, 4.3012, 4.2918, 3.9952, 4.0765, 4.3241, 4.0519, 4.1120], device='cuda:2'), covar=tensor([0.0649, 0.0882, 0.0314, 0.0312, 0.0706, 0.0622, 0.0687, 0.0680], device='cuda:2'), in_proj_covar=tensor([0.0318, 0.0478, 0.0372, 0.0374, 0.0370, 0.0429, 0.0255, 0.0446], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:26:08,320 INFO [train.py:904] (2/8) Epoch 28, batch 1850, loss[loss=0.1544, simple_loss=0.2428, pruned_loss=0.03293, over 16662.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2512, pruned_loss=0.03776, over 3324586.09 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:26:33,061 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275921.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:16,689 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275952.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:17,450 INFO [train.py:904] (2/8) Epoch 28, batch 1900, loss[loss=0.1423, simple_loss=0.2327, pruned_loss=0.02597, over 16013.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2508, pruned_loss=0.03736, over 3314616.32 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:27:31,521 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275963.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:47,686 INFO [optim.py:368] (2/8) 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 12:28:30,501 INFO [train.py:904] (2/8) Epoch 28, batch 1950, loss[loss=0.1784, simple_loss=0.2585, pruned_loss=0.04913, over 16914.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2504, pruned_loss=0.03711, over 3321578.02 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:28:30,897 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:28:43,428 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:28:56,247 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9077, 4.3837, 3.1457, 2.3401, 2.6925, 2.7615, 4.8075, 3.5985], device='cuda:2'), covar=tensor([0.2951, 0.0617, 0.1884, 0.3259, 0.3173, 0.2109, 0.0382, 0.1533], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0276, 0.0313, 0.0327, 0.0305, 0.0278, 0.0305, 0.0353], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:29:24,692 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:36,494 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2831, 5.6944, 5.8229, 5.4806, 5.6384, 6.1975, 5.6492, 5.4193], device='cuda:2'), covar=tensor([0.0925, 0.2130, 0.2356, 0.2280, 0.2629, 0.1059, 0.1568, 0.2230], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0648, 0.0712, 0.0530, 0.0699, 0.0740, 0.0555, 0.0700], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:29:37,515 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276051.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:29:40,564 INFO [train.py:904] (2/8) Epoch 28, batch 2000, loss[loss=0.1379, simple_loss=0.226, pruned_loss=0.02489, over 16842.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2513, pruned_loss=0.03747, over 3307644.41 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:29:56,398 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9329, 4.9050, 4.7979, 4.2624, 4.8693, 1.9829, 4.6324, 4.6188], device='cuda:2'), covar=tensor([0.0204, 0.0163, 0.0255, 0.0426, 0.0149, 0.3099, 0.0204, 0.0265], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0177, 0.0214, 0.0187, 0.0190, 0.0220, 0.0203, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:30:11,362 INFO [optim.py:368] (2/8) 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 12:30:31,277 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276089.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:30:32,510 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1965, 5.2310, 5.6323, 5.5984, 5.6567, 5.3198, 5.2258, 5.0785], device='cuda:2'), covar=tensor([0.0361, 0.0684, 0.0379, 0.0481, 0.0480, 0.0397, 0.0992, 0.0454], device='cuda:2'), in_proj_covar=tensor([0.0442, 0.0496, 0.0480, 0.0442, 0.0528, 0.0506, 0.0581, 0.0405], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 12:30:39,259 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-02 12:30:50,228 INFO [train.py:904] (2/8) Epoch 28, batch 2050, loss[loss=0.1626, simple_loss=0.2568, pruned_loss=0.0342, over 16509.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2514, pruned_loss=0.03786, over 3308222.22 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:32:00,358 INFO [train.py:904] (2/8) Epoch 28, batch 2100, loss[loss=0.1521, simple_loss=0.2437, pruned_loss=0.03027, over 17239.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2523, pruned_loss=0.0384, over 3310033.13 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:32:30,698 INFO [optim.py:368] (2/8) 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 12:33:09,370 INFO [train.py:904] (2/8) Epoch 28, batch 2150, loss[loss=0.1662, simple_loss=0.2521, pruned_loss=0.04017, over 16819.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2537, pruned_loss=0.03924, over 3313084.46 frames. ], batch size: 83, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:33:27,937 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276216.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:34:18,059 INFO [train.py:904] (2/8) Epoch 28, batch 2200, loss[loss=0.1775, simple_loss=0.2663, pruned_loss=0.04432, over 17063.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2547, pruned_loss=0.04016, over 3318896.50 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:34:50,520 INFO [optim.py:368] (2/8) 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 12:35:27,983 INFO [train.py:904] (2/8) Epoch 28, batch 2250, loss[loss=0.1655, simple_loss=0.2444, pruned_loss=0.04333, over 16723.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2551, pruned_loss=0.03976, over 3317284.72 frames. ], batch size: 83, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:35:35,327 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:35:48,028 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:36:37,084 INFO [train.py:904] (2/8) Epoch 28, batch 2300, loss[loss=0.1682, simple_loss=0.2668, pruned_loss=0.03485, over 16696.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2543, pruned_loss=0.03921, over 3319981.95 frames. ], batch size: 62, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:36:54,919 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8228, 2.7701, 2.5845, 4.6471, 3.6204, 4.1529, 1.7816, 3.0709], device='cuda:2'), covar=tensor([0.1464, 0.0893, 0.1320, 0.0304, 0.0281, 0.0474, 0.1697, 0.0843], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0204, 0.0207, 0.0220, 0.0210, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 12:37:08,703 INFO [optim.py:368] (2/8) 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 12:37:12,900 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276378.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:37:46,502 INFO [train.py:904] (2/8) Epoch 28, batch 2350, loss[loss=0.1788, simple_loss=0.2778, pruned_loss=0.0399, over 17035.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2551, pruned_loss=0.03992, over 3317966.57 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:38:09,039 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3246, 3.4453, 3.6031, 2.6287, 3.2586, 3.7113, 3.4304, 2.2904], device='cuda:2'), covar=tensor([0.0530, 0.0152, 0.0070, 0.0377, 0.0149, 0.0100, 0.0110, 0.0494], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 12:38:54,360 INFO [train.py:904] (2/8) Epoch 28, batch 2400, loss[loss=0.1551, simple_loss=0.2497, pruned_loss=0.03031, over 16844.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2546, pruned_loss=0.03955, over 3330004.30 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:39:16,535 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2209, 5.7758, 5.9191, 5.5679, 5.7274, 6.2585, 5.7461, 5.3916], device='cuda:2'), covar=tensor([0.0917, 0.1856, 0.2357, 0.2196, 0.2460, 0.0917, 0.1526, 0.2528], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0650, 0.0716, 0.0531, 0.0701, 0.0740, 0.0556, 0.0702], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:39:26,374 INFO [optim.py:368] (2/8) 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 12:39:59,692 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 12:40:04,311 INFO [train.py:904] (2/8) Epoch 28, batch 2450, loss[loss=0.1805, simple_loss=0.2675, pruned_loss=0.04674, over 16724.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.255, pruned_loss=0.03926, over 3329864.48 frames. ], batch size: 89, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:40:23,745 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276516.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:40:58,057 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0722, 5.1575, 4.9887, 4.4949, 4.4642, 5.0890, 5.0333, 4.6273], device='cuda:2'), covar=tensor([0.0748, 0.0614, 0.0438, 0.0502, 0.1418, 0.0585, 0.0384, 0.0878], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0482, 0.0375, 0.0378, 0.0371, 0.0432, 0.0257, 0.0450], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:41:13,994 INFO [train.py:904] (2/8) Epoch 28, batch 2500, loss[loss=0.1843, simple_loss=0.2632, pruned_loss=0.05273, over 16776.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2552, pruned_loss=0.03893, over 3334249.92 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:41:24,533 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6336, 4.0928, 4.5872, 2.6699, 4.7171, 4.8533, 3.6357, 3.9967], device='cuda:2'), covar=tensor([0.0529, 0.0274, 0.0197, 0.1034, 0.0082, 0.0187, 0.0382, 0.0317], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0110, 0.0102, 0.0139, 0.0086, 0.0131, 0.0129, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 12:41:30,324 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:41:45,567 INFO [optim.py:368] (2/8) 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 12:42:24,130 INFO [train.py:904] (2/8) Epoch 28, batch 2550, loss[loss=0.1407, simple_loss=0.2337, pruned_loss=0.02384, over 17199.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2546, pruned_loss=0.03863, over 3338952.71 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:42:31,779 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:42:34,410 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2936, 3.4761, 3.7340, 2.1792, 3.1201, 2.4204, 3.7031, 3.6657], device='cuda:2'), covar=tensor([0.0300, 0.0929, 0.0577, 0.2115, 0.0890, 0.1032, 0.0631, 0.1125], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0158, 0.0149, 0.0134, 0.0148, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 12:42:35,374 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276611.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:42:45,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8704, 4.0381, 2.6203, 4.7043, 3.1445, 4.5904, 2.8656, 3.4285], device='cuda:2'), covar=tensor([0.0347, 0.0405, 0.1586, 0.0313, 0.0892, 0.0553, 0.1477, 0.0698], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0186, 0.0200, 0.0177, 0.0183, 0.0226, 0.0208, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 12:43:33,174 INFO [train.py:904] (2/8) Epoch 28, batch 2600, loss[loss=0.1571, simple_loss=0.2432, pruned_loss=0.03548, over 16966.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2549, pruned_loss=0.03866, over 3327203.95 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:43:37,973 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276656.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:51,616 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4064, 5.3672, 5.2575, 4.7450, 4.8892, 5.3031, 5.2467, 4.9094], device='cuda:2'), covar=tensor([0.0564, 0.0567, 0.0310, 0.0343, 0.1024, 0.0461, 0.0309, 0.0786], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0482, 0.0376, 0.0378, 0.0372, 0.0433, 0.0257, 0.0451], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 12:43:55,728 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:01,047 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276672.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:01,969 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276673.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:05,671 INFO [optim.py:368] (2/8) 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 12:44:11,704 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 12:44:43,681 INFO [train.py:904] (2/8) Epoch 28, batch 2650, loss[loss=0.156, simple_loss=0.2458, pruned_loss=0.03312, over 17120.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2551, pruned_loss=0.03843, over 3328665.57 frames. ], batch size: 47, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:45:19,180 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:45:50,738 INFO [train.py:904] (2/8) Epoch 28, batch 2700, loss[loss=0.1632, simple_loss=0.2523, pruned_loss=0.03705, over 16053.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03843, over 3335906.57 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:46:14,449 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 12:46:23,524 INFO [optim.py:368] (2/8) 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 12:46:42,557 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:47:00,577 INFO [train.py:904] (2/8) Epoch 28, batch 2750, loss[loss=0.1628, simple_loss=0.2609, pruned_loss=0.03234, over 17219.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2556, pruned_loss=0.03781, over 3337135.42 frames. ], batch size: 46, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:07,254 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276851.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:09,124 INFO [train.py:904] (2/8) Epoch 28, batch 2800, loss[loss=0.1584, simple_loss=0.2513, pruned_loss=0.03277, over 17231.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.256, pruned_loss=0.03785, over 3334889.09 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:24,924 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276863.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:41,263 INFO [optim.py:368] (2/8) 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 12:49:20,035 INFO [train.py:904] (2/8) Epoch 28, batch 2850, loss[loss=0.168, simple_loss=0.2497, pruned_loss=0.04314, over 16543.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2556, pruned_loss=0.0378, over 3331226.87 frames. ], batch size: 75, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:49:50,185 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276924.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:49:57,778 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:27,302 INFO [train.py:904] (2/8) Epoch 28, batch 2900, loss[loss=0.1774, simple_loss=0.2508, pruned_loss=0.05203, over 16887.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2544, pruned_loss=0.0382, over 3321380.96 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:50:30,775 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 12:50:46,671 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:54,972 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:56,673 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-02 12:50:58,047 INFO [optim.py:368] (2/8) 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 12:51:21,999 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276992.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 12:51:36,311 INFO [train.py:904] (2/8) Epoch 28, batch 2950, loss[loss=0.1698, simple_loss=0.2571, pruned_loss=0.04129, over 17205.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2548, pruned_loss=0.0394, over 3317820.54 frames. ], batch size: 46, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:51:48,560 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:00,581 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:05,335 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:45,413 INFO [train.py:904] (2/8) Epoch 28, batch 3000, loss[loss=0.1786, simple_loss=0.2598, pruned_loss=0.04864, over 16394.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2542, pruned_loss=0.03946, over 3317320.35 frames. ], batch size: 146, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:52:45,413 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 12:52:54,783 INFO [train.py:938] (2/8) Epoch 28, validation: loss=0.1335, simple_loss=0.2385, pruned_loss=0.01427, over 944034.00 frames. 2023-05-02 12:52:54,784 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 12:53:21,167 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277072.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:53:25,662 INFO [optim.py:368] (2/8) 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 12:54:02,141 INFO [train.py:904] (2/8) Epoch 28, batch 3050, loss[loss=0.178, simple_loss=0.2762, pruned_loss=0.03989, over 16640.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2545, pruned_loss=0.03975, over 3315697.21 frames. ], batch size: 62, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:54:06,413 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 12:54:32,909 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6410, 4.6092, 4.5290, 3.9668, 4.5867, 1.7700, 4.3027, 4.1286], device='cuda:2'), covar=tensor([0.0158, 0.0129, 0.0202, 0.0369, 0.0119, 0.3086, 0.0181, 0.0260], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0179, 0.0217, 0.0190, 0.0193, 0.0223, 0.0206, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 12:55:01,328 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277146.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:55:11,783 INFO [train.py:904] (2/8) Epoch 28, batch 3100, loss[loss=0.1589, simple_loss=0.2473, pruned_loss=0.0352, over 16417.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2537, pruned_loss=0.03963, over 3321969.86 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:55:43,285 INFO [optim.py:368] (2/8) 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 12:55:45,411 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:21,102 INFO [train.py:904] (2/8) Epoch 28, batch 3150, loss[loss=0.1586, simple_loss=0.2537, pruned_loss=0.03179, over 17039.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2528, pruned_loss=0.03937, over 3326787.89 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:56:44,543 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:44,649 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:08,732 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-02 12:57:10,209 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277238.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:30,727 INFO [train.py:904] (2/8) Epoch 28, batch 3200, loss[loss=0.1413, simple_loss=0.2222, pruned_loss=0.03014, over 17006.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2516, pruned_loss=0.03868, over 3327455.73 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:57:32,609 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-05-02 12:57:50,762 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277267.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:01,813 INFO [optim.py:368] (2/8) 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 12:58:07,134 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277280.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:17,847 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277287.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:58:39,067 INFO [train.py:904] (2/8) Epoch 28, batch 3250, loss[loss=0.1754, simple_loss=0.2666, pruned_loss=0.04209, over 17134.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2518, pruned_loss=0.03857, over 3335002.13 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:58:42,472 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:09,109 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:28,970 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9677, 2.7063, 2.6032, 4.2532, 3.4153, 4.1103, 1.7499, 2.9970], device='cuda:2'), covar=tensor([0.1330, 0.0746, 0.1167, 0.0174, 0.0148, 0.0382, 0.1558, 0.0814], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0205, 0.0207, 0.0220, 0.0210, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 12:59:47,703 INFO [train.py:904] (2/8) Epoch 28, batch 3300, loss[loss=0.1538, simple_loss=0.2377, pruned_loss=0.03492, over 16754.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.252, pruned_loss=0.03863, over 3330734.30 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:00:05,286 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:06,235 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277367.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:09,453 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0695, 2.1755, 2.2972, 3.7000, 2.1776, 2.5157, 2.2507, 2.3058], device='cuda:2'), covar=tensor([0.1659, 0.3857, 0.3132, 0.0744, 0.3986, 0.2611, 0.3955, 0.3132], device='cuda:2'), in_proj_covar=tensor([0.0427, 0.0479, 0.0391, 0.0342, 0.0449, 0.0550, 0.0451, 0.0561], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:00:12,572 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 13:00:55,902 INFO [train.py:904] (2/8) Epoch 28, batch 3350, loss[loss=0.1562, simple_loss=0.2427, pruned_loss=0.03483, over 16809.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2531, pruned_loss=0.03914, over 3318799.20 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:01:16,991 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 13:01:54,972 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:02:03,962 INFO [train.py:904] (2/8) Epoch 28, batch 3400, loss[loss=0.1944, simple_loss=0.2823, pruned_loss=0.05323, over 12361.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2534, pruned_loss=0.03913, over 3311149.60 frames. ], batch size: 248, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:02:34,599 INFO [optim.py:368] (2/8) 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 13:02:39,760 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4197, 4.2878, 4.4751, 4.6270, 4.7211, 4.3251, 4.5373, 4.7315], device='cuda:2'), covar=tensor([0.1962, 0.1334, 0.1581, 0.0854, 0.0741, 0.1260, 0.2723, 0.1125], device='cuda:2'), in_proj_covar=tensor([0.0709, 0.0865, 0.1006, 0.0881, 0.0669, 0.0703, 0.0736, 0.0849], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:03:01,407 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:13,660 INFO [train.py:904] (2/8) Epoch 28, batch 3450, loss[loss=0.1734, simple_loss=0.2672, pruned_loss=0.03982, over 17072.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2512, pruned_loss=0.03829, over 3317601.48 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:03:35,940 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277519.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:52,855 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5175, 3.6325, 4.1100, 2.1380, 3.3484, 2.4553, 4.0636, 3.8987], device='cuda:2'), covar=tensor([0.0261, 0.0959, 0.0494, 0.2293, 0.0820, 0.1078, 0.0526, 0.0936], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0174, 0.0172, 0.0159, 0.0150, 0.0134, 0.0148, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 13:03:55,440 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:15,602 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 13:04:23,507 INFO [train.py:904] (2/8) Epoch 28, batch 3500, loss[loss=0.1873, simple_loss=0.2763, pruned_loss=0.04913, over 15388.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2502, pruned_loss=0.03799, over 3313678.62 frames. ], batch size: 191, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:04:25,852 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:42,185 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:47,004 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:54,522 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:55,301 INFO [optim.py:368] (2/8) 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 13:05:10,640 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277587.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 13:05:13,700 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9024, 4.7911, 4.7894, 4.4199, 4.4959, 4.8583, 4.6769, 4.5888], device='cuda:2'), covar=tensor([0.0778, 0.0998, 0.0610, 0.0436, 0.1102, 0.0652, 0.0509, 0.0807], device='cuda:2'), in_proj_covar=tensor([0.0326, 0.0491, 0.0382, 0.0384, 0.0379, 0.0440, 0.0260, 0.0457], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 13:05:24,434 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8548, 2.5760, 2.5796, 4.0782, 3.3402, 4.1234, 1.5975, 2.9492], device='cuda:2'), covar=tensor([0.1412, 0.0782, 0.1206, 0.0196, 0.0136, 0.0381, 0.1673, 0.0846], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0206, 0.0208, 0.0221, 0.0210, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 13:05:32,630 INFO [train.py:904] (2/8) Epoch 28, batch 3550, loss[loss=0.1554, simple_loss=0.24, pruned_loss=0.03542, over 16561.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2493, pruned_loss=0.03727, over 3323258.07 frames. ], batch size: 75, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:05:49,370 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277615.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:11,024 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277631.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:16,855 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:34,496 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 13:06:42,325 INFO [train.py:904] (2/8) Epoch 28, batch 3600, loss[loss=0.1566, simple_loss=0.2423, pruned_loss=0.03547, over 16412.00 frames. ], tot_loss[loss=0.162, simple_loss=0.249, pruned_loss=0.03751, over 3325256.88 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:06:52,860 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:02,018 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277667.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:14,463 INFO [optim.py:368] (2/8) 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 13:07:36,945 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:53,647 INFO [train.py:904] (2/8) Epoch 28, batch 3650, loss[loss=0.1484, simple_loss=0.2242, pruned_loss=0.03626, over 16376.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2484, pruned_loss=0.03806, over 3311810.38 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:08:12,450 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:09:06,604 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277752.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 13:09:07,228 INFO [train.py:904] (2/8) Epoch 28, batch 3700, loss[loss=0.168, simple_loss=0.2405, pruned_loss=0.04777, over 16472.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2476, pruned_loss=0.03957, over 3306880.86 frames. ], batch size: 75, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:09:22,150 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8657, 2.2627, 2.4944, 3.1187, 2.3077, 2.4057, 2.4828, 2.3344], device='cuda:2'), covar=tensor([0.1422, 0.3196, 0.2581, 0.0762, 0.3848, 0.2232, 0.2883, 0.3382], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0482, 0.0392, 0.0344, 0.0451, 0.0553, 0.0453, 0.0563], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:09:41,445 INFO [optim.py:368] (2/8) 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 13:10:09,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3808, 3.5828, 3.7432, 2.4845, 3.4523, 3.8887, 3.5818, 2.2243], device='cuda:2'), covar=tensor([0.0569, 0.0190, 0.0072, 0.0442, 0.0127, 0.0095, 0.0108, 0.0508], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0091, 0.0091, 0.0135, 0.0103, 0.0115, 0.0099, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 13:10:22,496 INFO [train.py:904] (2/8) Epoch 28, batch 3750, loss[loss=0.1838, simple_loss=0.2581, pruned_loss=0.05479, over 16812.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.249, pruned_loss=0.04105, over 3267597.19 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:10:27,414 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2281, 3.2891, 3.5622, 2.3219, 3.0387, 2.4258, 3.7748, 3.6764], device='cuda:2'), covar=tensor([0.0224, 0.0918, 0.0582, 0.1922, 0.0839, 0.1002, 0.0454, 0.0815], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0174, 0.0171, 0.0158, 0.0150, 0.0133, 0.0148, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 13:11:06,890 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:11:35,741 INFO [train.py:904] (2/8) Epoch 28, batch 3800, loss[loss=0.1929, simple_loss=0.2721, pruned_loss=0.05686, over 15653.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2502, pruned_loss=0.04219, over 3267305.14 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:08,762 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:11,083 INFO [optim.py:368] (2/8) 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 13:12:17,663 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:48,151 INFO [train.py:904] (2/8) Epoch 28, batch 3850, loss[loss=0.1537, simple_loss=0.2331, pruned_loss=0.03717, over 16685.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2495, pruned_loss=0.04276, over 3279742.81 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:59,661 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:18,278 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:54,793 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4890, 4.3623, 4.5364, 4.6755, 4.7958, 4.3845, 4.6198, 4.7990], device='cuda:2'), covar=tensor([0.1851, 0.1386, 0.1653, 0.0892, 0.0710, 0.1036, 0.2496, 0.1048], device='cuda:2'), in_proj_covar=tensor([0.0705, 0.0859, 0.0998, 0.0876, 0.0665, 0.0697, 0.0730, 0.0844], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:14:00,814 INFO [train.py:904] (2/8) Epoch 28, batch 3900, loss[loss=0.1496, simple_loss=0.2308, pruned_loss=0.03414, over 16839.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2491, pruned_loss=0.04306, over 3280992.63 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:14:05,040 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 13:14:13,180 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277961.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:27,568 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:36,142 INFO [optim.py:368] (2/8) 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 13:15:16,801 INFO [train.py:904] (2/8) Epoch 28, batch 3950, loss[loss=0.1649, simple_loss=0.2504, pruned_loss=0.03969, over 16528.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2484, pruned_loss=0.04372, over 3281059.44 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:15:25,692 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278009.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:15:54,379 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4111, 4.3781, 4.3918, 3.8314, 4.3884, 1.7808, 4.1583, 3.8998], device='cuda:2'), covar=tensor([0.0165, 0.0145, 0.0202, 0.0306, 0.0115, 0.3016, 0.0164, 0.0270], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0179, 0.0218, 0.0191, 0.0194, 0.0223, 0.0207, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:15:58,205 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278032.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:16:04,644 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6852, 4.6154, 4.5941, 3.9722, 4.6425, 1.7616, 4.3939, 4.1732], device='cuda:2'), covar=tensor([0.0201, 0.0202, 0.0210, 0.0406, 0.0125, 0.3062, 0.0206, 0.0276], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0179, 0.0218, 0.0191, 0.0194, 0.0223, 0.0207, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:16:19,768 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278047.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 13:16:26,864 INFO [train.py:904] (2/8) Epoch 28, batch 4000, loss[loss=0.1601, simple_loss=0.2451, pruned_loss=0.0375, over 16845.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2488, pruned_loss=0.04427, over 3283939.11 frames. ], batch size: 90, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:17:00,795 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4136, 5.7340, 5.4693, 5.5744, 5.2241, 5.0869, 5.1739, 5.8890], device='cuda:2'), covar=tensor([0.1442, 0.0901, 0.1036, 0.0888, 0.0846, 0.0745, 0.1163, 0.0786], device='cuda:2'), in_proj_covar=tensor([0.0737, 0.0901, 0.0733, 0.0691, 0.0567, 0.0563, 0.0753, 0.0699], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:17:01,538 INFO [optim.py:368] (2/8) 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 13:17:37,863 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7116, 4.4958, 4.3500, 2.8311, 3.8230, 4.4117, 3.8083, 2.7199], device='cuda:2'), covar=tensor([0.0525, 0.0033, 0.0049, 0.0408, 0.0109, 0.0074, 0.0115, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0090, 0.0091, 0.0135, 0.0103, 0.0115, 0.0099, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 13:17:38,523 INFO [train.py:904] (2/8) Epoch 28, batch 4050, loss[loss=0.1653, simple_loss=0.2492, pruned_loss=0.0407, over 17040.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2498, pruned_loss=0.04381, over 3275476.99 frames. ], batch size: 55, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:18:52,741 INFO [train.py:904] (2/8) Epoch 28, batch 4100, loss[loss=0.1985, simple_loss=0.2855, pruned_loss=0.05579, over 15286.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2518, pruned_loss=0.0434, over 3277546.01 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:19:10,521 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6693, 2.8266, 2.3237, 2.7214, 3.0980, 2.7875, 3.1611, 3.2622], device='cuda:2'), covar=tensor([0.0108, 0.0442, 0.0592, 0.0440, 0.0280, 0.0396, 0.0258, 0.0300], device='cuda:2'), in_proj_covar=tensor([0.0238, 0.0251, 0.0238, 0.0238, 0.0251, 0.0249, 0.0250, 0.0249], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:19:19,936 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 13:19:29,061 INFO [optim.py:368] (2/8) 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 13:19:29,563 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6048, 4.8485, 4.6350, 4.6482, 4.4272, 4.3248, 4.3301, 4.9181], device='cuda:2'), covar=tensor([0.1144, 0.0862, 0.1020, 0.0910, 0.0779, 0.1502, 0.1149, 0.0813], device='cuda:2'), in_proj_covar=tensor([0.0731, 0.0895, 0.0727, 0.0686, 0.0563, 0.0559, 0.0747, 0.0694], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:20:09,078 INFO [train.py:904] (2/8) Epoch 28, batch 4150, loss[loss=0.1952, simple_loss=0.2874, pruned_loss=0.05147, over 15436.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2586, pruned_loss=0.04526, over 3269902.87 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:20:19,261 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:20:42,410 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:22,144 INFO [train.py:904] (2/8) Epoch 28, batch 4200, loss[loss=0.175, simple_loss=0.2724, pruned_loss=0.03879, over 17199.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2655, pruned_loss=0.04671, over 3240534.93 frames. ], batch size: 46, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:21:22,538 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6932, 5.0539, 5.3064, 5.2297, 5.2707, 4.9921, 4.6515, 4.6843], device='cuda:2'), covar=tensor([0.0590, 0.0525, 0.0477, 0.0622, 0.0671, 0.0587, 0.1512, 0.0609], device='cuda:2'), in_proj_covar=tensor([0.0440, 0.0497, 0.0477, 0.0442, 0.0526, 0.0504, 0.0580, 0.0406], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 13:21:30,673 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:43,939 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3345, 4.2200, 4.3866, 4.5576, 4.6735, 4.2610, 4.6842, 4.7403], device='cuda:2'), covar=tensor([0.1982, 0.1482, 0.1883, 0.0914, 0.0713, 0.1443, 0.0995, 0.0781], device='cuda:2'), in_proj_covar=tensor([0.0696, 0.0849, 0.0985, 0.0862, 0.0657, 0.0688, 0.0720, 0.0831], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:21:54,355 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278274.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:57,721 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 13:21:58,171 INFO [optim.py:368] (2/8) 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 13:22:25,921 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 13:22:37,393 INFO [train.py:904] (2/8) Epoch 28, batch 4250, loss[loss=0.1691, simple_loss=0.2648, pruned_loss=0.03675, over 17025.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2678, pruned_loss=0.04613, over 3205711.24 frames. ], batch size: 41, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:23:07,791 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:13,431 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:43,993 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278347.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 13:23:52,556 INFO [train.py:904] (2/8) Epoch 28, batch 4300, loss[loss=0.1887, simple_loss=0.283, pruned_loss=0.04723, over 16405.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2692, pruned_loss=0.04539, over 3210992.09 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:24:25,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5538, 3.5449, 3.4924, 2.7379, 3.2795, 2.0196, 3.0957, 2.7469], device='cuda:2'), covar=tensor([0.0175, 0.0184, 0.0225, 0.0271, 0.0125, 0.2787, 0.0146, 0.0291], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0179, 0.0217, 0.0190, 0.0193, 0.0222, 0.0206, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:24:25,609 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278374.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:29,318 INFO [optim.py:368] (2/8) 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 13:24:40,214 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:56,293 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:25:07,419 INFO [train.py:904] (2/8) Epoch 28, batch 4350, loss[loss=0.1999, simple_loss=0.2915, pruned_loss=0.05413, over 16721.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2727, pruned_loss=0.04674, over 3162896.65 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:25:38,833 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9477, 2.1380, 2.5916, 2.9819, 2.8873, 3.4346, 2.1369, 3.4184], device='cuda:2'), covar=tensor([0.0312, 0.0582, 0.0388, 0.0357, 0.0357, 0.0206, 0.0645, 0.0162], device='cuda:2'), in_proj_covar=tensor([0.0201, 0.0199, 0.0189, 0.0194, 0.0210, 0.0168, 0.0204, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:25:57,112 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:26:00,922 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 13:26:22,393 INFO [train.py:904] (2/8) Epoch 28, batch 4400, loss[loss=0.1881, simple_loss=0.2736, pruned_loss=0.05129, over 15511.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2749, pruned_loss=0.04803, over 3185524.01 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:26:58,177 INFO [optim.py:368] (2/8) 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 13:27:26,404 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9149, 2.7328, 2.8440, 2.0905, 2.6942, 2.1181, 2.7634, 2.8958], device='cuda:2'), covar=tensor([0.0240, 0.0775, 0.0495, 0.1864, 0.0749, 0.0904, 0.0539, 0.0791], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0170, 0.0157, 0.0149, 0.0132, 0.0146, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 13:27:35,966 INFO [train.py:904] (2/8) Epoch 28, batch 4450, loss[loss=0.1985, simple_loss=0.2961, pruned_loss=0.05048, over 15389.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2779, pruned_loss=0.049, over 3199128.53 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:27:47,319 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:28:25,767 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8390, 2.0799, 2.1221, 3.2692, 2.0801, 2.3145, 2.2458, 2.1812], device='cuda:2'), covar=tensor([0.1617, 0.3448, 0.3019, 0.0810, 0.4432, 0.2512, 0.3156, 0.3615], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0476, 0.0385, 0.0338, 0.0445, 0.0546, 0.0447, 0.0556], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:28:35,195 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-05-02 13:28:48,400 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5602, 2.6607, 2.1848, 2.3928, 3.0542, 2.5705, 3.0958, 3.2340], device='cuda:2'), covar=tensor([0.0110, 0.0420, 0.0568, 0.0476, 0.0248, 0.0393, 0.0222, 0.0236], device='cuda:2'), in_proj_covar=tensor([0.0234, 0.0247, 0.0235, 0.0235, 0.0247, 0.0245, 0.0245, 0.0246], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:28:49,073 INFO [train.py:904] (2/8) Epoch 28, batch 4500, loss[loss=0.2009, simple_loss=0.2894, pruned_loss=0.05619, over 16556.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2789, pruned_loss=0.04993, over 3194363.10 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:29:17,571 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 13:30:01,296 INFO [train.py:904] (2/8) Epoch 28, batch 4550, loss[loss=0.1943, simple_loss=0.2803, pruned_loss=0.05411, over 15418.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2796, pruned_loss=0.0505, over 3214595.58 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:30:37,057 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278627.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:12,169 INFO [train.py:904] (2/8) Epoch 28, batch 4600, loss[loss=0.1794, simple_loss=0.2702, pruned_loss=0.04423, over 17219.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2805, pruned_loss=0.05081, over 3212185.31 frames. ], batch size: 45, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:31:31,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6292, 4.8716, 4.7123, 4.7256, 4.4387, 4.3228, 4.4048, 4.9311], device='cuda:2'), covar=tensor([0.1186, 0.0871, 0.1011, 0.0859, 0.0724, 0.1492, 0.1045, 0.0875], device='cuda:2'), in_proj_covar=tensor([0.0723, 0.0881, 0.0718, 0.0677, 0.0555, 0.0551, 0.0735, 0.0686], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:31:38,342 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 13:31:43,899 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:46,040 INFO [optim.py:368] (2/8) 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 13:31:49,269 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:50,534 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2714, 5.2885, 5.5604, 5.5584, 5.6801, 5.2774, 5.2329, 4.8692], device='cuda:2'), covar=tensor([0.0290, 0.0431, 0.0395, 0.0404, 0.0433, 0.0353, 0.0921, 0.0481], device='cuda:2'), in_proj_covar=tensor([0.0436, 0.0492, 0.0473, 0.0438, 0.0522, 0.0498, 0.0575, 0.0403], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 13:32:22,022 INFO [train.py:904] (2/8) Epoch 28, batch 4650, loss[loss=0.1888, simple_loss=0.276, pruned_loss=0.05074, over 16598.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.28, pruned_loss=0.05126, over 3191144.11 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:32:34,031 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 13:33:00,841 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:33:14,917 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-05-02 13:33:33,716 INFO [train.py:904] (2/8) Epoch 28, batch 4700, loss[loss=0.1716, simple_loss=0.2614, pruned_loss=0.04086, over 16760.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2775, pruned_loss=0.05035, over 3200982.91 frames. ], batch size: 89, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:34:07,034 INFO [optim.py:368] (2/8) 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 13:34:45,677 INFO [train.py:904] (2/8) Epoch 28, batch 4750, loss[loss=0.1751, simple_loss=0.2608, pruned_loss=0.04472, over 16496.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.274, pruned_loss=0.04863, over 3202049.14 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:35:12,587 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-05-02 13:35:57,544 INFO [train.py:904] (2/8) Epoch 28, batch 4800, loss[loss=0.172, simple_loss=0.2678, pruned_loss=0.03812, over 17213.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2705, pruned_loss=0.04683, over 3185465.21 frames. ], batch size: 45, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:36:18,310 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:36:27,718 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4677, 4.7094, 4.5460, 4.5588, 4.2735, 4.2166, 4.2602, 4.7481], device='cuda:2'), covar=tensor([0.1129, 0.0796, 0.0867, 0.0766, 0.0736, 0.1502, 0.0939, 0.0865], device='cuda:2'), in_proj_covar=tensor([0.0719, 0.0879, 0.0715, 0.0675, 0.0553, 0.0549, 0.0733, 0.0683], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:36:32,582 INFO [optim.py:368] (2/8) 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 13:37:13,014 INFO [train.py:904] (2/8) Epoch 28, batch 4850, loss[loss=0.1698, simple_loss=0.2603, pruned_loss=0.03972, over 16461.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2711, pruned_loss=0.04609, over 3161589.81 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:38:20,010 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5260, 3.7125, 2.7843, 2.1949, 2.3990, 2.4402, 4.0201, 3.1503], device='cuda:2'), covar=tensor([0.3079, 0.0638, 0.1970, 0.2898, 0.2723, 0.2123, 0.0433, 0.1447], device='cuda:2'), in_proj_covar=tensor([0.0335, 0.0276, 0.0313, 0.0327, 0.0307, 0.0277, 0.0304, 0.0354], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 13:38:26,573 INFO [train.py:904] (2/8) Epoch 28, batch 4900, loss[loss=0.1607, simple_loss=0.2598, pruned_loss=0.03082, over 16807.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.27, pruned_loss=0.04455, over 3173051.07 frames. ], batch size: 102, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:00,726 INFO [optim.py:368] (2/8) 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 13:39:04,623 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:39:39,802 INFO [train.py:904] (2/8) Epoch 28, batch 4950, loss[loss=0.1854, simple_loss=0.2815, pruned_loss=0.04461, over 16769.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.27, pruned_loss=0.0439, over 3183829.03 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:50,914 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9091, 2.6341, 2.8300, 2.0544, 2.6731, 2.1381, 2.7848, 2.8220], device='cuda:2'), covar=tensor([0.0272, 0.0862, 0.0620, 0.1922, 0.0818, 0.0956, 0.0568, 0.0747], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0171, 0.0170, 0.0156, 0.0148, 0.0132, 0.0145, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 13:40:14,413 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:19,454 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 5000, loss[loss=0.1687, simple_loss=0.277, pruned_loss=0.03024, over 16885.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2713, pruned_loss=0.04381, over 3201474.93 frames. ], batch size: 102, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:41:26,571 INFO [optim.py:368] (2/8) 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 13:41:27,923 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279078.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:41:39,662 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5609, 3.6397, 3.3747, 3.0931, 3.1909, 3.5121, 3.3506, 3.3592], device='cuda:2'), covar=tensor([0.0532, 0.0536, 0.0303, 0.0292, 0.0578, 0.0440, 0.1277, 0.0466], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0468, 0.0365, 0.0367, 0.0363, 0.0420, 0.0248, 0.0435], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:41:55,810 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2234, 5.7218, 5.9128, 5.4962, 5.6971, 6.1875, 5.6985, 5.3764], device='cuda:2'), covar=tensor([0.0749, 0.1514, 0.1763, 0.2135, 0.2136, 0.0803, 0.1385, 0.2103], device='cuda:2'), in_proj_covar=tensor([0.0425, 0.0632, 0.0693, 0.0516, 0.0685, 0.0721, 0.0541, 0.0687], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 13:42:04,193 INFO [train.py:904] (2/8) Epoch 28, batch 5050, loss[loss=0.1865, simple_loss=0.2791, pruned_loss=0.04693, over 16482.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2717, pruned_loss=0.0436, over 3202233.19 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:42:45,664 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:17,191 INFO [train.py:904] (2/8) Epoch 28, batch 5100, loss[loss=0.1721, simple_loss=0.2576, pruned_loss=0.04329, over 15461.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2694, pruned_loss=0.043, over 3206888.65 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:43:37,511 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:52,643 INFO [optim.py:368] (2/8) 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 13:43:54,495 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0161, 2.7070, 2.8109, 2.1089, 2.6263, 2.1656, 2.7796, 2.9271], device='cuda:2'), covar=tensor([0.0319, 0.0833, 0.0643, 0.1850, 0.0859, 0.0976, 0.0613, 0.0779], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0148, 0.0132, 0.0146, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 13:44:15,314 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:44:19,165 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 13:44:30,682 INFO [train.py:904] (2/8) Epoch 28, batch 5150, loss[loss=0.1753, simple_loss=0.2625, pruned_loss=0.04407, over 17019.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2694, pruned_loss=0.04243, over 3180443.80 frames. ], batch size: 55, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:44:48,994 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:45:42,233 INFO [train.py:904] (2/8) Epoch 28, batch 5200, loss[loss=0.1927, simple_loss=0.2838, pruned_loss=0.05079, over 12338.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2683, pruned_loss=0.04198, over 3169035.29 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:46:17,347 INFO [optim.py:368] (2/8) 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 13:46:34,596 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 13:46:53,552 INFO [train.py:904] (2/8) Epoch 28, batch 5250, loss[loss=0.165, simple_loss=0.2518, pruned_loss=0.03908, over 16402.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2659, pruned_loss=0.04178, over 3168773.58 frames. ], batch size: 75, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:47:29,342 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5687, 4.7226, 4.8675, 4.6255, 4.7168, 5.2053, 4.6969, 4.3827], device='cuda:2'), covar=tensor([0.1241, 0.1775, 0.1917, 0.2035, 0.2493, 0.0906, 0.1535, 0.2636], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0630, 0.0690, 0.0513, 0.0683, 0.0718, 0.0540, 0.0685], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 13:47:34,308 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0451, 5.0635, 5.3457, 5.3329, 5.3656, 5.0622, 4.9817, 4.7918], device='cuda:2'), covar=tensor([0.0257, 0.0517, 0.0350, 0.0343, 0.0427, 0.0327, 0.0859, 0.0441], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0488, 0.0473, 0.0435, 0.0517, 0.0497, 0.0573, 0.0400], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 13:48:06,861 INFO [train.py:904] (2/8) Epoch 28, batch 5300, loss[loss=0.1688, simple_loss=0.2602, pruned_loss=0.03867, over 16490.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2624, pruned_loss=0.04077, over 3177507.43 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:48:41,218 INFO [optim.py:368] (2/8) 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 13:48:49,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3166, 1.7189, 2.0737, 2.2888, 2.4176, 2.5954, 1.8859, 2.5143], device='cuda:2'), covar=tensor([0.0255, 0.0595, 0.0310, 0.0373, 0.0334, 0.0218, 0.0582, 0.0154], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0199, 0.0188, 0.0194, 0.0209, 0.0168, 0.0204, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:49:10,618 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-02 13:49:20,977 INFO [train.py:904] (2/8) Epoch 28, batch 5350, loss[loss=0.1656, simple_loss=0.2644, pruned_loss=0.03339, over 16812.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2612, pruned_loss=0.04042, over 3172044.77 frames. ], batch size: 102, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:50:32,511 INFO [train.py:904] (2/8) Epoch 28, batch 5400, loss[loss=0.1926, simple_loss=0.2812, pruned_loss=0.05197, over 12487.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2637, pruned_loss=0.04065, over 3174078.31 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:51:08,323 INFO [optim.py:368] (2/8) 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 13:51:24,273 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:51:50,352 INFO [train.py:904] (2/8) Epoch 28, batch 5450, loss[loss=0.2005, simple_loss=0.2828, pruned_loss=0.05913, over 11894.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2667, pruned_loss=0.042, over 3171662.42 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:52:31,379 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279529.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:52:47,267 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-05-02 13:52:52,421 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279542.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:53:09,374 INFO [train.py:904] (2/8) Epoch 28, batch 5500, loss[loss=0.1999, simple_loss=0.2931, pruned_loss=0.05336, over 16378.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2726, pruned_loss=0.04494, over 3168657.73 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:53:17,312 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0245, 4.4236, 3.3460, 2.8009, 3.2030, 2.8974, 4.9617, 3.9987], device='cuda:2'), covar=tensor([0.2603, 0.0501, 0.1581, 0.2369, 0.2619, 0.1827, 0.0321, 0.1108], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0275, 0.0312, 0.0326, 0.0305, 0.0276, 0.0304, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 13:53:47,381 INFO [optim.py:368] (2/8) 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 13:54:06,412 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 13:54:09,290 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:28,765 INFO [train.py:904] (2/8) Epoch 28, batch 5550, loss[loss=0.2102, simple_loss=0.3011, pruned_loss=0.05965, over 15221.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2791, pruned_loss=0.04944, over 3154469.22 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:54:29,372 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:59,585 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4919, 3.4664, 3.4615, 2.6454, 3.3101, 2.0460, 3.1235, 2.7716], device='cuda:2'), covar=tensor([0.0205, 0.0177, 0.0218, 0.0321, 0.0140, 0.2787, 0.0167, 0.0349], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0176, 0.0213, 0.0187, 0.0190, 0.0218, 0.0202, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 13:55:33,499 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3869, 2.9410, 2.7380, 2.3917, 2.3799, 2.3950, 2.9623, 2.9195], device='cuda:2'), covar=tensor([0.2124, 0.0582, 0.1469, 0.2263, 0.2173, 0.2056, 0.0447, 0.1137], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0274, 0.0312, 0.0327, 0.0305, 0.0276, 0.0304, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 13:55:48,649 INFO [train.py:904] (2/8) Epoch 28, batch 5600, loss[loss=0.2721, simple_loss=0.3346, pruned_loss=0.1048, over 11319.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2837, pruned_loss=0.05398, over 3098405.89 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:56:28,891 INFO [optim.py:368] (2/8) 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 13:56:53,542 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 13:57:11,930 INFO [train.py:904] (2/8) Epoch 28, batch 5650, loss[loss=0.2003, simple_loss=0.2862, pruned_loss=0.05716, over 16733.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2885, pruned_loss=0.05821, over 3063089.77 frames. ], batch size: 76, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:57:14,868 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 13:57:51,244 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 13:58:29,304 INFO [train.py:904] (2/8) Epoch 28, batch 5700, loss[loss=0.2664, simple_loss=0.3276, pruned_loss=0.1025, over 11215.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2904, pruned_loss=0.06024, over 3047066.41 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:59:05,447 INFO [optim.py:368] (2/8) 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 13:59:22,121 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:59:46,949 INFO [train.py:904] (2/8) Epoch 28, batch 5750, loss[loss=0.1982, simple_loss=0.2832, pruned_loss=0.05659, over 16270.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2935, pruned_loss=0.06152, over 3044314.83 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:00:13,790 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.48 vs. limit=5.0 2023-05-02 14:00:30,798 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3138, 3.0495, 3.3929, 1.8158, 3.5160, 3.5646, 2.8893, 2.7190], device='cuda:2'), covar=tensor([0.0883, 0.0330, 0.0231, 0.1246, 0.0114, 0.0251, 0.0449, 0.0523], device='cuda:2'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0137, 0.0086, 0.0129, 0.0128, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-02 14:00:39,834 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:01:07,307 INFO [train.py:904] (2/8) Epoch 28, batch 5800, loss[loss=0.2048, simple_loss=0.2962, pruned_loss=0.05668, over 16289.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2935, pruned_loss=0.06045, over 3053159.60 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:01:18,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3488, 2.5041, 2.4427, 4.1187, 2.3634, 2.8630, 2.5127, 2.6058], device='cuda:2'), covar=tensor([0.1343, 0.3162, 0.2851, 0.0528, 0.3765, 0.2164, 0.3403, 0.2966], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0473, 0.0382, 0.0335, 0.0443, 0.0540, 0.0443, 0.0550], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:01:46,121 INFO [optim.py:368] (2/8) 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 14:01:57,431 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:02,549 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0101, 2.1539, 2.2814, 3.5420, 2.1165, 2.4892, 2.2975, 2.3090], device='cuda:2'), covar=tensor([0.1582, 0.3689, 0.3138, 0.0700, 0.4300, 0.2606, 0.3573, 0.3508], device='cuda:2'), in_proj_covar=tensor([0.0420, 0.0473, 0.0383, 0.0335, 0.0443, 0.0541, 0.0444, 0.0551], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:02:15,486 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-05-02 14:02:16,884 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279898.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:25,208 INFO [train.py:904] (2/8) Epoch 28, batch 5850, loss[loss=0.2063, simple_loss=0.2886, pruned_loss=0.062, over 16880.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2913, pruned_loss=0.0591, over 3056835.82 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:02:50,424 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3728, 3.4901, 2.1261, 3.8247, 2.6138, 3.8332, 2.2995, 2.7780], device='cuda:2'), covar=tensor([0.0307, 0.0368, 0.1718, 0.0282, 0.0899, 0.0613, 0.1607, 0.0896], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0172, 0.0180, 0.0220, 0.0203, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 14:03:44,211 INFO [train.py:904] (2/8) Epoch 28, batch 5900, loss[loss=0.1939, simple_loss=0.2845, pruned_loss=0.05166, over 16842.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2911, pruned_loss=0.05851, over 3077477.32 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:04:08,369 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3740, 4.2122, 4.4026, 4.5599, 4.7306, 4.3038, 4.6994, 4.7559], device='cuda:2'), covar=tensor([0.2103, 0.1295, 0.1722, 0.0868, 0.0707, 0.1185, 0.0903, 0.0747], device='cuda:2'), in_proj_covar=tensor([0.0674, 0.0823, 0.0954, 0.0839, 0.0641, 0.0670, 0.0701, 0.0810], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:04:26,171 INFO [optim.py:368] (2/8) 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 14:04:43,286 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5974, 2.6047, 1.9075, 2.7158, 2.1197, 2.7707, 2.1323, 2.3650], device='cuda:2'), covar=tensor([0.0328, 0.0383, 0.1298, 0.0335, 0.0721, 0.0449, 0.1235, 0.0659], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0172, 0.0180, 0.0219, 0.0203, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 14:05:07,641 INFO [train.py:904] (2/8) Epoch 28, batch 5950, loss[loss=0.2046, simple_loss=0.2938, pruned_loss=0.05772, over 17053.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2915, pruned_loss=0.05747, over 3070601.46 frames. ], batch size: 55, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:20,818 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280051.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:06:23,832 INFO [train.py:904] (2/8) Epoch 28, batch 6000, loss[loss=0.1763, simple_loss=0.2637, pruned_loss=0.04443, over 16837.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2909, pruned_loss=0.05694, over 3086252.52 frames. ], batch size: 102, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:23,832 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 14:06:32,251 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1871, 4.5088, 5.0276, 3.5604, 4.4137, 3.6068, 4.8158, 4.7053], device='cuda:2'), covar=tensor([0.0157, 0.0719, 0.0342, 0.1482, 0.0513, 0.0738, 0.0386, 0.0724], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0173, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 14:06:34,271 INFO [train.py:938] (2/8) Epoch 28, validation: loss=0.148, simple_loss=0.2602, pruned_loss=0.0179, over 944034.00 frames. 2023-05-02 14:06:34,272 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 14:07:11,167 INFO [optim.py:368] (2/8) 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 14:07:18,458 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8284, 2.7095, 2.5958, 1.8640, 2.5737, 2.6652, 2.5437, 1.8993], device='cuda:2'), covar=tensor([0.0474, 0.0104, 0.0106, 0.0422, 0.0157, 0.0157, 0.0140, 0.0435], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0135, 0.0102, 0.0115, 0.0098, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 14:07:25,920 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280088.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:07:51,266 INFO [train.py:904] (2/8) Epoch 28, batch 6050, loss[loss=0.2182, simple_loss=0.285, pruned_loss=0.07574, over 11340.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2896, pruned_loss=0.05661, over 3075346.43 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:08:05,142 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7058, 4.6908, 4.5242, 3.6701, 4.5950, 1.6862, 4.3395, 4.1692], device='cuda:2'), covar=tensor([0.0164, 0.0145, 0.0228, 0.0409, 0.0147, 0.3149, 0.0213, 0.0307], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0186, 0.0189, 0.0218, 0.0201, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:08:06,508 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:08:53,999 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-02 14:09:02,495 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280149.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:09:08,866 INFO [train.py:904] (2/8) Epoch 28, batch 6100, loss[loss=0.189, simple_loss=0.2815, pruned_loss=0.04824, over 16315.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2882, pruned_loss=0.05533, over 3086901.99 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:09:51,397 INFO [optim.py:368] (2/8) 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 14:10:01,925 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280185.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:12,100 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0184, 1.9419, 2.5923, 2.9484, 2.7695, 3.3782, 2.2232, 3.3767], device='cuda:2'), covar=tensor([0.0261, 0.0631, 0.0359, 0.0372, 0.0367, 0.0204, 0.0628, 0.0151], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0193, 0.0210, 0.0167, 0.0204, 0.0166], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:10:22,446 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:25,455 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2062, 2.0590, 1.7428, 1.8194, 2.3501, 1.9933, 1.9984, 2.3932], device='cuda:2'), covar=tensor([0.0212, 0.0405, 0.0569, 0.0473, 0.0280, 0.0392, 0.0216, 0.0294], device='cuda:2'), in_proj_covar=tensor([0.0229, 0.0243, 0.0231, 0.0232, 0.0244, 0.0241, 0.0239, 0.0242], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:10:29,421 INFO [train.py:904] (2/8) Epoch 28, batch 6150, loss[loss=0.1817, simple_loss=0.2755, pruned_loss=0.0439, over 16691.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2863, pruned_loss=0.05458, over 3098102.95 frames. ], batch size: 89, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:10:41,270 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0219, 4.8515, 5.0434, 5.2388, 5.4055, 4.7741, 5.3786, 5.4075], device='cuda:2'), covar=tensor([0.1980, 0.1255, 0.1655, 0.0760, 0.0564, 0.0975, 0.0619, 0.0662], device='cuda:2'), in_proj_covar=tensor([0.0670, 0.0818, 0.0949, 0.0834, 0.0638, 0.0665, 0.0697, 0.0805], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:11:16,582 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280233.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:21,482 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-05-02 14:11:35,647 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280246.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:45,496 INFO [train.py:904] (2/8) Epoch 28, batch 6200, loss[loss=0.1816, simple_loss=0.2702, pruned_loss=0.04648, over 16748.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2843, pruned_loss=0.0546, over 3083672.04 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:12:24,023 INFO [optim.py:368] (2/8) 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 14:13:00,713 INFO [train.py:904] (2/8) Epoch 28, batch 6250, loss[loss=0.1555, simple_loss=0.25, pruned_loss=0.03055, over 17031.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2835, pruned_loss=0.05402, over 3093289.74 frames. ], batch size: 50, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:13:05,980 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6958, 4.9635, 4.7583, 4.7572, 4.5036, 4.4612, 4.3625, 5.0408], device='cuda:2'), covar=tensor([0.1264, 0.1003, 0.1067, 0.0887, 0.0820, 0.1202, 0.1292, 0.0963], device='cuda:2'), in_proj_covar=tensor([0.0719, 0.0876, 0.0716, 0.0674, 0.0551, 0.0548, 0.0728, 0.0678], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:13:36,624 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:14:07,425 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-02 14:14:14,430 INFO [train.py:904] (2/8) Epoch 28, batch 6300, loss[loss=0.198, simple_loss=0.2885, pruned_loss=0.05376, over 16768.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2834, pruned_loss=0.05342, over 3101157.27 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:14:53,563 INFO [optim.py:368] (2/8) 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 14:15:05,401 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-05-02 14:15:09,868 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:15:31,107 INFO [train.py:904] (2/8) Epoch 28, batch 6350, loss[loss=0.2066, simple_loss=0.2919, pruned_loss=0.06062, over 16389.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2847, pruned_loss=0.0548, over 3089326.18 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:15:36,782 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:02,721 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8826, 2.7459, 2.7970, 2.1795, 2.6439, 2.1683, 2.7239, 2.9609], device='cuda:2'), covar=tensor([0.0267, 0.0752, 0.0569, 0.1792, 0.0799, 0.0976, 0.0549, 0.0656], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 14:16:18,055 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 14:16:31,363 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:43,551 INFO [train.py:904] (2/8) Epoch 28, batch 6400, loss[loss=0.1819, simple_loss=0.2677, pruned_loss=0.04802, over 16843.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2844, pruned_loss=0.05562, over 3078892.90 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:16:47,913 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7615, 3.1400, 3.2174, 2.0317, 2.7980, 2.1143, 3.3820, 3.4321], device='cuda:2'), covar=tensor([0.0260, 0.0757, 0.0645, 0.2108, 0.0906, 0.1075, 0.0584, 0.0829], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 14:16:57,546 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 14:17:02,599 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 14:17:19,323 INFO [optim.py:368] (2/8) 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 14:17:56,109 INFO [train.py:904] (2/8) Epoch 28, batch 6450, loss[loss=0.1752, simple_loss=0.2645, pruned_loss=0.04291, over 16369.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2845, pruned_loss=0.05492, over 3097181.29 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:18:16,111 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0267, 4.0109, 3.9724, 2.9322, 3.9685, 1.7147, 3.7454, 3.3379], device='cuda:2'), covar=tensor([0.0168, 0.0147, 0.0217, 0.0424, 0.0121, 0.3464, 0.0172, 0.0461], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0186, 0.0190, 0.0218, 0.0201, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:18:18,362 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:18:29,816 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4721, 4.4511, 4.3773, 3.5083, 4.4009, 1.7623, 4.1601, 3.8282], device='cuda:2'), covar=tensor([0.0109, 0.0104, 0.0166, 0.0311, 0.0090, 0.2943, 0.0140, 0.0325], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0186, 0.0190, 0.0218, 0.0201, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:18:30,227 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 14:18:41,545 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:18:52,009 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 14:19:02,811 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:08,873 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 14:19:13,553 INFO [train.py:904] (2/8) Epoch 28, batch 6500, loss[loss=0.2033, simple_loss=0.2795, pruned_loss=0.06356, over 11368.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2828, pruned_loss=0.0544, over 3096571.28 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:19:49,977 INFO [optim.py:368] (2/8) 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 14:19:52,112 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:02,959 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 14:20:15,894 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:18,850 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3624, 3.0329, 3.4337, 1.6965, 3.5813, 3.6049, 2.9461, 2.6788], device='cuda:2'), covar=tensor([0.0846, 0.0328, 0.0216, 0.1301, 0.0102, 0.0212, 0.0431, 0.0517], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0140, 0.0087, 0.0131, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 14:20:28,521 INFO [train.py:904] (2/8) Epoch 28, batch 6550, loss[loss=0.2007, simple_loss=0.3017, pruned_loss=0.04991, over 16425.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2857, pruned_loss=0.05513, over 3097201.34 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:20:37,405 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:45,334 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:21:44,264 INFO [train.py:904] (2/8) Epoch 28, batch 6600, loss[loss=0.1943, simple_loss=0.2845, pruned_loss=0.05208, over 16495.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2875, pruned_loss=0.05532, over 3110701.62 frames. ], batch size: 75, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:22:11,771 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 14:22:16,973 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2999, 4.3949, 4.2270, 3.9606, 3.9257, 4.3419, 4.0335, 4.0793], device='cuda:2'), covar=tensor([0.0681, 0.0651, 0.0314, 0.0327, 0.0819, 0.0502, 0.0814, 0.0698], device='cuda:2'), in_proj_covar=tensor([0.0310, 0.0468, 0.0362, 0.0365, 0.0360, 0.0418, 0.0248, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:22:18,021 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:21,695 INFO [optim.py:368] (2/8) 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 14:22:28,860 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:59,891 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:01,186 INFO [train.py:904] (2/8) Epoch 28, batch 6650, loss[loss=0.1958, simple_loss=0.2844, pruned_loss=0.05359, over 16687.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.288, pruned_loss=0.0563, over 3105453.16 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:23:07,195 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:16,665 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9269, 4.6908, 4.8524, 5.1479, 5.2478, 4.7470, 5.3688, 5.3133], device='cuda:2'), covar=tensor([0.2221, 0.1566, 0.2083, 0.0907, 0.0853, 0.0939, 0.0760, 0.0912], device='cuda:2'), in_proj_covar=tensor([0.0668, 0.0815, 0.0944, 0.0830, 0.0636, 0.0663, 0.0697, 0.0801], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:24:01,099 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0273, 5.0293, 5.3982, 5.3661, 5.4207, 5.0578, 5.0032, 4.8261], device='cuda:2'), covar=tensor([0.0302, 0.0522, 0.0362, 0.0396, 0.0425, 0.0353, 0.0986, 0.0482], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0489, 0.0473, 0.0438, 0.0519, 0.0499, 0.0575, 0.0400], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 14:24:02,723 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280744.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:15,397 INFO [train.py:904] (2/8) Epoch 28, batch 6700, loss[loss=0.1676, simple_loss=0.2583, pruned_loss=0.03843, over 16678.00 frames. ], tot_loss[loss=0.2, simple_loss=0.287, pruned_loss=0.05653, over 3107543.07 frames. ], batch size: 89, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:24:18,762 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:19,021 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8527, 3.1840, 3.4122, 2.1444, 2.9438, 2.1549, 3.4005, 3.4917], device='cuda:2'), covar=tensor([0.0266, 0.0844, 0.0595, 0.2095, 0.0886, 0.1065, 0.0668, 0.0936], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0134, 0.0147, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 14:24:29,375 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5795, 4.5646, 4.4377, 3.6625, 4.5170, 1.6729, 4.2560, 4.0289], device='cuda:2'), covar=tensor([0.0094, 0.0096, 0.0189, 0.0368, 0.0089, 0.3036, 0.0125, 0.0309], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0175, 0.0213, 0.0186, 0.0189, 0.0218, 0.0201, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:24:30,555 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280763.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:47,763 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1582, 4.2012, 4.5149, 4.4655, 4.5026, 4.2098, 4.2074, 4.1762], device='cuda:2'), covar=tensor([0.0381, 0.0630, 0.0426, 0.0468, 0.0496, 0.0453, 0.0983, 0.0554], device='cuda:2'), in_proj_covar=tensor([0.0435, 0.0489, 0.0474, 0.0438, 0.0520, 0.0500, 0.0577, 0.0401], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 14:24:51,657 INFO [optim.py:368] (2/8) 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 14:25:12,496 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 6750, loss[loss=0.172, simple_loss=0.262, pruned_loss=0.04103, over 16728.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2856, pruned_loss=0.05663, over 3106132.40 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:26:06,846 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3709, 3.4597, 2.1243, 3.9259, 2.6311, 3.8912, 2.2032, 2.7813], device='cuda:2'), covar=tensor([0.0354, 0.0451, 0.1825, 0.0225, 0.1006, 0.0585, 0.1794, 0.0899], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0172, 0.0180, 0.0220, 0.0204, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 14:26:41,172 INFO [train.py:904] (2/8) Epoch 28, batch 6800, loss[loss=0.2304, simple_loss=0.3056, pruned_loss=0.07758, over 11434.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2855, pruned_loss=0.05668, over 3097134.17 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:13,224 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:18,675 INFO [optim.py:368] (2/8) 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 14:27:36,559 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 6850, loss[loss=0.2095, simple_loss=0.3107, pruned_loss=0.05417, over 16982.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2868, pruned_loss=0.05706, over 3100642.51 frames. ], batch size: 109, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:55,590 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:29:06,761 INFO [train.py:904] (2/8) Epoch 28, batch 6900, loss[loss=0.1783, simple_loss=0.2708, pruned_loss=0.04285, over 16759.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2895, pruned_loss=0.05649, over 3115989.06 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:29:34,342 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 14:29:53,906 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:30:22,542 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3946, 4.5244, 4.6535, 4.4435, 4.5543, 5.0533, 4.5932, 4.3209], device='cuda:2'), covar=tensor([0.1536, 0.1971, 0.2669, 0.1933, 0.2284, 0.0985, 0.1662, 0.2463], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0639, 0.0704, 0.0516, 0.0692, 0.0727, 0.0546, 0.0693], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 14:30:23,418 INFO [train.py:904] (2/8) Epoch 28, batch 6950, loss[loss=0.1964, simple_loss=0.2879, pruned_loss=0.05246, over 16260.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2904, pruned_loss=0.05703, over 3125241.14 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:31:03,952 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:31:26,211 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0804, 3.3529, 3.3603, 2.1161, 3.1510, 3.4124, 3.2130, 1.9725], device='cuda:2'), covar=tensor([0.0596, 0.0082, 0.0088, 0.0522, 0.0119, 0.0134, 0.0113, 0.0534], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0135, 0.0102, 0.0115, 0.0098, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 14:31:30,250 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281049.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:31:35,802 INFO [train.py:904] (2/8) Epoch 28, batch 7000, loss[loss=0.2092, simple_loss=0.2997, pruned_loss=0.05933, over 16370.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2905, pruned_loss=0.05683, over 3116681.10 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:31:44,073 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281058.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:32:12,984 INFO [optim.py:368] (2/8) 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 14:32:50,411 INFO [train.py:904] (2/8) Epoch 28, batch 7050, loss[loss=0.2379, simple_loss=0.3026, pruned_loss=0.08658, over 11435.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2916, pruned_loss=0.0573, over 3089196.31 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:33:01,150 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281110.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 14:34:04,177 INFO [train.py:904] (2/8) Epoch 28, batch 7100, loss[loss=0.2007, simple_loss=0.2903, pruned_loss=0.0555, over 15219.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2905, pruned_loss=0.05699, over 3103175.84 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:34:29,372 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281169.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:36,230 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2501, 3.9436, 3.8300, 2.3509, 3.5667, 3.9688, 3.5933, 2.0664], device='cuda:2'), covar=tensor([0.0703, 0.0073, 0.0088, 0.0577, 0.0120, 0.0139, 0.0114, 0.0578], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0136, 0.0102, 0.0116, 0.0098, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 14:34:37,322 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281174.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:43,391 INFO [optim.py:368] (2/8) 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 14:34:59,565 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:35:20,739 INFO [train.py:904] (2/8) Epoch 28, batch 7150, loss[loss=0.2188, simple_loss=0.3009, pruned_loss=0.06835, over 15445.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2891, pruned_loss=0.05714, over 3093049.39 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:35:21,015 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:00,668 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281230.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:36:09,136 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:29,667 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281251.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:31,500 INFO [train.py:904] (2/8) Epoch 28, batch 7200, loss[loss=0.2019, simple_loss=0.2843, pruned_loss=0.05974, over 11856.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2878, pruned_loss=0.05621, over 3060319.56 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:36:33,019 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 14:36:52,362 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4809, 4.2466, 4.1188, 2.8436, 3.6904, 4.1915, 3.6680, 2.5377], device='cuda:2'), covar=tensor([0.0564, 0.0052, 0.0060, 0.0404, 0.0117, 0.0119, 0.0114, 0.0433], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0101, 0.0115, 0.0097, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 14:36:56,616 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281270.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:37:07,756 INFO [optim.py:368] (2/8) 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 14:37:46,524 INFO [train.py:904] (2/8) Epoch 28, batch 7250, loss[loss=0.1894, simple_loss=0.2757, pruned_loss=0.05149, over 16836.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2848, pruned_loss=0.05448, over 3064835.05 frames. ], batch size: 102, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:38:09,206 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:38:59,568 INFO [train.py:904] (2/8) Epoch 28, batch 7300, loss[loss=0.1887, simple_loss=0.2824, pruned_loss=0.04751, over 15163.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.284, pruned_loss=0.05431, over 3071247.66 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:39:08,085 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281358.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:39:11,116 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3646, 4.0366, 3.8752, 2.6364, 3.5408, 4.0046, 3.5285, 2.3781], device='cuda:2'), covar=tensor([0.0594, 0.0064, 0.0072, 0.0467, 0.0127, 0.0126, 0.0126, 0.0487], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0102, 0.0115, 0.0098, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 14:39:39,518 INFO [optim.py:368] (2/8) 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 14:40:06,380 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0857, 2.3140, 2.5420, 1.8883, 2.6497, 2.6865, 2.4155, 2.3210], device='cuda:2'), covar=tensor([0.0771, 0.0326, 0.0299, 0.1019, 0.0146, 0.0316, 0.0478, 0.0479], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0110, 0.0103, 0.0138, 0.0086, 0.0130, 0.0129, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 14:40:13,645 INFO [train.py:904] (2/8) Epoch 28, batch 7350, loss[loss=0.2105, simple_loss=0.2973, pruned_loss=0.06188, over 15429.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2852, pruned_loss=0.05546, over 3058629.71 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:40:16,307 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281405.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 14:40:17,486 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:40:18,933 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4838, 1.7439, 2.1304, 2.3420, 2.4587, 2.7176, 1.8640, 2.6718], device='cuda:2'), covar=tensor([0.0261, 0.0594, 0.0385, 0.0450, 0.0343, 0.0274, 0.0634, 0.0187], device='cuda:2'), in_proj_covar=tensor([0.0195, 0.0197, 0.0185, 0.0190, 0.0206, 0.0164, 0.0201, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:40:49,416 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:41:28,059 INFO [train.py:904] (2/8) Epoch 28, batch 7400, loss[loss=0.2009, simple_loss=0.293, pruned_loss=0.05442, over 16950.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2862, pruned_loss=0.05555, over 3076727.84 frames. ], batch size: 41, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:41:33,666 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:41:55,404 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5976, 3.0455, 3.1977, 1.9867, 2.8643, 2.0842, 3.2040, 3.3866], device='cuda:2'), covar=tensor([0.0314, 0.0831, 0.0634, 0.2243, 0.0870, 0.1144, 0.0669, 0.0799], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0172, 0.0172, 0.0158, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 14:42:08,140 INFO [optim.py:368] (2/8) 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 14:42:18,965 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 7450, loss[loss=0.217, simple_loss=0.3056, pruned_loss=0.0642, over 15287.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2877, pruned_loss=0.05641, over 3086292.52 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:42:49,628 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1965, 4.2420, 4.5439, 4.5093, 4.5246, 4.2760, 4.2747, 4.2215], device='cuda:2'), covar=tensor([0.0349, 0.0613, 0.0421, 0.0449, 0.0488, 0.0440, 0.0855, 0.0492], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0485, 0.0470, 0.0434, 0.0516, 0.0495, 0.0570, 0.0398], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 14:43:05,942 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:43:17,712 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281525.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 14:43:51,031 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6438, 4.6519, 4.4934, 3.7104, 4.5666, 1.6842, 4.3179, 4.0764], device='cuda:2'), covar=tensor([0.0095, 0.0107, 0.0210, 0.0351, 0.0096, 0.3077, 0.0132, 0.0312], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0174, 0.0213, 0.0185, 0.0188, 0.0218, 0.0201, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:43:58,005 INFO [train.py:904] (2/8) Epoch 28, batch 7500, loss[loss=0.2141, simple_loss=0.2938, pruned_loss=0.06725, over 15274.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2878, pruned_loss=0.05602, over 3070247.72 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:44:23,861 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9021, 4.9354, 4.7853, 4.3663, 4.4416, 4.8451, 4.7391, 4.5376], device='cuda:2'), covar=tensor([0.0622, 0.0519, 0.0333, 0.0376, 0.1023, 0.0497, 0.0369, 0.0723], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0464, 0.0359, 0.0362, 0.0357, 0.0414, 0.0247, 0.0429], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:44:36,766 INFO [optim.py:368] (2/8) 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 14:45:11,672 INFO [train.py:904] (2/8) Epoch 28, batch 7550, loss[loss=0.206, simple_loss=0.3008, pruned_loss=0.05554, over 15420.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2874, pruned_loss=0.05661, over 3052788.77 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:46:25,965 INFO [train.py:904] (2/8) Epoch 28, batch 7600, loss[loss=0.206, simple_loss=0.3023, pruned_loss=0.05488, over 16940.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2861, pruned_loss=0.05623, over 3062956.44 frames. ], batch size: 109, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:46:40,964 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3878, 3.3751, 3.4297, 3.4902, 3.5109, 3.2800, 3.4991, 3.5706], device='cuda:2'), covar=tensor([0.1244, 0.0929, 0.1007, 0.0637, 0.0698, 0.2426, 0.1145, 0.0879], device='cuda:2'), in_proj_covar=tensor([0.0660, 0.0808, 0.0935, 0.0821, 0.0627, 0.0655, 0.0689, 0.0793], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:47:04,822 INFO [optim.py:368] (2/8) 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 14:47:40,203 INFO [train.py:904] (2/8) Epoch 28, batch 7650, loss[loss=0.2434, simple_loss=0.3175, pruned_loss=0.08469, over 11566.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2864, pruned_loss=0.05667, over 3065929.80 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:47:43,578 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281705.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:47:51,656 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3920, 4.5072, 4.3124, 4.0142, 4.0401, 4.4128, 4.1463, 4.1401], device='cuda:2'), covar=tensor([0.0678, 0.0600, 0.0297, 0.0316, 0.0825, 0.0532, 0.0614, 0.0635], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0463, 0.0357, 0.0360, 0.0356, 0.0413, 0.0247, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:47:56,480 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5113, 3.5994, 3.3378, 3.0222, 3.2278, 3.4905, 3.3143, 3.2986], device='cuda:2'), covar=tensor([0.0608, 0.0678, 0.0298, 0.0299, 0.0508, 0.0517, 0.1469, 0.0506], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0463, 0.0357, 0.0360, 0.0356, 0.0413, 0.0247, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:48:51,546 INFO [train.py:904] (2/8) Epoch 28, batch 7700, loss[loss=0.2128, simple_loss=0.3045, pruned_loss=0.06053, over 16810.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2862, pruned_loss=0.05679, over 3086291.65 frames. ], batch size: 39, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:48:52,554 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281753.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:49:09,510 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6005, 3.8777, 2.8379, 2.3111, 2.5629, 2.4641, 4.1031, 3.3862], device='cuda:2'), covar=tensor([0.3188, 0.0652, 0.1976, 0.2921, 0.2749, 0.2270, 0.0483, 0.1392], device='cuda:2'), in_proj_covar=tensor([0.0337, 0.0273, 0.0313, 0.0328, 0.0305, 0.0278, 0.0305, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 14:49:31,077 INFO [optim.py:368] (2/8) 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 14:49:35,122 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:07,527 INFO [train.py:904] (2/8) Epoch 28, batch 7750, loss[loss=0.2047, simple_loss=0.2917, pruned_loss=0.0588, over 16630.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2866, pruned_loss=0.05667, over 3088400.33 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:50:21,322 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:40,376 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281825.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:20,212 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9211, 2.1427, 2.4542, 3.0911, 2.2531, 2.3822, 2.3784, 2.3039], device='cuda:2'), covar=tensor([0.1563, 0.3424, 0.2622, 0.0792, 0.4206, 0.2415, 0.3199, 0.3260], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0473, 0.0385, 0.0336, 0.0445, 0.0541, 0.0447, 0.0553], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:51:21,178 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281852.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:21,508 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-05-02 14:51:21,935 INFO [train.py:904] (2/8) Epoch 28, batch 7800, loss[loss=0.2413, simple_loss=0.3083, pruned_loss=0.08715, over 11434.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2874, pruned_loss=0.05731, over 3071753.89 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:51:52,697 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:52:03,257 INFO [optim.py:368] (2/8) 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 14:52:23,158 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7502, 6.1058, 5.8304, 5.9148, 5.5402, 5.4887, 5.4472, 6.2056], device='cuda:2'), covar=tensor([0.1260, 0.0844, 0.1000, 0.0922, 0.0820, 0.0613, 0.1291, 0.0864], device='cuda:2'), in_proj_covar=tensor([0.0716, 0.0869, 0.0712, 0.0671, 0.0547, 0.0551, 0.0724, 0.0674], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:52:37,335 INFO [train.py:904] (2/8) Epoch 28, batch 7850, loss[loss=0.2217, simple_loss=0.3098, pruned_loss=0.06675, over 15446.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2877, pruned_loss=0.05696, over 3066417.75 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:52:47,422 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281910.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:52:51,792 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281913.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:07,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6359, 2.5895, 2.4484, 4.1232, 2.8755, 3.8513, 1.5805, 2.8315], device='cuda:2'), covar=tensor([0.1480, 0.0906, 0.1413, 0.0183, 0.0215, 0.0440, 0.1837, 0.0927], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0201, 0.0203, 0.0208, 0.0219, 0.0210, 0.0199], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 14:53:37,323 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:48,938 INFO [train.py:904] (2/8) Epoch 28, batch 7900, loss[loss=0.1866, simple_loss=0.2782, pruned_loss=0.04747, over 17115.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2864, pruned_loss=0.0559, over 3082556.89 frames. ], batch size: 47, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:54:16,298 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281971.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:54:29,910 INFO [optim.py:368] (2/8) 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 14:55:09,795 INFO [train.py:904] (2/8) Epoch 28, batch 7950, loss[loss=0.2049, simple_loss=0.2866, pruned_loss=0.06165, over 16134.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2869, pruned_loss=0.05615, over 3083093.75 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:55:14,728 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 8000, loss[loss=0.2151, simple_loss=0.296, pruned_loss=0.06707, over 17056.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2874, pruned_loss=0.05662, over 3086455.25 frames. ], batch size: 53, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:56:34,736 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7105, 4.9261, 5.1022, 4.8369, 4.9565, 5.4861, 4.8659, 4.6376], device='cuda:2'), covar=tensor([0.1136, 0.1759, 0.2439, 0.1949, 0.2189, 0.0928, 0.1786, 0.2516], device='cuda:2'), in_proj_covar=tensor([0.0427, 0.0638, 0.0702, 0.0516, 0.0692, 0.0726, 0.0548, 0.0693], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 14:57:07,769 INFO [optim.py:368] (2/8) 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 14:57:11,253 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:33,321 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282096.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:42,452 INFO [train.py:904] (2/8) Epoch 28, batch 8050, loss[loss=0.2073, simple_loss=0.2994, pruned_loss=0.0576, over 16630.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2873, pruned_loss=0.05633, over 3092931.69 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:54,973 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6924, 1.8118, 1.6381, 1.4362, 1.9403, 1.5819, 1.5702, 1.9085], device='cuda:2'), covar=tensor([0.0218, 0.0317, 0.0449, 0.0388, 0.0243, 0.0303, 0.0192, 0.0235], device='cuda:2'), in_proj_covar=tensor([0.0226, 0.0240, 0.0231, 0.0231, 0.0242, 0.0240, 0.0237, 0.0240], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:57:56,177 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:22,175 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:36,497 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0624, 5.3460, 5.1173, 5.0972, 4.8913, 4.8789, 4.7474, 5.4553], device='cuda:2'), covar=tensor([0.1191, 0.0840, 0.0921, 0.0929, 0.0778, 0.0869, 0.1306, 0.0765], device='cuda:2'), in_proj_covar=tensor([0.0713, 0.0862, 0.0707, 0.0666, 0.0543, 0.0548, 0.0720, 0.0670], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 14:58:57,913 INFO [train.py:904] (2/8) Epoch 28, batch 8100, loss[loss=0.1962, simple_loss=0.2835, pruned_loss=0.05441, over 16401.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2866, pruned_loss=0.05588, over 3092049.01 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:59:03,731 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282157.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:08,426 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:38,412 INFO [optim.py:368] (2/8) 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 15:00:13,692 INFO [train.py:904] (2/8) Epoch 28, batch 8150, loss[loss=0.1772, simple_loss=0.2704, pruned_loss=0.042, over 16700.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2838, pruned_loss=0.05473, over 3120192.02 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:00:21,752 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282208.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:00:44,423 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9499, 2.1143, 2.3062, 3.3481, 2.1013, 2.3679, 2.2937, 2.2923], device='cuda:2'), covar=tensor([0.1652, 0.3770, 0.3042, 0.0751, 0.4474, 0.2752, 0.3575, 0.3445], device='cuda:2'), in_proj_covar=tensor([0.0420, 0.0474, 0.0384, 0.0335, 0.0446, 0.0542, 0.0447, 0.0554], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:01:02,591 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-05-02 15:01:29,268 INFO [train.py:904] (2/8) Epoch 28, batch 8200, loss[loss=0.2056, simple_loss=0.2837, pruned_loss=0.06377, over 11623.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2824, pruned_loss=0.05504, over 3098448.78 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:01:50,074 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282266.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:02:13,182 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:02:49,557 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7570, 4.5824, 4.7953, 4.9622, 5.1629, 4.6197, 5.2001, 5.1586], device='cuda:2'), covar=tensor([0.2170, 0.1467, 0.1903, 0.0889, 0.0658, 0.1076, 0.0615, 0.0733], device='cuda:2'), in_proj_covar=tensor([0.0662, 0.0812, 0.0938, 0.0824, 0.0629, 0.0656, 0.0691, 0.0798], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:02:50,325 INFO [train.py:904] (2/8) Epoch 28, batch 8250, loss[loss=0.172, simple_loss=0.2626, pruned_loss=0.04066, over 16584.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2813, pruned_loss=0.05258, over 3073098.25 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:03:04,665 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 15:04:07,365 INFO [train.py:904] (2/8) Epoch 28, batch 8300, loss[loss=0.1922, simple_loss=0.2882, pruned_loss=0.04806, over 15345.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2786, pruned_loss=0.04998, over 3051339.51 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:04:50,784 INFO [optim.py:368] (2/8) 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 15:04:58,653 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 15:05:26,438 INFO [train.py:904] (2/8) Epoch 28, batch 8350, loss[loss=0.1819, simple_loss=0.2743, pruned_loss=0.04479, over 15352.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2789, pruned_loss=0.0483, over 3077098.37 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:06:16,370 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 15:06:27,224 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-05-02 15:06:43,064 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282452.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:06:43,824 INFO [train.py:904] (2/8) Epoch 28, batch 8400, loss[loss=0.1702, simple_loss=0.266, pruned_loss=0.03723, over 16703.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.276, pruned_loss=0.04618, over 3072992.25 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:07:26,988 INFO [optim.py:368] (2/8) 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 15:07:39,972 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:08:02,089 INFO [train.py:904] (2/8) Epoch 28, batch 8450, loss[loss=0.1683, simple_loss=0.2647, pruned_loss=0.03597, over 16237.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2741, pruned_loss=0.04435, over 3093180.13 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:08:11,482 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:08:12,814 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0833, 3.1650, 3.1587, 2.2101, 2.8739, 3.1913, 3.1120, 2.0059], device='cuda:2'), covar=tensor([0.0531, 0.0078, 0.0082, 0.0413, 0.0151, 0.0119, 0.0102, 0.0517], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0133, 0.0100, 0.0114, 0.0097, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 15:08:33,371 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 15:09:17,252 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 8500, loss[loss=0.1733, simple_loss=0.2607, pruned_loss=0.04296, over 16650.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2704, pruned_loss=0.04225, over 3078568.20 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:09:27,450 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:43,312 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282566.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:09:45,717 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 15:10:07,072 INFO [optim.py:368] (2/8) 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 15:10:27,989 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 15:10:42,384 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 8550, loss[loss=0.1846, simple_loss=0.2808, pruned_loss=0.04414, over 16082.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2678, pruned_loss=0.04141, over 3041120.70 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:11:05,481 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282614.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:12:14,469 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 8600, loss[loss=0.1746, simple_loss=0.2661, pruned_loss=0.04152, over 16572.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2671, pruned_loss=0.04, over 3044001.29 frames. ], batch size: 62, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:12:32,784 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1250, 4.4114, 4.0774, 3.8226, 3.4901, 4.3061, 4.0000, 3.9520], device='cuda:2'), covar=tensor([0.1146, 0.1073, 0.0573, 0.0556, 0.1744, 0.0906, 0.1104, 0.0877], device='cuda:2'), in_proj_covar=tensor([0.0305, 0.0463, 0.0357, 0.0358, 0.0353, 0.0413, 0.0247, 0.0426], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:12:32,807 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282658.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:06,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8297, 4.3993, 3.1457, 2.2507, 2.6445, 2.7299, 4.6928, 3.6066], device='cuda:2'), covar=tensor([0.3015, 0.0457, 0.1837, 0.3350, 0.2991, 0.2076, 0.0339, 0.1303], device='cuda:2'), in_proj_covar=tensor([0.0332, 0.0269, 0.0309, 0.0324, 0.0300, 0.0275, 0.0301, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 15:13:11,877 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282677.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:17,326 INFO [optim.py:368] (2/8) 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 15:13:45,616 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-05-02 15:13:50,838 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282698.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:13:58,782 INFO [train.py:904] (2/8) Epoch 28, batch 8650, loss[loss=0.1597, simple_loss=0.2648, pruned_loss=0.02725, over 16195.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2655, pruned_loss=0.03843, over 3050521.64 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:14:36,253 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282719.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:14,409 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:40,213 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:40,925 INFO [train.py:904] (2/8) Epoch 28, batch 8700, loss[loss=0.163, simple_loss=0.2557, pruned_loss=0.03513, over 16909.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.263, pruned_loss=0.03735, over 3053559.80 frames. ], batch size: 116, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:15:53,239 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282759.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:16:20,489 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5830, 4.8643, 4.7030, 4.6747, 4.4238, 4.4219, 4.3658, 4.9333], device='cuda:2'), covar=tensor([0.1170, 0.0961, 0.0940, 0.0858, 0.0882, 0.1323, 0.1277, 0.0962], device='cuda:2'), in_proj_covar=tensor([0.0701, 0.0848, 0.0695, 0.0656, 0.0535, 0.0538, 0.0707, 0.0659], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:16:29,664 INFO [optim.py:368] (2/8) 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 15:16:36,733 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4304, 3.6355, 3.7465, 1.8627, 3.9766, 4.1561, 3.1648, 2.9932], device='cuda:2'), covar=tensor([0.1028, 0.0219, 0.0253, 0.1382, 0.0099, 0.0153, 0.0436, 0.0558], device='cuda:2'), in_proj_covar=tensor([0.0144, 0.0108, 0.0099, 0.0135, 0.0083, 0.0127, 0.0126, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 15:17:06,890 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 8750, loss[loss=0.187, simple_loss=0.2854, pruned_loss=0.04431, over 16782.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2635, pruned_loss=0.03688, over 3055868.80 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:17:14,506 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 15:17:18,875 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282805.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:18:46,444 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:18:51,399 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 15:19:02,829 INFO [train.py:904] (2/8) Epoch 28, batch 8800, loss[loss=0.1687, simple_loss=0.2644, pruned_loss=0.03647, over 16120.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2617, pruned_loss=0.03593, over 3052381.59 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:19:29,117 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:20:00,725 INFO [optim.py:368] (2/8) 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 15:20:47,892 INFO [train.py:904] (2/8) Epoch 28, batch 8850, loss[loss=0.161, simple_loss=0.2537, pruned_loss=0.03416, over 12795.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2643, pruned_loss=0.03553, over 3044136.35 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:22:35,072 INFO [train.py:904] (2/8) Epoch 28, batch 8900, loss[loss=0.1642, simple_loss=0.2554, pruned_loss=0.03645, over 12576.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2645, pruned_loss=0.03512, over 3035932.08 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:23:38,028 INFO [optim.py:368] (2/8) 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 15:24:39,869 INFO [train.py:904] (2/8) Epoch 28, batch 8950, loss[loss=0.152, simple_loss=0.2509, pruned_loss=0.02655, over 16298.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2647, pruned_loss=0.03558, over 3047926.46 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:25:04,564 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283014.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:25:45,366 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:25:45,783 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.04 vs. limit=5.0 2023-05-02 15:26:27,356 INFO [train.py:904] (2/8) Epoch 28, batch 9000, loss[loss=0.1457, simple_loss=0.2417, pruned_loss=0.02488, over 16172.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2612, pruned_loss=0.03417, over 3069367.27 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:26:27,357 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 15:26:35,522 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7050, 5.6423, 5.6979, 5.8651, 5.9563, 5.4327, 5.9283, 5.9491], device='cuda:2'), covar=tensor([0.1573, 0.0974, 0.1225, 0.0507, 0.0466, 0.0502, 0.0438, 0.0561], device='cuda:2'), in_proj_covar=tensor([0.0641, 0.0786, 0.0906, 0.0802, 0.0610, 0.0637, 0.0670, 0.0775], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:26:38,040 INFO [train.py:938] (2/8) Epoch 28, validation: loss=0.1436, simple_loss=0.2472, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 15:26:38,041 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 15:26:41,579 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283054.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:27:36,918 INFO [optim.py:368] (2/8) 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 15:28:19,164 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4836, 1.9830, 1.4587, 1.6144, 2.2162, 1.9145, 2.0743, 2.4719], device='cuda:2'), covar=tensor([0.0251, 0.0518, 0.0789, 0.0588, 0.0360, 0.0450, 0.0231, 0.0327], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0237, 0.0228, 0.0228, 0.0238, 0.0236, 0.0232, 0.0235], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:28:21,091 INFO [train.py:904] (2/8) Epoch 28, batch 9050, loss[loss=0.1555, simple_loss=0.2533, pruned_loss=0.02884, over 16880.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2624, pruned_loss=0.03473, over 3082903.52 frames. ], batch size: 102, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:28:29,097 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 15:28:48,317 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:46,409 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:48,309 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:03,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4120, 3.5316, 3.6835, 3.6580, 3.6822, 3.5071, 3.5449, 3.5770], device='cuda:2'), covar=tensor([0.0411, 0.0730, 0.0478, 0.0516, 0.0485, 0.0574, 0.0792, 0.0529], device='cuda:2'), in_proj_covar=tensor([0.0428, 0.0480, 0.0467, 0.0429, 0.0512, 0.0490, 0.0563, 0.0396], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 15:30:04,160 INFO [train.py:904] (2/8) Epoch 28, batch 9100, loss[loss=0.1587, simple_loss=0.2526, pruned_loss=0.0324, over 12431.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2617, pruned_loss=0.035, over 3068322.97 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:30:20,072 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283161.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:30,942 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-02 15:30:48,690 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 15:30:57,973 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:31:08,545 INFO [optim.py:368] (2/8) 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 15:31:38,531 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283193.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:31:44,386 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4200, 3.5460, 2.1813, 3.9691, 2.5955, 3.8911, 2.1813, 2.8776], device='cuda:2'), covar=tensor([0.0340, 0.0388, 0.1717, 0.0204, 0.0929, 0.0459, 0.1737, 0.0785], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0175, 0.0192, 0.0166, 0.0175, 0.0213, 0.0200, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 15:32:01,011 INFO [train.py:904] (2/8) Epoch 28, batch 9150, loss[loss=0.177, simple_loss=0.2655, pruned_loss=0.04424, over 15421.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2622, pruned_loss=0.03476, over 3065229.34 frames. ], batch size: 192, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:32:06,350 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283205.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:33:44,271 INFO [train.py:904] (2/8) Epoch 28, batch 9200, loss[loss=0.1741, simple_loss=0.2693, pruned_loss=0.0394, over 16266.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2574, pruned_loss=0.03352, over 3073281.79 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:34:34,270 INFO [optim.py:368] (2/8) 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 15:34:39,351 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:35:20,274 INFO [train.py:904] (2/8) Epoch 28, batch 9250, loss[loss=0.1555, simple_loss=0.2539, pruned_loss=0.02853, over 16373.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2577, pruned_loss=0.03401, over 3075379.62 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:35:44,258 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283314.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:35:51,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6320, 4.4740, 4.7038, 4.8420, 4.9968, 4.4774, 5.0369, 5.0296], device='cuda:2'), covar=tensor([0.2191, 0.1307, 0.1654, 0.0802, 0.0584, 0.0925, 0.0510, 0.0708], device='cuda:2'), in_proj_covar=tensor([0.0646, 0.0789, 0.0910, 0.0806, 0.0613, 0.0640, 0.0674, 0.0778], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:36:28,195 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:55,351 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:37:14,589 INFO [train.py:904] (2/8) Epoch 28, batch 9300, loss[loss=0.1457, simple_loss=0.2375, pruned_loss=0.02692, over 16757.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2561, pruned_loss=0.03368, over 3068330.13 frames. ], batch size: 76, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:37:16,901 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283354.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:37:34,850 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:37:35,084 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5319, 3.7036, 2.7598, 2.2197, 2.2055, 2.4040, 3.8590, 3.1566], device='cuda:2'), covar=tensor([0.3102, 0.0588, 0.1952, 0.3074, 0.2946, 0.2314, 0.0415, 0.1389], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0268, 0.0307, 0.0321, 0.0297, 0.0272, 0.0298, 0.0344], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 15:37:35,108 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0836, 2.1739, 2.1616, 3.6800, 2.1129, 2.5068, 2.2913, 2.3064], device='cuda:2'), covar=tensor([0.1471, 0.3913, 0.3486, 0.0627, 0.4439, 0.2750, 0.4009, 0.3672], device='cuda:2'), in_proj_covar=tensor([0.0411, 0.0464, 0.0380, 0.0326, 0.0438, 0.0530, 0.0438, 0.0542], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:38:16,507 INFO [optim.py:368] (2/8) 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 15:38:17,488 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283381.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:58,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283402.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:38:59,594 INFO [train.py:904] (2/8) Epoch 28, batch 9350, loss[loss=0.16, simple_loss=0.2562, pruned_loss=0.03187, over 16633.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2561, pruned_loss=0.03345, over 3080273.42 frames. ], batch size: 89, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:40:41,025 INFO [train.py:904] (2/8) Epoch 28, batch 9400, loss[loss=0.1705, simple_loss=0.2796, pruned_loss=0.0307, over 16351.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2564, pruned_loss=0.03347, over 3069955.37 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:41:00,042 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283461.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:18,987 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:39,543 INFO [optim.py:368] (2/8) 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 15:42:20,565 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283500.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:42:24,700 INFO [train.py:904] (2/8) Epoch 28, batch 9450, loss[loss=0.1635, simple_loss=0.2541, pruned_loss=0.03648, over 16592.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2578, pruned_loss=0.03345, over 3065460.74 frames. ], batch size: 62, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:42:36,846 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283509.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:43:04,129 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1029, 5.0515, 4.8413, 4.2776, 4.9477, 1.9243, 4.7018, 4.6202], device='cuda:2'), covar=tensor([0.0089, 0.0116, 0.0204, 0.0346, 0.0088, 0.2680, 0.0123, 0.0240], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0169, 0.0206, 0.0178, 0.0183, 0.0212, 0.0195, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:43:41,157 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8686, 3.7676, 3.9119, 4.0173, 4.0927, 3.7213, 4.0674, 4.1421], device='cuda:2'), covar=tensor([0.1587, 0.1124, 0.1263, 0.0692, 0.0573, 0.1741, 0.0755, 0.0767], device='cuda:2'), in_proj_covar=tensor([0.0641, 0.0783, 0.0904, 0.0799, 0.0608, 0.0633, 0.0668, 0.0772], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:43:48,953 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283544.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:44:06,790 INFO [train.py:904] (2/8) Epoch 28, batch 9500, loss[loss=0.1703, simple_loss=0.2618, pruned_loss=0.0394, over 16535.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2568, pruned_loss=0.03301, over 3077026.16 frames. ], batch size: 62, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:44:32,405 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2873, 3.6420, 3.9000, 2.2555, 3.2878, 2.5742, 3.6354, 3.7540], device='cuda:2'), covar=tensor([0.0294, 0.0859, 0.0483, 0.2120, 0.0737, 0.0945, 0.0659, 0.0960], device='cuda:2'), in_proj_covar=tensor([0.0156, 0.0164, 0.0165, 0.0154, 0.0143, 0.0129, 0.0141, 0.0176], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 15:45:03,521 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 15:45:03,873 INFO [optim.py:368] (2/8) 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 15:45:10,594 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2637, 4.3630, 4.4734, 4.2218, 4.3733, 4.8446, 4.4213, 4.0834], device='cuda:2'), covar=tensor([0.1574, 0.2177, 0.2360, 0.2269, 0.2537, 0.1024, 0.1665, 0.2646], device='cuda:2'), in_proj_covar=tensor([0.0409, 0.0612, 0.0675, 0.0497, 0.0665, 0.0702, 0.0526, 0.0665], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 15:45:39,809 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 15:45:53,231 INFO [train.py:904] (2/8) Epoch 28, batch 9550, loss[loss=0.168, simple_loss=0.2732, pruned_loss=0.03145, over 16083.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2562, pruned_loss=0.03291, over 3075112.33 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:59,233 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283605.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:47:09,794 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283639.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:47:34,671 INFO [train.py:904] (2/8) Epoch 28, batch 9600, loss[loss=0.1637, simple_loss=0.2502, pruned_loss=0.03857, over 12607.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2567, pruned_loss=0.03351, over 3048875.12 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:48:29,440 INFO [optim.py:368] (2/8) 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 15:48:58,322 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4358, 3.3509, 2.7436, 2.1624, 2.1884, 2.3213, 3.5476, 3.0237], device='cuda:2'), covar=tensor([0.3057, 0.0630, 0.1824, 0.3063, 0.2780, 0.2371, 0.0415, 0.1513], device='cuda:2'), in_proj_covar=tensor([0.0328, 0.0266, 0.0305, 0.0319, 0.0295, 0.0271, 0.0296, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 15:49:23,006 INFO [train.py:904] (2/8) Epoch 28, batch 9650, loss[loss=0.1549, simple_loss=0.2465, pruned_loss=0.03171, over 12475.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2594, pruned_loss=0.03396, over 3065775.10 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:51:10,205 INFO [train.py:904] (2/8) Epoch 28, batch 9700, loss[loss=0.1703, simple_loss=0.2639, pruned_loss=0.03836, over 16342.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2589, pruned_loss=0.03403, over 3057352.33 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:51:46,235 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 15:51:47,244 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283771.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:08,986 INFO [optim.py:368] (2/8) 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 15:52:48,122 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283800.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:52,808 INFO [train.py:904] (2/8) Epoch 28, batch 9750, loss[loss=0.1717, simple_loss=0.2657, pruned_loss=0.03882, over 16682.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2578, pruned_loss=0.0342, over 3058260.19 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:53:24,763 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283819.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:12,775 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283842.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:24,070 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283848.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:31,904 INFO [train.py:904] (2/8) Epoch 28, batch 9800, loss[loss=0.1522, simple_loss=0.2436, pruned_loss=0.0304, over 12319.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2573, pruned_loss=0.03325, over 3056301.42 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:54:41,158 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6367, 4.6189, 4.9821, 4.9671, 4.9716, 4.6820, 4.6202, 4.5654], device='cuda:2'), covar=tensor([0.0335, 0.0653, 0.0467, 0.0418, 0.0498, 0.0456, 0.1099, 0.0440], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0475, 0.0462, 0.0423, 0.0507, 0.0483, 0.0556, 0.0391], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 15:55:23,051 INFO [optim.py:368] (2/8) 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 15:55:37,847 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9627, 3.8732, 3.9770, 4.1389, 4.2223, 3.9046, 4.2144, 4.2755], device='cuda:2'), covar=tensor([0.1869, 0.1344, 0.1587, 0.0852, 0.0691, 0.1448, 0.0814, 0.0835], device='cuda:2'), in_proj_covar=tensor([0.0640, 0.0784, 0.0903, 0.0799, 0.0607, 0.0633, 0.0667, 0.0773], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:56:11,010 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283900.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:56:15,639 INFO [train.py:904] (2/8) Epoch 28, batch 9850, loss[loss=0.1525, simple_loss=0.2418, pruned_loss=0.0316, over 12445.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2581, pruned_loss=0.03298, over 3046795.85 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:56:17,406 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:57:36,910 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:57:51,021 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9915, 2.0853, 2.4821, 2.9809, 2.7139, 3.3772, 2.3358, 3.3597], device='cuda:2'), covar=tensor([0.0244, 0.0642, 0.0415, 0.0321, 0.0425, 0.0202, 0.0559, 0.0177], device='cuda:2'), in_proj_covar=tensor([0.0192, 0.0194, 0.0181, 0.0185, 0.0203, 0.0160, 0.0198, 0.0160], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 15:58:06,900 INFO [train.py:904] (2/8) Epoch 28, batch 9900, loss[loss=0.1459, simple_loss=0.2525, pruned_loss=0.01961, over 17156.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2585, pruned_loss=0.03283, over 3057336.33 frames. ], batch size: 48, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:58:38,568 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:59:13,294 INFO [optim.py:368] (2/8) 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 15:59:30,051 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:00:07,759 INFO [train.py:904] (2/8) Epoch 28, batch 9950, loss[loss=0.1574, simple_loss=0.2553, pruned_loss=0.02977, over 16521.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2603, pruned_loss=0.03289, over 3059065.57 frames. ], batch size: 68, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:00:44,584 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0940, 3.1514, 1.9014, 3.2729, 2.3852, 3.2686, 2.0853, 2.6537], device='cuda:2'), covar=tensor([0.0273, 0.0373, 0.1621, 0.0334, 0.0749, 0.0662, 0.1540, 0.0658], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0176, 0.0193, 0.0166, 0.0176, 0.0213, 0.0201, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:01:08,220 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:02:08,463 INFO [train.py:904] (2/8) Epoch 28, batch 10000, loss[loss=0.1604, simple_loss=0.2543, pruned_loss=0.03329, over 17022.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2592, pruned_loss=0.03268, over 3074501.26 frames. ], batch size: 55, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:02:52,028 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9708, 5.0960, 5.4250, 5.3673, 5.4204, 5.1480, 5.0353, 4.9291], device='cuda:2'), covar=tensor([0.0428, 0.0710, 0.0467, 0.0564, 0.0504, 0.0434, 0.1053, 0.0375], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0469, 0.0456, 0.0419, 0.0502, 0.0479, 0.0550, 0.0386], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 16:03:03,885 INFO [optim.py:368] (2/8) 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 16:03:50,388 INFO [train.py:904] (2/8) Epoch 28, batch 10050, loss[loss=0.1532, simple_loss=0.2541, pruned_loss=0.02609, over 16669.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2595, pruned_loss=0.03285, over 3092229.65 frames. ], batch size: 83, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:05:23,764 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4594, 4.5281, 4.3431, 4.0263, 4.0863, 4.4707, 4.1137, 4.1752], device='cuda:2'), covar=tensor([0.0719, 0.0908, 0.0423, 0.0411, 0.0841, 0.1054, 0.0680, 0.0704], device='cuda:2'), in_proj_covar=tensor([0.0297, 0.0446, 0.0348, 0.0347, 0.0341, 0.0401, 0.0239, 0.0413], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:05:24,422 INFO [train.py:904] (2/8) Epoch 28, batch 10100, loss[loss=0.1587, simple_loss=0.2504, pruned_loss=0.0335, over 16679.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2593, pruned_loss=0.03266, over 3090805.02 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:06:20,722 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:06:42,003 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 28, batch 10150, loss[loss=0.1507, simple_loss=0.2397, pruned_loss=0.03083, over 12223.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2582, pruned_loss=0.03282, over 3064834.31 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:07:10,315 INFO [train.py:904] (2/8) Epoch 29, batch 0, loss[loss=0.2097, simple_loss=0.2868, pruned_loss=0.06627, over 16682.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2868, pruned_loss=0.06627, over 16682.00 frames. ], batch size: 134, lr: 2.34e-03, grad_scale: 8.0 2023-05-02 16:07:10,316 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 16:07:17,739 INFO [train.py:938] (2/8) Epoch 29, validation: loss=0.1427, simple_loss=0.246, pruned_loss=0.0197, over 944034.00 frames. 2023-05-02 16:07:17,740 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 16:08:18,231 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284248.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:08:26,888 INFO [train.py:904] (2/8) Epoch 29, batch 50, loss[loss=0.1624, simple_loss=0.2595, pruned_loss=0.03266, over 17259.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2671, pruned_loss=0.04745, over 748246.71 frames. ], batch size: 52, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:09:08,278 INFO [optim.py:368] (2/8) 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 16:09:37,377 INFO [train.py:904] (2/8) Epoch 29, batch 100, loss[loss=0.1492, simple_loss=0.2416, pruned_loss=0.02835, over 17215.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2602, pruned_loss=0.04337, over 1319921.67 frames. ], batch size: 44, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:09:44,055 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 16:09:57,902 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-02 16:10:02,072 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:10:19,741 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 16:10:46,097 INFO [train.py:904] (2/8) Epoch 29, batch 150, loss[loss=0.1964, simple_loss=0.2658, pruned_loss=0.06352, over 16758.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2591, pruned_loss=0.0421, over 1757034.86 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:10:50,138 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2005, 3.2965, 3.6376, 2.2466, 3.1572, 2.4123, 3.6511, 3.7059], device='cuda:2'), covar=tensor([0.0266, 0.1023, 0.0612, 0.2057, 0.0869, 0.1146, 0.0534, 0.0858], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0168, 0.0155, 0.0145, 0.0131, 0.0143, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:11:10,526 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-05-02 16:11:22,485 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7209, 4.5775, 4.6277, 4.3067, 4.3777, 4.6544, 4.5606, 4.4220], device='cuda:2'), covar=tensor([0.0661, 0.1118, 0.0378, 0.0389, 0.0898, 0.0694, 0.0404, 0.0767], device='cuda:2'), in_proj_covar=tensor([0.0304, 0.0456, 0.0355, 0.0354, 0.0349, 0.0409, 0.0243, 0.0422], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:11:22,711 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-05-02 16:11:25,624 INFO [optim.py:368] (2/8) 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 16:11:49,571 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6457, 2.6135, 1.8821, 2.7517, 2.1048, 2.8354, 2.1085, 2.3851], device='cuda:2'), covar=tensor([0.0313, 0.0370, 0.1326, 0.0251, 0.0721, 0.0483, 0.1248, 0.0676], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0179, 0.0196, 0.0170, 0.0179, 0.0217, 0.0204, 0.0182], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:11:55,142 INFO [train.py:904] (2/8) Epoch 29, batch 200, loss[loss=0.1689, simple_loss=0.2624, pruned_loss=0.03769, over 15894.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2607, pruned_loss=0.0429, over 2094157.61 frames. ], batch size: 35, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:12:51,518 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284443.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:13:04,663 INFO [train.py:904] (2/8) Epoch 29, batch 250, loss[loss=0.1931, simple_loss=0.2647, pruned_loss=0.0607, over 16866.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2579, pruned_loss=0.04153, over 2373004.21 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:13:11,635 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0920, 5.6558, 5.7232, 5.4788, 5.5207, 6.1460, 5.5394, 5.3111], device='cuda:2'), covar=tensor([0.0956, 0.1979, 0.2476, 0.2214, 0.2830, 0.1023, 0.1755, 0.2382], device='cuda:2'), in_proj_covar=tensor([0.0417, 0.0625, 0.0693, 0.0510, 0.0680, 0.0716, 0.0538, 0.0679], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:13:47,576 INFO [optim.py:368] (2/8) 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 16:13:54,860 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7238, 6.0880, 5.8047, 5.8743, 5.4196, 5.5060, 5.5093, 6.2101], device='cuda:2'), covar=tensor([0.1491, 0.0964, 0.1175, 0.0945, 0.0989, 0.0673, 0.1322, 0.0975], device='cuda:2'), in_proj_covar=tensor([0.0698, 0.0847, 0.0693, 0.0655, 0.0535, 0.0534, 0.0709, 0.0662], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:14:08,387 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284498.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:16,992 INFO [train.py:904] (2/8) Epoch 29, batch 300, loss[loss=0.1441, simple_loss=0.2338, pruned_loss=0.02719, over 16865.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2549, pruned_loss=0.04025, over 2575271.69 frames. ], batch size: 42, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:14:17,386 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284504.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:26,621 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4231, 5.4106, 5.2077, 4.8084, 5.2444, 2.2121, 5.0581, 5.0871], device='cuda:2'), covar=tensor([0.0094, 0.0089, 0.0215, 0.0324, 0.0086, 0.2630, 0.0119, 0.0205], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0178, 0.0184, 0.0214, 0.0195, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:15:14,271 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284546.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:15:16,801 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8077, 2.8591, 2.4718, 2.7678, 3.1279, 2.9463, 3.4162, 3.4155], device='cuda:2'), covar=tensor([0.0231, 0.0539, 0.0633, 0.0515, 0.0372, 0.0479, 0.0376, 0.0322], device='cuda:2'), in_proj_covar=tensor([0.0229, 0.0245, 0.0235, 0.0235, 0.0245, 0.0244, 0.0239, 0.0242], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:15:23,226 INFO [train.py:904] (2/8) Epoch 29, batch 350, loss[loss=0.1381, simple_loss=0.2232, pruned_loss=0.02652, over 16812.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2518, pruned_loss=0.03852, over 2742658.48 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:15:37,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9756, 2.2000, 2.5837, 2.9924, 2.7473, 3.4288, 2.3996, 3.4026], device='cuda:2'), covar=tensor([0.0297, 0.0574, 0.0393, 0.0367, 0.0443, 0.0266, 0.0561, 0.0213], device='cuda:2'), in_proj_covar=tensor([0.0196, 0.0198, 0.0186, 0.0190, 0.0208, 0.0165, 0.0202, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:16:02,935 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 29, batch 400, loss[loss=0.1565, simple_loss=0.242, pruned_loss=0.03547, over 17061.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2503, pruned_loss=0.03813, over 2871659.74 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:16:57,004 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:17:41,207 INFO [train.py:904] (2/8) Epoch 29, batch 450, loss[loss=0.163, simple_loss=0.2483, pruned_loss=0.03884, over 16781.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2486, pruned_loss=0.03753, over 2979933.69 frames. ], batch size: 83, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:18:02,994 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284670.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:18:18,521 INFO [optim.py:368] (2/8) 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 16:18:36,768 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8677, 2.8671, 2.7178, 5.1128, 4.1117, 4.3942, 1.7956, 3.1922], device='cuda:2'), covar=tensor([0.1404, 0.0852, 0.1305, 0.0217, 0.0231, 0.0426, 0.1673, 0.0809], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0179, 0.0199, 0.0201, 0.0203, 0.0216, 0.0209, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:18:47,400 INFO [train.py:904] (2/8) Epoch 29, batch 500, loss[loss=0.1549, simple_loss=0.2372, pruned_loss=0.03634, over 15618.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2478, pruned_loss=0.03641, over 3063010.68 frames. ], batch size: 191, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:19:48,343 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 16:19:56,202 INFO [train.py:904] (2/8) Epoch 29, batch 550, loss[loss=0.1434, simple_loss=0.2349, pruned_loss=0.02589, over 16866.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2476, pruned_loss=0.03626, over 3119195.43 frames. ], batch size: 42, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:20:34,980 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7078, 4.6671, 4.5698, 4.0810, 4.6282, 1.9392, 4.4067, 4.2372], device='cuda:2'), covar=tensor([0.0193, 0.0142, 0.0231, 0.0335, 0.0126, 0.2719, 0.0175, 0.0258], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0173, 0.0211, 0.0181, 0.0186, 0.0217, 0.0198, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:20:35,715 INFO [optim.py:368] (2/8) 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 16:20:51,647 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1422, 5.1479, 5.0481, 4.5639, 4.7023, 5.0970, 4.9412, 4.7221], device='cuda:2'), covar=tensor([0.0667, 0.0639, 0.0392, 0.0388, 0.1129, 0.0505, 0.0426, 0.0846], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0469, 0.0364, 0.0366, 0.0360, 0.0421, 0.0251, 0.0434], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:20:57,936 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:21:04,192 INFO [train.py:904] (2/8) Epoch 29, batch 600, loss[loss=0.1694, simple_loss=0.2635, pruned_loss=0.03762, over 17112.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2471, pruned_loss=0.03672, over 3165617.87 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:21:21,962 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:22:12,365 INFO [train.py:904] (2/8) Epoch 29, batch 650, loss[loss=0.139, simple_loss=0.2292, pruned_loss=0.02443, over 16981.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.2458, pruned_loss=0.03686, over 3195509.24 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:22:46,757 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 16:23:22,615 INFO [train.py:904] (2/8) Epoch 29, batch 700, loss[loss=0.1417, simple_loss=0.2357, pruned_loss=0.02385, over 17211.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2457, pruned_loss=0.03669, over 3219553.85 frames. ], batch size: 44, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:24:04,473 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0058, 2.1586, 2.2267, 3.5346, 2.1847, 2.4195, 2.2401, 2.2933], device='cuda:2'), covar=tensor([0.1628, 0.3707, 0.3385, 0.0770, 0.3940, 0.2613, 0.4031, 0.3402], device='cuda:2'), in_proj_covar=tensor([0.0420, 0.0474, 0.0389, 0.0336, 0.0447, 0.0542, 0.0448, 0.0555], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:24:12,211 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:24:30,127 INFO [train.py:904] (2/8) Epoch 29, batch 750, loss[loss=0.1523, simple_loss=0.2459, pruned_loss=0.02933, over 16638.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2461, pruned_loss=0.03702, over 3236700.78 frames. ], batch size: 62, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:25:13,140 INFO [optim.py:368] (2/8) 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 16:25:30,055 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:37,920 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:42,935 INFO [train.py:904] (2/8) Epoch 29, batch 800, loss[loss=0.1649, simple_loss=0.2595, pruned_loss=0.03512, over 17064.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2459, pruned_loss=0.03711, over 3259232.06 frames. ], batch size: 53, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:26:08,925 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1696, 5.5280, 5.2978, 5.3439, 5.0503, 5.0438, 4.9635, 5.6560], device='cuda:2'), covar=tensor([0.1360, 0.1001, 0.1123, 0.0862, 0.0813, 0.0842, 0.1252, 0.0899], device='cuda:2'), in_proj_covar=tensor([0.0720, 0.0869, 0.0712, 0.0673, 0.0549, 0.0549, 0.0730, 0.0681], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:26:33,711 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285042.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:26:49,076 INFO [train.py:904] (2/8) Epoch 29, batch 850, loss[loss=0.1631, simple_loss=0.2414, pruned_loss=0.04243, over 16906.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2452, pruned_loss=0.03662, over 3281543.72 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:26:52,992 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:26:54,202 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9764, 2.7565, 2.7332, 4.4582, 3.5676, 4.1780, 1.6799, 3.1529], device='cuda:2'), covar=tensor([0.1376, 0.0765, 0.1171, 0.0181, 0.0208, 0.0398, 0.1680, 0.0804], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0200, 0.0203, 0.0204, 0.0217, 0.0210, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:27:00,118 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0079, 3.0509, 2.8375, 2.9989, 3.3068, 3.0192, 3.5805, 3.4812], device='cuda:2'), covar=tensor([0.0180, 0.0456, 0.0526, 0.0449, 0.0323, 0.0418, 0.0255, 0.0275], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0246, 0.0236, 0.0236, 0.0247, 0.0246, 0.0241, 0.0245], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:27:09,243 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9507, 4.4257, 4.4842, 3.2412, 3.6265, 4.4170, 3.9091, 2.7250], device='cuda:2'), covar=tensor([0.0484, 0.0087, 0.0045, 0.0376, 0.0171, 0.0102, 0.0114, 0.0455], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0102, 0.0115, 0.0098, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 16:27:31,496 INFO [optim.py:368] (2/8) 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 16:27:47,509 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 16:27:49,507 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:54,227 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285103.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:27:55,642 INFO [train.py:904] (2/8) Epoch 29, batch 900, loss[loss=0.1697, simple_loss=0.255, pruned_loss=0.04214, over 16739.00 frames. ], tot_loss[loss=0.1584, simple_loss=0.2447, pruned_loss=0.03608, over 3284158.57 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:28:20,427 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7975, 4.3940, 3.1486, 2.4321, 2.6770, 2.7067, 4.7574, 3.6009], device='cuda:2'), covar=tensor([0.3043, 0.0550, 0.1835, 0.3117, 0.3122, 0.2153, 0.0356, 0.1517], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0274, 0.0314, 0.0328, 0.0304, 0.0279, 0.0305, 0.0353], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:28:55,575 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285147.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:05,599 INFO [train.py:904] (2/8) Epoch 29, batch 950, loss[loss=0.1686, simple_loss=0.2654, pruned_loss=0.03595, over 16729.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2449, pruned_loss=0.03605, over 3291803.55 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:29:30,493 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:47,164 INFO [optim.py:368] (2/8) 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 16:29:52,611 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9178, 1.4106, 1.7243, 1.7417, 1.8487, 1.9650, 1.6922, 1.8713], device='cuda:2'), covar=tensor([0.0313, 0.0539, 0.0288, 0.0405, 0.0361, 0.0268, 0.0543, 0.0238], device='cuda:2'), in_proj_covar=tensor([0.0201, 0.0202, 0.0190, 0.0195, 0.0212, 0.0169, 0.0206, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 16:30:14,201 INFO [train.py:904] (2/8) Epoch 29, batch 1000, loss[loss=0.1561, simple_loss=0.2515, pruned_loss=0.0304, over 17124.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2433, pruned_loss=0.03586, over 3293036.72 frames. ], batch size: 48, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:30:24,350 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2630, 3.9377, 4.3927, 2.3838, 4.5937, 4.6843, 3.5015, 3.7112], device='cuda:2'), covar=tensor([0.0679, 0.0283, 0.0302, 0.1130, 0.0090, 0.0196, 0.0419, 0.0414], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0140, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:31:24,212 INFO [train.py:904] (2/8) Epoch 29, batch 1050, loss[loss=0.1602, simple_loss=0.2401, pruned_loss=0.04015, over 16782.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2425, pruned_loss=0.0355, over 3294522.34 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:32:05,303 INFO [optim.py:368] (2/8) 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 16:32:19,761 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:32:31,238 INFO [train.py:904] (2/8) Epoch 29, batch 1100, loss[loss=0.1687, simple_loss=0.2439, pruned_loss=0.04669, over 12197.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2422, pruned_loss=0.03535, over 3298275.84 frames. ], batch size: 247, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:33:37,332 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:33:40,187 INFO [train.py:904] (2/8) Epoch 29, batch 1150, loss[loss=0.1635, simple_loss=0.2372, pruned_loss=0.0449, over 16868.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2415, pruned_loss=0.03537, over 3300317.79 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:34:22,235 INFO [optim.py:368] (2/8) 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 16:34:40,275 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285398.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:34:47,458 INFO [train.py:904] (2/8) Epoch 29, batch 1200, loss[loss=0.145, simple_loss=0.2268, pruned_loss=0.03158, over 16425.00 frames. ], tot_loss[loss=0.1558, simple_loss=0.2416, pruned_loss=0.03501, over 3305594.77 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:35:47,422 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9831, 5.3277, 5.4626, 5.2082, 5.2863, 5.8809, 5.3950, 5.1837], device='cuda:2'), covar=tensor([0.1195, 0.2135, 0.2860, 0.2148, 0.2588, 0.1068, 0.1707, 0.2326], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0645, 0.0717, 0.0524, 0.0703, 0.0737, 0.0555, 0.0701], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:35:56,648 INFO [train.py:904] (2/8) Epoch 29, batch 1250, loss[loss=0.1558, simple_loss=0.2488, pruned_loss=0.03143, over 16618.00 frames. ], tot_loss[loss=0.156, simple_loss=0.2415, pruned_loss=0.03522, over 3305369.69 frames. ], batch size: 62, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:36:21,861 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:36:36,035 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 16:36:38,696 INFO [optim.py:368] (2/8) 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 16:36:40,482 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3551, 2.9775, 2.6913, 2.2767, 2.2258, 2.3545, 2.9767, 2.8118], device='cuda:2'), covar=tensor([0.2797, 0.0716, 0.1762, 0.2679, 0.2528, 0.2346, 0.0540, 0.1408], device='cuda:2'), in_proj_covar=tensor([0.0339, 0.0275, 0.0315, 0.0329, 0.0306, 0.0281, 0.0306, 0.0355], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:37:05,023 INFO [train.py:904] (2/8) Epoch 29, batch 1300, loss[loss=0.1446, simple_loss=0.2309, pruned_loss=0.02914, over 16759.00 frames. ], tot_loss[loss=0.1559, simple_loss=0.2418, pruned_loss=0.03501, over 3312339.38 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:37:27,838 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285520.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:38:04,194 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 16:38:12,058 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1901, 5.2640, 5.6533, 5.6246, 5.6595, 5.2904, 5.2292, 5.1055], device='cuda:2'), covar=tensor([0.0415, 0.0641, 0.0441, 0.0499, 0.0553, 0.0456, 0.1073, 0.0486], device='cuda:2'), in_proj_covar=tensor([0.0444, 0.0502, 0.0483, 0.0445, 0.0532, 0.0507, 0.0582, 0.0407], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 16:38:13,879 INFO [train.py:904] (2/8) Epoch 29, batch 1350, loss[loss=0.175, simple_loss=0.2593, pruned_loss=0.04532, over 12488.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2423, pruned_loss=0.03497, over 3315491.68 frames. ], batch size: 247, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:38:57,453 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8758, 5.1833, 5.2950, 5.0581, 5.1380, 5.7291, 5.1373, 4.8716], device='cuda:2'), covar=tensor([0.1347, 0.2224, 0.3019, 0.2397, 0.2596, 0.1032, 0.1930, 0.2682], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0643, 0.0716, 0.0524, 0.0701, 0.0734, 0.0554, 0.0699], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:38:58,360 INFO [optim.py:368] (2/8) 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 16:39:11,741 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285595.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:39:22,083 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5509, 5.9731, 5.6780, 5.8066, 5.2816, 5.4086, 5.3431, 6.0735], device='cuda:2'), covar=tensor([0.1603, 0.1023, 0.1157, 0.0947, 0.1036, 0.0759, 0.1389, 0.0982], device='cuda:2'), in_proj_covar=tensor([0.0731, 0.0881, 0.0722, 0.0685, 0.0558, 0.0557, 0.0742, 0.0692], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:39:24,823 INFO [train.py:904] (2/8) Epoch 29, batch 1400, loss[loss=0.1587, simple_loss=0.236, pruned_loss=0.04066, over 16747.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2429, pruned_loss=0.03507, over 3327464.14 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:39:49,620 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:12,991 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2665, 5.8745, 6.0071, 5.6999, 5.9327, 6.3533, 5.8451, 5.4895], device='cuda:2'), covar=tensor([0.0925, 0.1900, 0.2335, 0.2007, 0.2026, 0.0860, 0.1536, 0.2437], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0643, 0.0715, 0.0524, 0.0702, 0.0734, 0.0554, 0.0700], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:40:19,785 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:31,671 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:33,555 INFO [train.py:904] (2/8) Epoch 29, batch 1450, loss[loss=0.1715, simple_loss=0.2437, pruned_loss=0.04971, over 16846.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2422, pruned_loss=0.03495, over 3317919.07 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:41:13,911 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:16,390 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285698.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:41:37,508 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 29, batch 1500, loss[loss=0.1697, simple_loss=0.2669, pruned_loss=0.03621, over 17091.00 frames. ], tot_loss[loss=0.1562, simple_loss=0.2423, pruned_loss=0.03506, over 3323028.06 frames. ], batch size: 53, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:41:43,599 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7580, 6.1114, 5.8277, 5.9685, 5.5075, 5.5379, 5.5394, 6.2373], device='cuda:2'), covar=tensor([0.1567, 0.1014, 0.1165, 0.0962, 0.0985, 0.0695, 0.1241, 0.0973], device='cuda:2'), in_proj_covar=tensor([0.0733, 0.0883, 0.0724, 0.0686, 0.0559, 0.0558, 0.0744, 0.0693], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:41:49,596 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6782, 4.3515, 4.3300, 3.0094, 3.5988, 4.2928, 3.8715, 2.5263], device='cuda:2'), covar=tensor([0.0576, 0.0091, 0.0065, 0.0448, 0.0196, 0.0134, 0.0133, 0.0555], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0117, 0.0099, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 16:42:40,635 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285746.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:42:51,734 INFO [train.py:904] (2/8) Epoch 29, batch 1550, loss[loss=0.168, simple_loss=0.2681, pruned_loss=0.03399, over 17135.00 frames. ], tot_loss[loss=0.1578, simple_loss=0.2437, pruned_loss=0.03593, over 3310342.68 frames. ], batch size: 48, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:43:07,242 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0241, 3.7566, 4.3382, 2.1956, 4.4726, 4.6471, 3.3447, 3.4906], device='cuda:2'), covar=tensor([0.0764, 0.0305, 0.0257, 0.1199, 0.0092, 0.0155, 0.0458, 0.0456], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:43:07,278 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7411, 2.6557, 2.3065, 2.5609, 2.9976, 2.7919, 3.2860, 3.2532], device='cuda:2'), covar=tensor([0.0193, 0.0526, 0.0649, 0.0545, 0.0351, 0.0515, 0.0284, 0.0354], device='cuda:2'), in_proj_covar=tensor([0.0237, 0.0251, 0.0240, 0.0240, 0.0251, 0.0250, 0.0247, 0.0249], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:43:34,663 INFO [optim.py:368] (2/8) 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 16:44:00,607 INFO [train.py:904] (2/8) Epoch 29, batch 1600, loss[loss=0.1604, simple_loss=0.2433, pruned_loss=0.03873, over 16750.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2445, pruned_loss=0.03577, over 3315252.38 frames. ], batch size: 89, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:44:17,142 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1590, 4.9801, 5.2104, 5.3790, 5.5217, 4.8502, 5.5012, 5.5316], device='cuda:2'), covar=tensor([0.1971, 0.1304, 0.1823, 0.0811, 0.0645, 0.1085, 0.0747, 0.0758], device='cuda:2'), in_proj_covar=tensor([0.0699, 0.0853, 0.0985, 0.0869, 0.0660, 0.0683, 0.0727, 0.0840], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:44:33,739 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2713, 2.4033, 2.5416, 4.0362, 2.3398, 2.7360, 2.3936, 2.5658], device='cuda:2'), covar=tensor([0.1637, 0.4044, 0.3193, 0.0768, 0.4166, 0.2737, 0.4237, 0.3193], device='cuda:2'), in_proj_covar=tensor([0.0424, 0.0478, 0.0391, 0.0339, 0.0449, 0.0548, 0.0451, 0.0562], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:44:53,664 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 16:45:09,517 INFO [train.py:904] (2/8) Epoch 29, batch 1650, loss[loss=0.179, simple_loss=0.2619, pruned_loss=0.04809, over 16266.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2468, pruned_loss=0.03674, over 3313693.48 frames. ], batch size: 165, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:45:50,989 INFO [optim.py:368] (2/8) 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 16:46:16,869 INFO [train.py:904] (2/8) Epoch 29, batch 1700, loss[loss=0.1642, simple_loss=0.2504, pruned_loss=0.03899, over 16765.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2485, pruned_loss=0.03691, over 3318019.81 frames. ], batch size: 83, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:47:12,928 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 16:47:24,308 INFO [train.py:904] (2/8) Epoch 29, batch 1750, loss[loss=0.1408, simple_loss=0.2234, pruned_loss=0.02912, over 17008.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2491, pruned_loss=0.03656, over 3320803.90 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:47:58,390 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:48:07,053 INFO [optim.py:368] (2/8) 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 16:48:36,742 INFO [train.py:904] (2/8) Epoch 29, batch 1800, loss[loss=0.1911, simple_loss=0.2739, pruned_loss=0.0542, over 16735.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2489, pruned_loss=0.0362, over 3315512.69 frames. ], batch size: 134, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:49:43,956 INFO [train.py:904] (2/8) Epoch 29, batch 1850, loss[loss=0.1523, simple_loss=0.251, pruned_loss=0.02682, over 17260.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2503, pruned_loss=0.03728, over 3302260.94 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:50:15,973 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-02 16:50:28,097 INFO [optim.py:368] (2/8) 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 16:50:53,098 INFO [train.py:904] (2/8) Epoch 29, batch 1900, loss[loss=0.1544, simple_loss=0.2407, pruned_loss=0.03403, over 16839.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2501, pruned_loss=0.03742, over 3304644.72 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:51:20,242 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286123.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:51:29,237 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:51:55,666 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 16:51:58,611 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 16:51:59,708 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 16:52:04,245 INFO [train.py:904] (2/8) Epoch 29, batch 1950, loss[loss=0.1623, simple_loss=0.2567, pruned_loss=0.03399, over 16635.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2505, pruned_loss=0.03751, over 3310620.26 frames. ], batch size: 62, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:52:23,318 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5120, 3.5164, 2.7516, 2.1600, 2.2545, 2.3152, 3.5829, 3.0325], device='cuda:2'), covar=tensor([0.2933, 0.0620, 0.1838, 0.3258, 0.2857, 0.2264, 0.0576, 0.1735], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0277, 0.0316, 0.0330, 0.0308, 0.0281, 0.0307, 0.0357], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:52:45,817 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2442, 2.3790, 2.4850, 4.0915, 2.3912, 2.7009, 2.4336, 2.5935], device='cuda:2'), covar=tensor([0.1651, 0.4028, 0.3265, 0.0667, 0.4081, 0.2765, 0.4123, 0.3173], device='cuda:2'), in_proj_covar=tensor([0.0424, 0.0478, 0.0390, 0.0339, 0.0448, 0.0548, 0.0451, 0.0560], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:52:46,884 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286184.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 16:52:48,709 INFO [optim.py:368] (2/8) 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 16:52:49,178 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9950, 2.5686, 2.1181, 2.4028, 2.9215, 2.6712, 2.8864, 2.9853], device='cuda:2'), covar=tensor([0.0318, 0.0562, 0.0682, 0.0544, 0.0329, 0.0460, 0.0342, 0.0374], device='cuda:2'), in_proj_covar=tensor([0.0239, 0.0253, 0.0240, 0.0241, 0.0253, 0.0251, 0.0249, 0.0251], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 16:52:55,391 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286190.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:53:10,078 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1379, 3.1013, 2.2062, 3.2847, 2.5235, 3.2952, 2.2598, 2.6956], device='cuda:2'), covar=tensor([0.0356, 0.0475, 0.1430, 0.0332, 0.0793, 0.0716, 0.1412, 0.0718], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0184, 0.0200, 0.0178, 0.0183, 0.0225, 0.0209, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 16:53:13,092 INFO [train.py:904] (2/8) Epoch 29, batch 2000, loss[loss=0.1594, simple_loss=0.2579, pruned_loss=0.03046, over 17124.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2506, pruned_loss=0.03731, over 3311266.23 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:21,795 INFO [train.py:904] (2/8) Epoch 29, batch 2050, loss[loss=0.1567, simple_loss=0.261, pruned_loss=0.0262, over 17290.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2505, pruned_loss=0.03699, over 3308426.26 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:54,809 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:55:04,677 INFO [optim.py:368] (2/8) 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 16:55:05,297 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0966, 4.6406, 3.3238, 2.5125, 2.7464, 2.7863, 4.9576, 3.6527], device='cuda:2'), covar=tensor([0.2735, 0.0498, 0.1800, 0.3136, 0.3254, 0.2213, 0.0324, 0.1543], device='cuda:2'), in_proj_covar=tensor([0.0341, 0.0278, 0.0318, 0.0331, 0.0310, 0.0283, 0.0309, 0.0358], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 16:55:18,573 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:55:30,415 INFO [train.py:904] (2/8) Epoch 29, batch 2100, loss[loss=0.1589, simple_loss=0.2607, pruned_loss=0.02851, over 17304.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2517, pruned_loss=0.0375, over 3313378.07 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:55:54,471 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 16:56:00,344 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286326.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:26,248 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:40,456 INFO [train.py:904] (2/8) Epoch 29, batch 2150, loss[loss=0.157, simple_loss=0.257, pruned_loss=0.02847, over 17130.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2517, pruned_loss=0.03777, over 3309073.78 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:44,725 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286356.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:57:24,983 INFO [optim.py:368] (2/8) 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 16:57:50,675 INFO [train.py:904] (2/8) Epoch 29, batch 2200, loss[loss=0.1381, simple_loss=0.233, pruned_loss=0.02163, over 17126.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2523, pruned_loss=0.03873, over 3301802.17 frames. ], batch size: 48, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:57:52,931 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:58:59,528 INFO [train.py:904] (2/8) Epoch 29, batch 2250, loss[loss=0.1939, simple_loss=0.271, pruned_loss=0.05838, over 16899.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2528, pruned_loss=0.03881, over 3313999.31 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:59:07,424 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 16:59:33,875 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286479.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:59:43,695 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286485.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:59:45,656 INFO [optim.py:368] (2/8) 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 16:59:48,328 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8508, 4.9316, 5.3315, 5.3212, 5.3264, 4.9701, 4.9460, 4.7502], device='cuda:2'), covar=tensor([0.0389, 0.0565, 0.0407, 0.0412, 0.0519, 0.0413, 0.0964, 0.0538], device='cuda:2'), in_proj_covar=tensor([0.0450, 0.0509, 0.0491, 0.0452, 0.0539, 0.0514, 0.0591, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 17:00:08,076 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8195, 4.3797, 3.0208, 2.3575, 2.5752, 2.5083, 4.7543, 3.5720], device='cuda:2'), covar=tensor([0.3112, 0.0570, 0.2032, 0.3238, 0.3368, 0.2416, 0.0342, 0.1557], device='cuda:2'), in_proj_covar=tensor([0.0340, 0.0278, 0.0318, 0.0331, 0.0309, 0.0283, 0.0309, 0.0358], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:00:08,671 INFO [train.py:904] (2/8) Epoch 29, batch 2300, loss[loss=0.1618, simple_loss=0.2574, pruned_loss=0.03311, over 17119.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2536, pruned_loss=0.03867, over 3323510.62 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:00:12,316 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2223, 3.3138, 3.7927, 2.2031, 3.1624, 2.5325, 3.6465, 3.6161], device='cuda:2'), covar=tensor([0.0280, 0.1008, 0.0517, 0.2142, 0.0864, 0.0995, 0.0595, 0.1074], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0173, 0.0172, 0.0160, 0.0150, 0.0134, 0.0148, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 17:00:28,094 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0124, 5.1113, 5.5078, 5.5057, 5.5064, 5.1568, 5.1307, 4.9311], device='cuda:2'), covar=tensor([0.0394, 0.0543, 0.0393, 0.0387, 0.0446, 0.0427, 0.0881, 0.0482], device='cuda:2'), in_proj_covar=tensor([0.0448, 0.0506, 0.0490, 0.0450, 0.0537, 0.0513, 0.0589, 0.0413], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 17:00:51,914 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0883, 2.9885, 2.7042, 4.3540, 3.6045, 4.1536, 1.7720, 3.1028], device='cuda:2'), covar=tensor([0.1259, 0.0633, 0.1095, 0.0173, 0.0134, 0.0375, 0.1483, 0.0794], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0206, 0.0207, 0.0221, 0.0211, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 17:01:13,721 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1490, 5.6335, 5.8236, 5.4872, 5.5914, 6.1641, 5.6937, 5.4125], device='cuda:2'), covar=tensor([0.0893, 0.1839, 0.2601, 0.1910, 0.2462, 0.0944, 0.1465, 0.2224], device='cuda:2'), in_proj_covar=tensor([0.0440, 0.0654, 0.0728, 0.0532, 0.0714, 0.0747, 0.0562, 0.0710], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:01:17,610 INFO [train.py:904] (2/8) Epoch 29, batch 2350, loss[loss=0.1475, simple_loss=0.2509, pruned_loss=0.02203, over 17129.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2539, pruned_loss=0.03858, over 3314622.45 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:01:22,208 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7542, 4.6332, 4.6610, 4.3293, 4.3967, 4.6822, 4.4825, 4.4503], device='cuda:2'), covar=tensor([0.0635, 0.0883, 0.0345, 0.0357, 0.0807, 0.0496, 0.0488, 0.0676], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0489, 0.0380, 0.0382, 0.0375, 0.0438, 0.0260, 0.0454], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:02:03,068 INFO [optim.py:368] (2/8) 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 17:02:04,872 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286588.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:02:27,440 INFO [train.py:904] (2/8) Epoch 29, batch 2400, loss[loss=0.1458, simple_loss=0.2286, pruned_loss=0.03146, over 16827.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2541, pruned_loss=0.03882, over 3320457.63 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 17:03:30,985 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:31,998 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9618, 4.9919, 5.4280, 5.3992, 5.4473, 5.0849, 5.0334, 4.8249], device='cuda:2'), covar=tensor([0.0380, 0.0568, 0.0382, 0.0441, 0.0511, 0.0424, 0.0992, 0.0494], device='cuda:2'), in_proj_covar=tensor([0.0449, 0.0509, 0.0491, 0.0452, 0.0539, 0.0515, 0.0592, 0.0415], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-02 17:03:34,191 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286651.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:37,417 INFO [train.py:904] (2/8) Epoch 29, batch 2450, loss[loss=0.1629, simple_loss=0.2602, pruned_loss=0.03282, over 16996.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2547, pruned_loss=0.03872, over 3327799.12 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:03:59,950 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:04:23,752 INFO [optim.py:368] (2/8) 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 17:04:41,155 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:04:46,153 INFO [train.py:904] (2/8) Epoch 29, batch 2500, loss[loss=0.1781, simple_loss=0.2603, pruned_loss=0.04789, over 16430.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2542, pruned_loss=0.03843, over 3335274.53 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:05:16,474 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-05-02 17:05:24,969 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286732.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:05:55,678 INFO [train.py:904] (2/8) Epoch 29, batch 2550, loss[loss=0.1604, simple_loss=0.2506, pruned_loss=0.03513, over 17192.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2551, pruned_loss=0.03884, over 3323256.24 frames. ], batch size: 46, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:06:18,989 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:32,964 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286779.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:06:40,999 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 17:07:07,735 INFO [train.py:904] (2/8) Epoch 29, batch 2600, loss[loss=0.152, simple_loss=0.2407, pruned_loss=0.0317, over 17001.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2545, pruned_loss=0.03822, over 3324440.84 frames. ], batch size: 41, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:07:39,060 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286827.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:07:40,558 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-05-02 17:07:44,485 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:07:47,330 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:08:15,902 INFO [train.py:904] (2/8) Epoch 29, batch 2650, loss[loss=0.1823, simple_loss=0.2743, pruned_loss=0.0451, over 16683.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2556, pruned_loss=0.03811, over 3321543.22 frames. ], batch size: 76, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:08:30,040 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-05-02 17:09:00,291 INFO [optim.py:368] (2/8) 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 17:09:22,332 INFO [train.py:904] (2/8) Epoch 29, batch 2700, loss[loss=0.1675, simple_loss=0.2502, pruned_loss=0.04236, over 16394.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2552, pruned_loss=0.03752, over 3331234.23 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:09:51,858 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5023, 3.7608, 3.9114, 2.7523, 3.5635, 3.9860, 3.6537, 2.3890], device='cuda:2'), covar=tensor([0.0571, 0.0220, 0.0073, 0.0435, 0.0138, 0.0111, 0.0118, 0.0488], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0091, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 17:10:17,436 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286944.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:26,952 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286951.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:30,031 INFO [train.py:904] (2/8) Epoch 29, batch 2750, loss[loss=0.1688, simple_loss=0.2672, pruned_loss=0.03518, over 16669.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2557, pruned_loss=0.03711, over 3338815.26 frames. ], batch size: 57, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:17,993 INFO [optim.py:368] (2/8) 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 17:11:32,917 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:35,335 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:40,740 INFO [train.py:904] (2/8) Epoch 29, batch 2800, loss[loss=0.1533, simple_loss=0.245, pruned_loss=0.03084, over 17234.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.255, pruned_loss=0.03721, over 3340658.87 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:41,084 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4766, 5.8833, 5.6344, 5.6802, 5.3404, 5.3285, 5.3124, 6.0428], device='cuda:2'), covar=tensor([0.1583, 0.1058, 0.1093, 0.0921, 0.0922, 0.0750, 0.1339, 0.0974], device='cuda:2'), in_proj_covar=tensor([0.0735, 0.0889, 0.0728, 0.0689, 0.0562, 0.0560, 0.0747, 0.0699], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:12:12,974 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:42,702 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:50,055 INFO [train.py:904] (2/8) Epoch 29, batch 2850, loss[loss=0.1659, simple_loss=0.2545, pruned_loss=0.03867, over 16822.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2544, pruned_loss=0.03685, over 3337811.85 frames. ], batch size: 39, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:13:39,986 INFO [optim.py:368] (2/8) 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 17:14:00,573 INFO [train.py:904] (2/8) Epoch 29, batch 2900, loss[loss=0.1433, simple_loss=0.2288, pruned_loss=0.02895, over 15994.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2524, pruned_loss=0.0372, over 3336439.59 frames. ], batch size: 35, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:14:31,456 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:15:10,985 INFO [train.py:904] (2/8) Epoch 29, batch 2950, loss[loss=0.1417, simple_loss=0.2303, pruned_loss=0.02651, over 17041.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2524, pruned_loss=0.03779, over 3322694.25 frames. ], batch size: 41, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:15:34,330 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7299, 4.1631, 3.0143, 2.3812, 2.6684, 2.6399, 4.4323, 3.4782], device='cuda:2'), covar=tensor([0.3085, 0.0630, 0.1914, 0.3138, 0.3090, 0.2214, 0.0420, 0.1551], device='cuda:2'), in_proj_covar=tensor([0.0339, 0.0278, 0.0317, 0.0331, 0.0309, 0.0282, 0.0309, 0.0358], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:15:59,392 INFO [optim.py:368] (2/8) 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 17:16:09,304 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8050, 3.7050, 3.8827, 3.5394, 3.8064, 4.2583, 3.9014, 3.5231], device='cuda:2'), covar=tensor([0.2275, 0.2699, 0.3147, 0.2903, 0.2927, 0.2199, 0.1918, 0.2868], device='cuda:2'), in_proj_covar=tensor([0.0442, 0.0658, 0.0731, 0.0537, 0.0718, 0.0751, 0.0565, 0.0712], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:16:20,172 INFO [train.py:904] (2/8) Epoch 29, batch 3000, loss[loss=0.1746, simple_loss=0.2493, pruned_loss=0.05001, over 16924.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2525, pruned_loss=0.03805, over 3324821.09 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:16:20,172 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 17:16:28,741 INFO [train.py:938] (2/8) Epoch 29, validation: loss=0.1336, simple_loss=0.2385, pruned_loss=0.01438, over 944034.00 frames. 2023-05-02 17:16:28,742 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 17:16:30,359 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6555, 3.8316, 2.4811, 4.5433, 3.0064, 4.4452, 2.5426, 3.2651], device='cuda:2'), covar=tensor([0.0409, 0.0464, 0.1727, 0.0317, 0.0919, 0.0600, 0.1669, 0.0766], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0185, 0.0200, 0.0179, 0.0183, 0.0227, 0.0208, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 17:16:37,791 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 17:16:53,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0641, 2.3406, 2.8143, 3.1084, 3.0346, 3.6215, 2.6778, 3.5658], device='cuda:2'), covar=tensor([0.0341, 0.0573, 0.0401, 0.0384, 0.0384, 0.0232, 0.0530, 0.0229], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0205, 0.0192, 0.0199, 0.0216, 0.0172, 0.0209, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 17:17:12,387 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.7206, 6.1025, 5.8017, 5.9139, 5.4929, 5.5887, 5.4460, 6.2228], device='cuda:2'), covar=tensor([0.1235, 0.0940, 0.1043, 0.0881, 0.0940, 0.0699, 0.1393, 0.0878], device='cuda:2'), in_proj_covar=tensor([0.0735, 0.0889, 0.0728, 0.0690, 0.0562, 0.0560, 0.0746, 0.0697], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:17:25,966 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:17:40,002 INFO [train.py:904] (2/8) Epoch 29, batch 3050, loss[loss=0.1697, simple_loss=0.2715, pruned_loss=0.034, over 16704.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03889, over 3301599.68 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:18:29,494 INFO [optim.py:368] (2/8) 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 17:18:34,331 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287292.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:18:49,717 INFO [train.py:904] (2/8) Epoch 29, batch 3100, loss[loss=0.1693, simple_loss=0.2629, pruned_loss=0.03783, over 16797.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2527, pruned_loss=0.03889, over 3309875.65 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:19:03,738 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 17:19:14,341 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 17:19:17,152 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9405, 2.7247, 2.6184, 4.7657, 3.7621, 4.1762, 1.5976, 3.1087], device='cuda:2'), covar=tensor([0.1425, 0.0863, 0.1304, 0.0235, 0.0246, 0.0421, 0.1822, 0.0893], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0207, 0.0207, 0.0221, 0.0211, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 17:19:22,428 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:00,321 INFO [train.py:904] (2/8) Epoch 29, batch 3150, loss[loss=0.1668, simple_loss=0.2452, pruned_loss=0.0442, over 16896.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2523, pruned_loss=0.03875, over 3300549.38 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:20:30,293 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:49,321 INFO [optim.py:368] (2/8) 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 17:20:52,002 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5805, 3.6516, 3.4047, 3.0664, 3.2863, 3.5324, 3.2847, 3.3724], device='cuda:2'), covar=tensor([0.0609, 0.0604, 0.0290, 0.0313, 0.0511, 0.0491, 0.1876, 0.0490], device='cuda:2'), in_proj_covar=tensor([0.0327, 0.0493, 0.0383, 0.0385, 0.0379, 0.0443, 0.0263, 0.0458], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:21:10,100 INFO [train.py:904] (2/8) Epoch 29, batch 3200, loss[loss=0.1712, simple_loss=0.2517, pruned_loss=0.0453, over 16414.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2517, pruned_loss=0.03892, over 3307265.74 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:21:20,081 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:21:40,352 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:18,322 INFO [train.py:904] (2/8) Epoch 29, batch 3250, loss[loss=0.1993, simple_loss=0.2788, pruned_loss=0.05991, over 15678.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2523, pruned_loss=0.03894, over 3312314.87 frames. ], batch size: 191, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:22:43,503 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:45,674 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287474.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:23:05,921 INFO [optim.py:368] (2/8) 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 17:23:26,857 INFO [train.py:904] (2/8) Epoch 29, batch 3300, loss[loss=0.1879, simple_loss=0.2715, pruned_loss=0.05212, over 16753.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2529, pruned_loss=0.03865, over 3316739.27 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:11,400 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:24:34,315 INFO [train.py:904] (2/8) Epoch 29, batch 3350, loss[loss=0.2051, simple_loss=0.2854, pruned_loss=0.06242, over 16408.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2537, pruned_loss=0.03893, over 3313401.50 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:46,521 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287562.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:14,550 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 17:25:33,305 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:35,701 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7715, 4.0453, 2.6162, 4.6608, 3.2036, 4.5588, 2.7715, 3.3848], device='cuda:2'), covar=tensor([0.0397, 0.0446, 0.1590, 0.0286, 0.0850, 0.0558, 0.1485, 0.0727], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0185, 0.0199, 0.0179, 0.0183, 0.0226, 0.0208, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 17:25:43,315 INFO [train.py:904] (2/8) Epoch 29, batch 3400, loss[loss=0.1729, simple_loss=0.2632, pruned_loss=0.0413, over 12409.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2532, pruned_loss=0.03846, over 3315333.95 frames. ], batch size: 247, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:26:10,381 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:40,174 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287644.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:53,544 INFO [train.py:904] (2/8) Epoch 29, batch 3450, loss[loss=0.1596, simple_loss=0.2617, pruned_loss=0.02875, over 16679.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2515, pruned_loss=0.03783, over 3314287.55 frames. ], batch size: 57, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:27:43,293 INFO [optim.py:368] (2/8) 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 17:27:44,452 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-05-02 17:27:51,324 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7837, 5.0926, 5.4274, 5.3858, 5.4302, 5.0998, 4.8329, 4.9108], device='cuda:2'), covar=tensor([0.0676, 0.0749, 0.0607, 0.0792, 0.0785, 0.0698, 0.1497, 0.0590], device='cuda:2'), in_proj_covar=tensor([0.0453, 0.0514, 0.0497, 0.0456, 0.0545, 0.0521, 0.0600, 0.0419], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-02 17:28:04,224 INFO [train.py:904] (2/8) Epoch 29, batch 3500, loss[loss=0.1323, simple_loss=0.2196, pruned_loss=0.02249, over 17261.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2496, pruned_loss=0.0372, over 3315497.26 frames. ], batch size: 45, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:28:06,056 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8133, 2.4538, 1.9925, 2.2135, 2.8040, 2.5741, 2.7341, 2.9293], device='cuda:2'), covar=tensor([0.0292, 0.0500, 0.0706, 0.0555, 0.0317, 0.0391, 0.0272, 0.0357], device='cuda:2'), in_proj_covar=tensor([0.0242, 0.0252, 0.0240, 0.0241, 0.0254, 0.0252, 0.0250, 0.0252], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:28:26,813 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:14,561 INFO [train.py:904] (2/8) Epoch 29, batch 3550, loss[loss=0.1419, simple_loss=0.2242, pruned_loss=0.02981, over 16999.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2487, pruned_loss=0.03717, over 3320929.83 frames. ], batch size: 41, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:29:33,116 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:42,479 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:53,968 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287781.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:30:07,185 INFO [optim.py:368] (2/8) 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 17:30:26,881 INFO [train.py:904] (2/8) Epoch 29, batch 3600, loss[loss=0.1715, simple_loss=0.2497, pruned_loss=0.04665, over 16711.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2477, pruned_loss=0.03711, over 3316587.91 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:31:10,894 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287834.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:31:40,144 INFO [train.py:904] (2/8) Epoch 29, batch 3650, loss[loss=0.1779, simple_loss=0.2673, pruned_loss=0.04426, over 17037.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2473, pruned_loss=0.03786, over 3315923.69 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:31:55,933 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9514, 2.5383, 2.1077, 2.3763, 2.9229, 2.6466, 2.8782, 3.0122], device='cuda:2'), covar=tensor([0.0230, 0.0487, 0.0646, 0.0506, 0.0295, 0.0371, 0.0289, 0.0346], device='cuda:2'), in_proj_covar=tensor([0.0242, 0.0252, 0.0241, 0.0241, 0.0254, 0.0252, 0.0250, 0.0253], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:32:35,500 INFO [optim.py:368] (2/8) 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 17:32:38,196 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287892.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:32:57,081 INFO [train.py:904] (2/8) Epoch 29, batch 3700, loss[loss=0.1587, simple_loss=0.2361, pruned_loss=0.04063, over 16916.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2471, pruned_loss=0.03915, over 3279621.53 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:33:18,785 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:33:48,732 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:34:09,992 INFO [train.py:904] (2/8) Epoch 29, batch 3750, loss[loss=0.1976, simple_loss=0.2741, pruned_loss=0.06054, over 11406.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2477, pruned_loss=0.04053, over 3264971.38 frames. ], batch size: 247, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:35:05,542 INFO [optim.py:368] (2/8) 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 17:35:29,273 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288003.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:35:30,035 INFO [train.py:904] (2/8) Epoch 29, batch 3800, loss[loss=0.1766, simple_loss=0.2562, pruned_loss=0.04846, over 16527.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2494, pruned_loss=0.0418, over 3253714.66 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:03,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1050, 2.2252, 2.3276, 3.8481, 2.2208, 2.5650, 2.2967, 2.3779], device='cuda:2'), covar=tensor([0.1685, 0.3942, 0.3191, 0.0679, 0.3971, 0.2666, 0.4238, 0.3035], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0484, 0.0393, 0.0344, 0.0451, 0.0554, 0.0455, 0.0567], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:36:29,046 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5593, 3.5533, 2.5891, 2.2603, 2.3568, 2.2206, 3.5911, 3.1155], device='cuda:2'), covar=tensor([0.2983, 0.0646, 0.2123, 0.3108, 0.2891, 0.2453, 0.0625, 0.1614], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0277, 0.0316, 0.0330, 0.0310, 0.0281, 0.0308, 0.0358], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:36:40,376 INFO [train.py:904] (2/8) Epoch 29, batch 3850, loss[loss=0.1617, simple_loss=0.2553, pruned_loss=0.03403, over 17108.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2492, pruned_loss=0.04248, over 3270405.87 frames. ], batch size: 47, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:55,430 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288064.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:36:59,078 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288067.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:37:12,500 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 17:37:50,142 INFO [train.py:904] (2/8) Epoch 29, batch 3900, loss[loss=0.1635, simple_loss=0.2401, pruned_loss=0.04343, over 16242.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2484, pruned_loss=0.04252, over 3273786.74 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:37:53,667 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288106.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:37:55,528 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9067, 3.0497, 3.2078, 2.1658, 2.8816, 2.2386, 3.3743, 3.4256], device='cuda:2'), covar=tensor([0.0312, 0.1058, 0.0720, 0.2058, 0.0932, 0.1118, 0.0614, 0.0964], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0174, 0.0172, 0.0158, 0.0149, 0.0134, 0.0147, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 17:38:05,839 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:27,297 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:39:00,679 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288152.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:39:03,100 INFO [train.py:904] (2/8) Epoch 29, batch 3950, loss[loss=0.2023, simple_loss=0.2798, pruned_loss=0.06247, over 12299.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2485, pruned_loss=0.04306, over 3273256.50 frames. ], batch size: 247, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:39:22,309 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288167.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:39:45,297 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7525, 2.7191, 2.5711, 4.3257, 3.4570, 4.1078, 1.6227, 2.8938], device='cuda:2'), covar=tensor([0.1455, 0.0756, 0.1216, 0.0195, 0.0177, 0.0368, 0.1704, 0.0917], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0207, 0.0207, 0.0220, 0.0210, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 17:39:55,369 INFO [optim.py:368] (2/8) 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 17:39:58,798 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:40:16,237 INFO [train.py:904] (2/8) Epoch 29, batch 4000, loss[loss=0.2098, simple_loss=0.2888, pruned_loss=0.06538, over 12519.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2493, pruned_loss=0.04394, over 3268152.27 frames. ], batch size: 245, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:40:30,228 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288213.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:40:36,638 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:00,388 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8290, 4.7626, 4.8973, 5.0445, 5.1783, 4.6931, 5.1661, 5.2333], device='cuda:2'), covar=tensor([0.1781, 0.1046, 0.1468, 0.0644, 0.0505, 0.0953, 0.0590, 0.0542], device='cuda:2'), in_proj_covar=tensor([0.0709, 0.0864, 0.0995, 0.0877, 0.0669, 0.0693, 0.0732, 0.0849], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:41:07,384 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288239.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:08,412 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:24,440 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8422, 2.2209, 2.4079, 3.1215, 2.2077, 2.3671, 2.3418, 2.3309], device='cuda:2'), covar=tensor([0.1555, 0.3386, 0.2802, 0.0841, 0.4046, 0.2486, 0.3437, 0.3073], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0484, 0.0392, 0.0344, 0.0451, 0.0555, 0.0455, 0.0567], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:41:29,245 INFO [train.py:904] (2/8) Epoch 29, batch 4050, loss[loss=0.1563, simple_loss=0.2439, pruned_loss=0.03432, over 16868.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.25, pruned_loss=0.04343, over 3268542.09 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:41:47,212 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288266.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:17,220 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:21,065 INFO [optim.py:368] (2/8) 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 17:42:42,062 INFO [train.py:904] (2/8) Epoch 29, batch 4100, loss[loss=0.187, simple_loss=0.2742, pruned_loss=0.04992, over 16854.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2521, pruned_loss=0.04294, over 3268374.83 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:43:10,490 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 17:43:17,870 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:45,762 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:59,219 INFO [train.py:904] (2/8) Epoch 29, batch 4150, loss[loss=0.2041, simple_loss=0.2943, pruned_loss=0.05698, over 16613.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.259, pruned_loss=0.04513, over 3233709.19 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:44:05,797 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 17:44:07,904 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:33,293 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:40,180 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-05-02 17:44:51,063 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 17:45:02,628 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:13,941 INFO [train.py:904] (2/8) Epoch 29, batch 4200, loss[loss=0.2061, simple_loss=0.3064, pruned_loss=0.05289, over 16428.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2659, pruned_loss=0.04647, over 3213077.36 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:45:18,113 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:45,768 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:53,645 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:54,691 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4699, 4.4869, 4.6974, 4.4232, 4.5134, 5.0453, 4.5601, 4.2399], device='cuda:2'), covar=tensor([0.1262, 0.2040, 0.1513, 0.1971, 0.2176, 0.0784, 0.1263, 0.2275], device='cuda:2'), in_proj_covar=tensor([0.0433, 0.0646, 0.0714, 0.0524, 0.0700, 0.0734, 0.0553, 0.0696], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:46:30,280 INFO [train.py:904] (2/8) Epoch 29, batch 4250, loss[loss=0.1578, simple_loss=0.2649, pruned_loss=0.02531, over 16913.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.269, pruned_loss=0.04589, over 3209193.28 frames. ], batch size: 96, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:46:36,676 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:43,221 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288462.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:46:54,844 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 17:47:05,783 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:47:22,571 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5331, 3.7691, 2.7385, 2.2652, 2.5000, 2.5874, 4.0640, 3.2602], device='cuda:2'), covar=tensor([0.3188, 0.0642, 0.2058, 0.2847, 0.2831, 0.2145, 0.0488, 0.1407], device='cuda:2'), in_proj_covar=tensor([0.0336, 0.0275, 0.0314, 0.0329, 0.0309, 0.0280, 0.0307, 0.0355], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:47:23,909 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5058, 3.5034, 2.6131, 2.1687, 2.3112, 2.3311, 3.6612, 3.1095], device='cuda:2'), covar=tensor([0.3003, 0.0620, 0.1992, 0.2885, 0.2890, 0.2337, 0.0555, 0.1366], device='cuda:2'), in_proj_covar=tensor([0.0336, 0.0275, 0.0314, 0.0329, 0.0309, 0.0280, 0.0307, 0.0355], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:47:24,431 INFO [optim.py:368] (2/8) 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 17:47:45,985 INFO [train.py:904] (2/8) Epoch 29, batch 4300, loss[loss=0.1827, simple_loss=0.2769, pruned_loss=0.04423, over 16901.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2699, pruned_loss=0.04492, over 3210883.24 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:47:51,124 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288508.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:48:59,343 INFO [train.py:904] (2/8) Epoch 29, batch 4350, loss[loss=0.1933, simple_loss=0.2977, pruned_loss=0.04447, over 16791.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2734, pruned_loss=0.04606, over 3206359.20 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:49:09,526 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6412, 2.5683, 2.0904, 2.4442, 2.9306, 2.5992, 3.0993, 3.2246], device='cuda:2'), covar=tensor([0.0113, 0.0512, 0.0722, 0.0535, 0.0329, 0.0467, 0.0279, 0.0290], device='cuda:2'), in_proj_covar=tensor([0.0237, 0.0248, 0.0236, 0.0237, 0.0248, 0.0247, 0.0245, 0.0247], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:49:42,906 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 2023-05-02 17:49:53,411 INFO [optim.py:368] (2/8) 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 17:50:13,841 INFO [train.py:904] (2/8) Epoch 29, batch 4400, loss[loss=0.1882, simple_loss=0.275, pruned_loss=0.05065, over 16707.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2758, pruned_loss=0.04778, over 3207153.28 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:50:22,259 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:50:59,550 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3541, 5.6302, 5.4001, 5.5039, 5.2234, 5.0518, 5.0944, 5.7643], device='cuda:2'), covar=tensor([0.1188, 0.0788, 0.0974, 0.0820, 0.0768, 0.0725, 0.1112, 0.0790], device='cuda:2'), in_proj_covar=tensor([0.0730, 0.0877, 0.0718, 0.0684, 0.0558, 0.0557, 0.0736, 0.0690], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:51:26,299 INFO [train.py:904] (2/8) Epoch 29, batch 4450, loss[loss=0.2014, simple_loss=0.2943, pruned_loss=0.05424, over 15369.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.279, pruned_loss=0.04907, over 3203896.81 frames. ], batch size: 190, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:51:34,144 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:49,616 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288670.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:52:08,448 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:18,786 INFO [optim.py:368] (2/8) 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 17:52:20,828 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0093, 4.1179, 4.3221, 4.2907, 4.3179, 4.0785, 4.1056, 4.0468], device='cuda:2'), covar=tensor([0.0337, 0.0480, 0.0380, 0.0394, 0.0418, 0.0414, 0.0838, 0.0526], device='cuda:2'), in_proj_covar=tensor([0.0441, 0.0499, 0.0481, 0.0442, 0.0529, 0.0506, 0.0586, 0.0410], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 17:52:34,490 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288701.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:37,603 INFO [train.py:904] (2/8) Epoch 29, batch 4500, loss[loss=0.2109, simple_loss=0.2977, pruned_loss=0.06202, over 16702.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2802, pruned_loss=0.0505, over 3195434.84 frames. ], batch size: 89, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:52:42,934 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:04,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6027, 4.6666, 4.4567, 3.0399, 3.9921, 4.5147, 3.9185, 2.8501], device='cuda:2'), covar=tensor([0.0577, 0.0024, 0.0049, 0.0406, 0.0089, 0.0078, 0.0093, 0.0400], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0136, 0.0103, 0.0117, 0.0099, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 17:53:49,128 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:51,867 INFO [train.py:904] (2/8) Epoch 29, batch 4550, loss[loss=0.1823, simple_loss=0.2743, pruned_loss=0.04519, over 16741.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2805, pruned_loss=0.05129, over 3204237.39 frames. ], batch size: 39, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:54:03,033 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288762.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:54:44,107 INFO [optim.py:368] (2/8) 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 17:54:53,765 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:55:04,695 INFO [train.py:904] (2/8) Epoch 29, batch 4600, loss[loss=0.2365, simple_loss=0.3058, pruned_loss=0.08356, over 11699.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2813, pruned_loss=0.05175, over 3193098.32 frames. ], batch size: 246, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:55:11,596 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288808.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:55:14,423 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288810.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:56:18,393 INFO [train.py:904] (2/8) Epoch 29, batch 4650, loss[loss=0.1863, simple_loss=0.277, pruned_loss=0.04779, over 16523.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2809, pruned_loss=0.05204, over 3213478.93 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:56:21,002 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288856.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:56:24,288 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:57:10,669 INFO [optim.py:368] (2/8) 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 17:57:23,796 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0189, 2.2515, 2.3011, 3.5935, 2.2151, 2.5622, 2.3409, 2.3930], device='cuda:2'), covar=tensor([0.1667, 0.3568, 0.3182, 0.0716, 0.4120, 0.2643, 0.3590, 0.3564], device='cuda:2'), in_proj_covar=tensor([0.0425, 0.0480, 0.0389, 0.0340, 0.0449, 0.0552, 0.0452, 0.0562], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:57:29,734 INFO [train.py:904] (2/8) Epoch 29, batch 4700, loss[loss=0.1761, simple_loss=0.2594, pruned_loss=0.04637, over 16575.00 frames. ], tot_loss[loss=0.19, simple_loss=0.278, pruned_loss=0.051, over 3218348.03 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:57:42,462 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8795, 4.9813, 5.2200, 5.2001, 5.2491, 4.9597, 4.8972, 4.6631], device='cuda:2'), covar=tensor([0.0296, 0.0419, 0.0331, 0.0327, 0.0393, 0.0314, 0.0863, 0.0477], device='cuda:2'), in_proj_covar=tensor([0.0436, 0.0493, 0.0477, 0.0437, 0.0523, 0.0500, 0.0580, 0.0405], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 17:58:00,045 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8626, 5.1839, 5.3439, 5.0895, 5.0967, 5.7027, 5.1823, 4.8594], device='cuda:2'), covar=tensor([0.0979, 0.1682, 0.1929, 0.1804, 0.2358, 0.0817, 0.1396, 0.2129], device='cuda:2'), in_proj_covar=tensor([0.0430, 0.0638, 0.0704, 0.0517, 0.0693, 0.0727, 0.0547, 0.0689], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:58:31,326 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0655, 5.1424, 4.9820, 4.5624, 4.5675, 5.0380, 4.9250, 4.7697], device='cuda:2'), covar=tensor([0.0635, 0.0517, 0.0294, 0.0323, 0.1099, 0.0573, 0.0313, 0.0573], device='cuda:2'), in_proj_covar=tensor([0.0316, 0.0479, 0.0371, 0.0373, 0.0367, 0.0428, 0.0254, 0.0443], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 17:58:41,593 INFO [train.py:904] (2/8) Epoch 29, batch 4750, loss[loss=0.155, simple_loss=0.2389, pruned_loss=0.03557, over 16571.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2735, pruned_loss=0.04895, over 3207598.16 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:58:52,603 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6245, 3.7953, 2.7558, 2.3125, 2.4673, 2.5300, 4.0831, 3.1754], device='cuda:2'), covar=tensor([0.2883, 0.0629, 0.2047, 0.2947, 0.2811, 0.2175, 0.0452, 0.1576], device='cuda:2'), in_proj_covar=tensor([0.0334, 0.0274, 0.0313, 0.0327, 0.0307, 0.0278, 0.0305, 0.0354], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 17:58:57,733 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288965.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:59:23,558 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:35,272 INFO [optim.py:368] (2/8) 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 17:59:50,955 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289001.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:54,498 INFO [train.py:904] (2/8) Epoch 29, batch 4800, loss[loss=0.1998, simple_loss=0.2818, pruned_loss=0.05895, over 12143.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2697, pruned_loss=0.04686, over 3204634.33 frames. ], batch size: 248, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:00:13,072 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0711, 2.4719, 2.5595, 1.9142, 2.7672, 2.8109, 2.4540, 2.3522], device='cuda:2'), covar=tensor([0.0739, 0.0281, 0.0268, 0.1005, 0.0134, 0.0267, 0.0487, 0.0480], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0088, 0.0133, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 18:00:37,056 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:00,408 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289046.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:01:05,206 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:05,808 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 18:01:08,914 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:12,124 INFO [train.py:904] (2/8) Epoch 29, batch 4850, loss[loss=0.1844, simple_loss=0.2633, pruned_loss=0.05274, over 17003.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2694, pruned_loss=0.04551, over 3192383.88 frames. ], batch size: 53, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:08,678 INFO [optim.py:368] (2/8) 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 18:02:22,360 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:02:28,898 INFO [train.py:904] (2/8) Epoch 29, batch 4900, loss[loss=0.1815, simple_loss=0.2744, pruned_loss=0.04427, over 16218.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2686, pruned_loss=0.04391, over 3186148.32 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:33,578 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289107.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:03:16,436 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289136.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:24,504 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1738, 2.5319, 2.0804, 2.2713, 2.8770, 2.4960, 2.6816, 3.0115], device='cuda:2'), covar=tensor([0.0193, 0.0518, 0.0678, 0.0572, 0.0325, 0.0503, 0.0290, 0.0327], device='cuda:2'), in_proj_covar=tensor([0.0236, 0.0247, 0.0236, 0.0237, 0.0248, 0.0246, 0.0244, 0.0247], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:03:42,180 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:43,041 INFO [train.py:904] (2/8) Epoch 29, batch 4950, loss[loss=0.1799, simple_loss=0.2698, pruned_loss=0.04503, over 17050.00 frames. ], tot_loss[loss=0.177, simple_loss=0.268, pruned_loss=0.04303, over 3188750.35 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:38,081 INFO [optim.py:368] (2/8) 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 18:04:46,906 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289197.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:04:56,967 INFO [train.py:904] (2/8) Epoch 29, batch 5000, loss[loss=0.1835, simple_loss=0.2798, pruned_loss=0.04355, over 16317.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2701, pruned_loss=0.043, over 3207616.14 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:59,456 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4368, 4.6840, 4.5912, 2.8661, 3.9346, 4.5592, 3.8954, 2.3503], device='cuda:2'), covar=tensor([0.0657, 0.0043, 0.0041, 0.0477, 0.0113, 0.0080, 0.0120, 0.0519], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0136, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 18:05:29,471 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 18:06:09,972 INFO [train.py:904] (2/8) Epoch 29, batch 5050, loss[loss=0.198, simple_loss=0.2873, pruned_loss=0.05429, over 15329.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2708, pruned_loss=0.04337, over 3205661.94 frames. ], batch size: 190, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:06:26,926 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289265.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:07:03,091 INFO [optim.py:368] (2/8) 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 18:07:22,109 INFO [train.py:904] (2/8) Epoch 29, batch 5100, loss[loss=0.1728, simple_loss=0.2601, pruned_loss=0.04277, over 16729.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2688, pruned_loss=0.04261, over 3218548.52 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:07:34,856 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:07:57,043 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1701, 5.2065, 5.0034, 4.5683, 4.6242, 5.0816, 4.9783, 4.7516], device='cuda:2'), covar=tensor([0.0617, 0.0395, 0.0316, 0.0328, 0.1143, 0.0539, 0.0297, 0.0640], device='cuda:2'), in_proj_covar=tensor([0.0313, 0.0475, 0.0368, 0.0369, 0.0365, 0.0426, 0.0253, 0.0440], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:07:57,271 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 18:08:36,372 INFO [train.py:904] (2/8) Epoch 29, batch 5150, loss[loss=0.1865, simple_loss=0.2838, pruned_loss=0.04464, over 16694.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.269, pruned_loss=0.04234, over 3193845.94 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:08:43,880 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1860, 4.2520, 4.5390, 4.4938, 4.5130, 4.2725, 4.2402, 4.2177], device='cuda:2'), covar=tensor([0.0350, 0.0631, 0.0362, 0.0400, 0.0458, 0.0390, 0.0872, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0436, 0.0493, 0.0477, 0.0437, 0.0523, 0.0501, 0.0580, 0.0405], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 18:09:29,038 INFO [optim.py:368] (2/8) 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 18:09:45,672 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289402.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:09:47,652 INFO [train.py:904] (2/8) Epoch 29, batch 5200, loss[loss=0.1501, simple_loss=0.2445, pruned_loss=0.0278, over 16837.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2676, pruned_loss=0.04189, over 3190819.26 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:10:19,775 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-02 18:10:38,471 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1121, 3.2533, 3.5552, 2.0164, 3.0583, 2.3978, 3.5011, 3.5226], device='cuda:2'), covar=tensor([0.0226, 0.0919, 0.0608, 0.2163, 0.0847, 0.0957, 0.0597, 0.0862], device='cuda:2'), in_proj_covar=tensor([0.0163, 0.0173, 0.0173, 0.0158, 0.0149, 0.0134, 0.0147, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 18:10:59,560 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 29, batch 5250, loss[loss=0.165, simple_loss=0.2619, pruned_loss=0.03404, over 15454.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2659, pruned_loss=0.04193, over 3192470.77 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:11:55,944 INFO [optim.py:368] (2/8) 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 18:11:58,153 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:10,717 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:14,373 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289503.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:15,680 INFO [train.py:904] (2/8) Epoch 29, batch 5300, loss[loss=0.1642, simple_loss=0.2487, pruned_loss=0.03986, over 16732.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2623, pruned_loss=0.04052, over 3197891.42 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:13:16,824 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 18:13:23,548 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8621, 2.1683, 2.3886, 3.1096, 2.1861, 2.3675, 2.3287, 2.2994], device='cuda:2'), covar=tensor([0.1580, 0.3713, 0.2910, 0.0803, 0.4334, 0.2496, 0.3564, 0.3416], device='cuda:2'), in_proj_covar=tensor([0.0423, 0.0478, 0.0387, 0.0339, 0.0446, 0.0550, 0.0450, 0.0559], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:13:26,898 INFO [train.py:904] (2/8) Epoch 29, batch 5350, loss[loss=0.197, simple_loss=0.2856, pruned_loss=0.05421, over 16408.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2602, pruned_loss=0.0396, over 3204967.54 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:13:42,081 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:13:53,793 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 18:14:21,420 INFO [optim.py:368] (2/8) 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 18:14:41,093 INFO [train.py:904] (2/8) Epoch 29, batch 5400, loss[loss=0.2093, simple_loss=0.2936, pruned_loss=0.0625, over 11662.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2627, pruned_loss=0.04024, over 3210348.60 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:15:34,739 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-05-02 18:15:42,189 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4265, 3.5312, 3.6704, 3.6468, 3.6581, 3.5119, 3.5315, 3.5473], device='cuda:2'), covar=tensor([0.0387, 0.0706, 0.0495, 0.0430, 0.0531, 0.0487, 0.0747, 0.0538], device='cuda:2'), in_proj_covar=tensor([0.0437, 0.0494, 0.0479, 0.0438, 0.0525, 0.0502, 0.0582, 0.0404], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 18:15:57,431 INFO [train.py:904] (2/8) Epoch 29, batch 5450, loss[loss=0.2145, simple_loss=0.3028, pruned_loss=0.0631, over 16168.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2651, pruned_loss=0.04139, over 3180296.02 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:16:07,381 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7164, 2.4987, 2.4246, 3.6990, 2.4076, 3.7599, 1.5280, 2.7733], device='cuda:2'), covar=tensor([0.1419, 0.0868, 0.1262, 0.0213, 0.0150, 0.0410, 0.1792, 0.0830], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0181, 0.0200, 0.0205, 0.0206, 0.0218, 0.0211, 0.0200], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 18:16:47,915 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3441, 5.3400, 5.0994, 4.3805, 5.2709, 1.8621, 4.9762, 4.8908], device='cuda:2'), covar=tensor([0.0120, 0.0130, 0.0232, 0.0511, 0.0119, 0.3092, 0.0141, 0.0259], device='cuda:2'), in_proj_covar=tensor([0.0185, 0.0179, 0.0218, 0.0190, 0.0195, 0.0222, 0.0206, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:16:53,639 INFO [optim.py:368] (2/8) 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 18:16:58,462 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0649, 2.4042, 2.5824, 1.9570, 2.7200, 2.7774, 2.3781, 2.3794], device='cuda:2'), covar=tensor([0.0745, 0.0299, 0.0248, 0.0959, 0.0136, 0.0306, 0.0499, 0.0476], device='cuda:2'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0139, 0.0087, 0.0131, 0.0130, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 18:17:11,478 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289702.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:17:14,657 INFO [train.py:904] (2/8) Epoch 29, batch 5500, loss[loss=0.2093, simple_loss=0.2992, pruned_loss=0.05968, over 16484.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2721, pruned_loss=0.04554, over 3138238.56 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:17:31,309 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5700, 4.8479, 4.6369, 4.6557, 4.3875, 4.3590, 4.3025, 4.9166], device='cuda:2'), covar=tensor([0.1312, 0.0884, 0.1038, 0.0931, 0.0871, 0.1472, 0.1227, 0.0929], device='cuda:2'), in_proj_covar=tensor([0.0721, 0.0866, 0.0714, 0.0671, 0.0553, 0.0550, 0.0726, 0.0682], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:18:26,454 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289750.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:18:31,713 INFO [train.py:904] (2/8) Epoch 29, batch 5550, loss[loss=0.2649, simple_loss=0.3246, pruned_loss=0.1026, over 11437.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2788, pruned_loss=0.04996, over 3116698.57 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:18:36,578 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 18:19:30,758 INFO [optim.py:368] (2/8) 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 18:19:33,602 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:19:46,078 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 18:19:53,194 INFO [train.py:904] (2/8) Epoch 29, batch 5600, loss[loss=0.2022, simple_loss=0.2912, pruned_loss=0.05655, over 17115.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2836, pruned_loss=0.05382, over 3091129.88 frames. ], batch size: 47, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:20:53,921 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289840.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:20:54,030 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6745, 4.7061, 5.0289, 4.9904, 5.0409, 4.7417, 4.7198, 4.5660], device='cuda:2'), covar=tensor([0.0332, 0.0587, 0.0416, 0.0412, 0.0445, 0.0383, 0.0914, 0.0523], device='cuda:2'), in_proj_covar=tensor([0.0439, 0.0497, 0.0482, 0.0441, 0.0527, 0.0504, 0.0585, 0.0407], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 18:21:16,880 INFO [train.py:904] (2/8) Epoch 29, batch 5650, loss[loss=0.25, simple_loss=0.3196, pruned_loss=0.09023, over 11158.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.288, pruned_loss=0.05731, over 3064631.95 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:21:25,653 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289859.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:22:15,728 INFO [optim.py:368] (2/8) 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 18:22:25,584 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0285, 3.2203, 3.5399, 2.0836, 3.0518, 2.2555, 3.4846, 3.6179], device='cuda:2'), covar=tensor([0.0247, 0.0929, 0.0571, 0.2233, 0.0858, 0.1094, 0.0673, 0.1069], device='cuda:2'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 18:22:35,917 INFO [train.py:904] (2/8) Epoch 29, batch 5700, loss[loss=0.2742, simple_loss=0.3289, pruned_loss=0.1098, over 11381.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2894, pruned_loss=0.05907, over 3044337.79 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:23:01,301 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289920.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:23:32,449 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3372, 3.3003, 3.3605, 3.4367, 3.4783, 3.2749, 3.4537, 3.5230], device='cuda:2'), covar=tensor([0.1264, 0.0937, 0.1040, 0.0669, 0.0737, 0.2468, 0.1151, 0.0881], device='cuda:2'), in_proj_covar=tensor([0.0686, 0.0836, 0.0964, 0.0851, 0.0648, 0.0671, 0.0707, 0.0823], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:23:55,710 INFO [train.py:904] (2/8) Epoch 29, batch 5750, loss[loss=0.1873, simple_loss=0.2843, pruned_loss=0.04514, over 16889.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2915, pruned_loss=0.06045, over 3011762.33 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:24:40,822 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289981.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:24:56,848 INFO [optim.py:368] (2/8) 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 18:25:08,727 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289998.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:25:20,974 INFO [train.py:904] (2/8) Epoch 29, batch 5800, loss[loss=0.1989, simple_loss=0.2946, pruned_loss=0.05158, over 16697.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.29, pruned_loss=0.05781, over 3042801.18 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:25:53,547 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5049, 3.4658, 3.4356, 2.6524, 3.2921, 2.1437, 3.1282, 2.7270], device='cuda:2'), covar=tensor([0.0207, 0.0176, 0.0224, 0.0262, 0.0136, 0.2528, 0.0161, 0.0325], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0178, 0.0217, 0.0188, 0.0194, 0.0221, 0.0204, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:26:20,299 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5475, 4.5419, 4.4286, 3.6613, 4.4897, 1.8566, 4.2610, 4.0864], device='cuda:2'), covar=tensor([0.0127, 0.0109, 0.0215, 0.0381, 0.0121, 0.2963, 0.0143, 0.0296], device='cuda:2'), in_proj_covar=tensor([0.0184, 0.0178, 0.0217, 0.0188, 0.0193, 0.0220, 0.0204, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:26:34,244 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290050.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:26:39,145 INFO [train.py:904] (2/8) Epoch 29, batch 5850, loss[loss=0.1883, simple_loss=0.2782, pruned_loss=0.04917, over 16641.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2882, pruned_loss=0.05663, over 3038685.88 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:26:48,245 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290059.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:27:37,751 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 18:27:39,837 INFO [optim.py:368] (2/8) 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 18:28:01,356 INFO [train.py:904] (2/8) Epoch 29, batch 5900, loss[loss=0.2401, simple_loss=0.3054, pruned_loss=0.08742, over 11454.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2879, pruned_loss=0.05665, over 3060414.88 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:28:15,106 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:28:47,495 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9363, 3.2134, 3.3545, 2.0975, 3.0613, 3.3492, 3.2427, 1.8695], device='cuda:2'), covar=tensor([0.0643, 0.0117, 0.0090, 0.0518, 0.0140, 0.0140, 0.0114, 0.0580], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 18:29:20,965 INFO [train.py:904] (2/8) Epoch 29, batch 5950, loss[loss=0.2119, simple_loss=0.3036, pruned_loss=0.06011, over 16223.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2887, pruned_loss=0.05534, over 3075975.80 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:29:30,202 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:30:18,749 INFO [optim.py:368] (2/8) 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 18:30:32,082 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 18:30:40,226 INFO [train.py:904] (2/8) Epoch 29, batch 6000, loss[loss=0.1642, simple_loss=0.2589, pruned_loss=0.03474, over 17012.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2882, pruned_loss=0.05477, over 3091154.86 frames. ], batch size: 50, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:30:40,226 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 18:30:50,243 INFO [train.py:938] (2/8) Epoch 29, validation: loss=0.1475, simple_loss=0.2594, pruned_loss=0.01778, over 944034.00 frames. 2023-05-02 18:30:50,244 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 18:30:55,860 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:32:11,209 INFO [train.py:904] (2/8) Epoch 29, batch 6050, loss[loss=0.1896, simple_loss=0.29, pruned_loss=0.04453, over 16265.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.287, pruned_loss=0.05473, over 3098936.59 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:32:15,006 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1047, 5.1868, 4.9221, 4.5462, 4.5851, 5.0301, 4.9230, 4.7318], device='cuda:2'), covar=tensor([0.0957, 0.1062, 0.0505, 0.0580, 0.1268, 0.0887, 0.0635, 0.1060], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0471, 0.0365, 0.0366, 0.0361, 0.0422, 0.0251, 0.0437], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:32:42,720 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290276.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:33:04,448 INFO [optim.py:368] (2/8) 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 18:33:29,590 INFO [train.py:904] (2/8) Epoch 29, batch 6100, loss[loss=0.2325, simple_loss=0.3022, pruned_loss=0.0814, over 11278.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.287, pruned_loss=0.05437, over 3093137.31 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:33:53,042 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:34:46,605 INFO [train.py:904] (2/8) Epoch 29, batch 6150, loss[loss=0.1731, simple_loss=0.2675, pruned_loss=0.03935, over 16786.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2861, pruned_loss=0.05443, over 3093132.07 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:34:47,113 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:27,398 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290379.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:44,980 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 29, batch 6200, loss[loss=0.1738, simple_loss=0.2579, pruned_loss=0.04482, over 17034.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2847, pruned_loss=0.05424, over 3100592.67 frames. ], batch size: 55, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:36:08,269 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:36:37,884 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 18:37:21,110 INFO [train.py:904] (2/8) Epoch 29, batch 6250, loss[loss=0.1817, simple_loss=0.2832, pruned_loss=0.04012, over 16869.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2836, pruned_loss=0.05364, over 3098745.73 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:38:06,760 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1775, 3.4390, 3.6130, 2.1355, 3.1560, 2.4525, 3.6609, 3.7857], device='cuda:2'), covar=tensor([0.0252, 0.0817, 0.0597, 0.2149, 0.0839, 0.0981, 0.0552, 0.0903], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0172, 0.0171, 0.0157, 0.0148, 0.0133, 0.0145, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 18:38:15,680 INFO [optim.py:368] (2/8) 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 18:38:28,955 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.62 vs. limit=5.0 2023-05-02 18:38:34,373 INFO [train.py:904] (2/8) Epoch 29, batch 6300, loss[loss=0.1695, simple_loss=0.2654, pruned_loss=0.03686, over 16844.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2834, pruned_loss=0.05294, over 3099397.73 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:39:29,529 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6127, 2.8234, 2.8067, 4.8543, 3.8145, 4.1584, 1.5757, 3.2064], device='cuda:2'), covar=tensor([0.1562, 0.0848, 0.1274, 0.0176, 0.0354, 0.0470, 0.1863, 0.0862], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0182, 0.0202, 0.0206, 0.0208, 0.0220, 0.0212, 0.0201], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 18:39:52,799 INFO [train.py:904] (2/8) Epoch 29, batch 6350, loss[loss=0.1867, simple_loss=0.2707, pruned_loss=0.05129, over 16534.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2842, pruned_loss=0.05405, over 3100388.42 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:40:19,062 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 18:40:28,172 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290576.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:40:50,035 INFO [optim.py:368] (2/8) 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 18:40:55,011 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 18:41:09,303 INFO [train.py:904] (2/8) Epoch 29, batch 6400, loss[loss=0.2538, simple_loss=0.3227, pruned_loss=0.09241, over 11098.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2857, pruned_loss=0.05601, over 3072921.98 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:41:39,261 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290624.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:42:23,290 INFO [train.py:904] (2/8) Epoch 29, batch 6450, loss[loss=0.1837, simple_loss=0.2737, pruned_loss=0.04687, over 16705.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2854, pruned_loss=0.05538, over 3069985.63 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:42:23,721 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:42:53,379 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:16,582 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4124, 3.6036, 3.3076, 3.0932, 3.0509, 3.5117, 3.2403, 3.3374], device='cuda:2'), covar=tensor([0.0848, 0.0747, 0.0466, 0.0398, 0.0956, 0.0607, 0.2026, 0.0624], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0473, 0.0365, 0.0366, 0.0362, 0.0422, 0.0251, 0.0438], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:43:19,925 INFO [optim.py:368] (2/8) 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 18:43:37,903 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:39,653 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 18:43:40,039 INFO [train.py:904] (2/8) Epoch 29, batch 6500, loss[loss=0.1986, simple_loss=0.2906, pruned_loss=0.05333, over 16816.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2834, pruned_loss=0.05459, over 3073380.78 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:43:44,299 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290706.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:44:59,497 INFO [train.py:904] (2/8) Epoch 29, batch 6550, loss[loss=0.1957, simple_loss=0.2974, pruned_loss=0.04699, over 16666.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2858, pruned_loss=0.05514, over 3091786.85 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:44:59,828 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:45:17,918 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-05-02 18:45:58,154 INFO [optim.py:368] (2/8) 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 18:46:19,466 INFO [train.py:904] (2/8) Epoch 29, batch 6600, loss[loss=0.2144, simple_loss=0.2948, pruned_loss=0.06695, over 11624.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.288, pruned_loss=0.05552, over 3087739.14 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:47:38,810 INFO [train.py:904] (2/8) Epoch 29, batch 6650, loss[loss=0.2042, simple_loss=0.2924, pruned_loss=0.05798, over 16348.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2884, pruned_loss=0.05655, over 3069437.89 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:48:04,874 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 18:48:10,445 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 18:48:35,967 INFO [optim.py:368] (2/8) 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 18:48:55,747 INFO [train.py:904] (2/8) Epoch 29, batch 6700, loss[loss=0.225, simple_loss=0.308, pruned_loss=0.07095, over 15437.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2869, pruned_loss=0.05673, over 3071686.23 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:49:08,359 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-02 18:49:25,431 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-02 18:49:48,205 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5511, 3.6053, 3.3750, 3.0390, 3.2558, 3.5176, 3.3123, 3.3571], device='cuda:2'), covar=tensor([0.0541, 0.0626, 0.0278, 0.0266, 0.0478, 0.0437, 0.1482, 0.0471], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0473, 0.0365, 0.0366, 0.0362, 0.0422, 0.0251, 0.0439], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:49:51,076 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7823, 3.8704, 3.9497, 3.7157, 3.9023, 4.2579, 3.9213, 3.6447], device='cuda:2'), covar=tensor([0.2325, 0.2337, 0.2847, 0.2496, 0.2541, 0.1751, 0.1839, 0.2836], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0644, 0.0714, 0.0520, 0.0698, 0.0732, 0.0554, 0.0697], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 18:50:12,971 INFO [train.py:904] (2/8) Epoch 29, batch 6750, loss[loss=0.1833, simple_loss=0.2709, pruned_loss=0.04784, over 16448.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2865, pruned_loss=0.05706, over 3066717.46 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:43,370 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290974.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:51:10,928 INFO [optim.py:368] (2/8) 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 18:51:12,166 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1188, 2.3584, 1.9990, 2.1660, 2.7409, 2.4042, 2.6626, 2.9328], device='cuda:2'), covar=tensor([0.0204, 0.0533, 0.0649, 0.0574, 0.0315, 0.0467, 0.0240, 0.0310], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0245, 0.0235, 0.0235, 0.0245, 0.0243, 0.0240, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:51:28,424 INFO [train.py:904] (2/8) Epoch 29, batch 6800, loss[loss=0.207, simple_loss=0.2996, pruned_loss=0.05718, over 16475.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2867, pruned_loss=0.0567, over 3088941.42 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:51:57,165 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291022.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:52:46,127 INFO [train.py:904] (2/8) Epoch 29, batch 6850, loss[loss=0.2416, simple_loss=0.3047, pruned_loss=0.08928, over 11080.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2881, pruned_loss=0.05727, over 3087141.20 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:53:09,404 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291069.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:53:42,653 INFO [optim.py:368] (2/8) 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 18:54:02,729 INFO [train.py:904] (2/8) Epoch 29, batch 6900, loss[loss=0.1891, simple_loss=0.2896, pruned_loss=0.04425, over 16885.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.29, pruned_loss=0.05657, over 3080127.56 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:54:41,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8630, 3.7985, 3.9105, 4.0271, 4.0951, 3.7280, 4.0558, 4.1162], device='cuda:2'), covar=tensor([0.1633, 0.1094, 0.1252, 0.0653, 0.0661, 0.1689, 0.0894, 0.0778], device='cuda:2'), in_proj_covar=tensor([0.0679, 0.0830, 0.0955, 0.0842, 0.0641, 0.0662, 0.0703, 0.0817], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:54:44,576 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 29, batch 6950, loss[loss=0.207, simple_loss=0.2939, pruned_loss=0.06003, over 15425.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2913, pruned_loss=0.05804, over 3068610.35 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:18,587 INFO [optim.py:368] (2/8) 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 18:56:25,436 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8584, 2.2143, 2.5210, 3.1693, 2.2252, 2.3976, 2.3663, 2.2967], device='cuda:2'), covar=tensor([0.1461, 0.3174, 0.2553, 0.0730, 0.4192, 0.2448, 0.3214, 0.3481], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0474, 0.0386, 0.0336, 0.0446, 0.0545, 0.0447, 0.0556], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:56:36,625 INFO [train.py:904] (2/8) Epoch 29, batch 7000, loss[loss=0.1775, simple_loss=0.2846, pruned_loss=0.03517, over 16807.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2911, pruned_loss=0.05784, over 3046504.82 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:57:28,520 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4321, 5.4595, 5.2711, 4.9297, 4.8877, 5.3110, 5.2287, 4.9873], device='cuda:2'), covar=tensor([0.0672, 0.0575, 0.0339, 0.0312, 0.1152, 0.0571, 0.0328, 0.0691], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0472, 0.0364, 0.0365, 0.0361, 0.0421, 0.0250, 0.0437], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 18:57:47,298 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6419, 2.5564, 2.4332, 4.0683, 2.8397, 3.9222, 1.5553, 2.8227], device='cuda:2'), covar=tensor([0.1469, 0.0898, 0.1406, 0.0235, 0.0252, 0.0436, 0.1792, 0.0919], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0182, 0.0203, 0.0207, 0.0209, 0.0221, 0.0213, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 18:57:50,284 INFO [train.py:904] (2/8) Epoch 29, batch 7050, loss[loss=0.2569, simple_loss=0.3167, pruned_loss=0.09856, over 11209.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2914, pruned_loss=0.05735, over 3058650.08 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:58:49,441 INFO [optim.py:368] (2/8) 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 18:59:07,335 INFO [train.py:904] (2/8) Epoch 29, batch 7100, loss[loss=0.2135, simple_loss=0.2738, pruned_loss=0.07663, over 11467.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2893, pruned_loss=0.05678, over 3068951.14 frames. ], batch size: 249, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:59:10,925 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 18:59:35,485 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 18:59:50,571 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.97 vs. limit=5.0 2023-05-02 19:00:25,886 INFO [train.py:904] (2/8) Epoch 29, batch 7150, loss[loss=0.1976, simple_loss=0.285, pruned_loss=0.05511, over 16892.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.288, pruned_loss=0.05708, over 3064738.65 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:01:23,565 INFO [optim.py:368] (2/8) 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 19:01:41,848 INFO [train.py:904] (2/8) Epoch 29, batch 7200, loss[loss=0.1712, simple_loss=0.2668, pruned_loss=0.03778, over 16796.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2858, pruned_loss=0.0556, over 3059551.89 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:02:14,455 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=291425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:02:42,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6252, 3.6933, 2.4471, 4.3967, 2.9456, 4.2654, 2.5108, 2.9698], device='cuda:2'), covar=tensor([0.0372, 0.0456, 0.1672, 0.0193, 0.0863, 0.0585, 0.1622, 0.0880], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0181, 0.0197, 0.0174, 0.0181, 0.0222, 0.0205, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 19:02:43,248 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6405, 4.9159, 5.0506, 4.8207, 4.9049, 5.3729, 4.8981, 4.6667], device='cuda:2'), covar=tensor([0.1300, 0.1796, 0.2282, 0.1857, 0.2248, 0.1050, 0.1639, 0.2420], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0642, 0.0711, 0.0519, 0.0694, 0.0730, 0.0551, 0.0695], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 19:03:00,075 INFO [train.py:904] (2/8) Epoch 29, batch 7250, loss[loss=0.1713, simple_loss=0.2575, pruned_loss=0.04261, over 15219.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2839, pruned_loss=0.05426, over 3074664.62 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:03:58,886 INFO [optim.py:368] (2/8) 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 19:04:16,115 INFO [train.py:904] (2/8) Epoch 29, batch 7300, loss[loss=0.2322, simple_loss=0.3129, pruned_loss=0.07569, over 15482.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2835, pruned_loss=0.05405, over 3069644.00 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:05:32,027 INFO [train.py:904] (2/8) Epoch 29, batch 7350, loss[loss=0.2065, simple_loss=0.2952, pruned_loss=0.05884, over 16463.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2848, pruned_loss=0.05511, over 3057449.11 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:32,137 INFO [optim.py:368] (2/8) 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 19:06:49,574 INFO [train.py:904] (2/8) Epoch 29, batch 7400, loss[loss=0.2045, simple_loss=0.2833, pruned_loss=0.06287, over 16302.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2855, pruned_loss=0.05523, over 3090594.49 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:51,483 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8669, 4.6555, 4.8847, 5.0617, 5.2534, 4.7062, 5.2313, 5.2547], device='cuda:2'), covar=tensor([0.1918, 0.1439, 0.1818, 0.0778, 0.0645, 0.0991, 0.0730, 0.0696], device='cuda:2'), in_proj_covar=tensor([0.0674, 0.0825, 0.0947, 0.0835, 0.0639, 0.0658, 0.0700, 0.0811], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:07:12,796 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4934, 3.5396, 3.2976, 2.9347, 3.1933, 3.4373, 3.3081, 3.2753], device='cuda:2'), covar=tensor([0.0619, 0.0776, 0.0298, 0.0307, 0.0475, 0.0521, 0.1257, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0468, 0.0361, 0.0362, 0.0358, 0.0416, 0.0248, 0.0434], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:08:07,282 INFO [train.py:904] (2/8) Epoch 29, batch 7450, loss[loss=0.1858, simple_loss=0.2718, pruned_loss=0.04994, over 17049.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2871, pruned_loss=0.05624, over 3094260.09 frames. ], batch size: 55, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:09:10,902 INFO [optim.py:368] (2/8) 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 19:09:25,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7275, 1.7307, 2.2957, 2.6781, 2.6440, 2.9823, 1.7353, 2.9844], device='cuda:2'), covar=tensor([0.0235, 0.0681, 0.0349, 0.0320, 0.0346, 0.0217, 0.0862, 0.0165], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0197, 0.0186, 0.0191, 0.0209, 0.0166, 0.0204, 0.0168], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:09:28,110 INFO [train.py:904] (2/8) Epoch 29, batch 7500, loss[loss=0.165, simple_loss=0.254, pruned_loss=0.03805, over 17010.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2866, pruned_loss=0.05506, over 3109022.33 frames. ], batch size: 53, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:09:57,726 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-02 19:10:01,545 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=291725.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:10:45,759 INFO [train.py:904] (2/8) Epoch 29, batch 7550, loss[loss=0.2224, simple_loss=0.2888, pruned_loss=0.07804, over 11461.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2854, pruned_loss=0.05517, over 3113839.31 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:11:04,494 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 19:11:15,345 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:19,748 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:44,714 INFO [optim.py:368] (2/8) 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 19:12:01,407 INFO [train.py:904] (2/8) Epoch 29, batch 7600, loss[loss=0.181, simple_loss=0.2715, pruned_loss=0.0453, over 16983.00 frames. ], tot_loss[loss=0.198, simple_loss=0.285, pruned_loss=0.05552, over 3098023.71 frames. ], batch size: 41, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:12:52,099 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291837.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:13:18,970 INFO [train.py:904] (2/8) Epoch 29, batch 7650, loss[loss=0.1799, simple_loss=0.2706, pruned_loss=0.04459, over 17108.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.286, pruned_loss=0.05639, over 3085734.15 frames. ], batch size: 47, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:13:27,128 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3568, 3.2759, 2.6760, 2.2117, 2.2442, 2.3228, 3.3692, 2.9869], device='cuda:2'), covar=tensor([0.3217, 0.0780, 0.1980, 0.2965, 0.2821, 0.2397, 0.0586, 0.1470], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0276, 0.0316, 0.0330, 0.0308, 0.0279, 0.0306, 0.0354], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 19:14:20,910 INFO [optim.py:368] (2/8) 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 19:14:21,336 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2055, 5.5404, 5.2754, 5.2837, 4.9836, 4.9602, 4.9256, 5.6438], device='cuda:2'), covar=tensor([0.1370, 0.0872, 0.1002, 0.0990, 0.0834, 0.0991, 0.1248, 0.0899], device='cuda:2'), in_proj_covar=tensor([0.0722, 0.0866, 0.0714, 0.0675, 0.0553, 0.0556, 0.0726, 0.0682], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:14:36,029 INFO [train.py:904] (2/8) Epoch 29, batch 7700, loss[loss=0.186, simple_loss=0.2727, pruned_loss=0.04964, over 16802.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2859, pruned_loss=0.05665, over 3088412.77 frames. ], batch size: 39, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:15:22,714 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0529, 5.2895, 4.8979, 4.6451, 4.1742, 5.1812, 5.1026, 4.7499], device='cuda:2'), covar=tensor([0.0962, 0.0831, 0.0588, 0.0494, 0.1974, 0.0629, 0.0436, 0.0961], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0468, 0.0361, 0.0362, 0.0358, 0.0416, 0.0249, 0.0433], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:15:44,900 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9579, 5.3774, 5.5705, 5.2970, 5.3779, 5.9069, 5.4297, 5.2188], device='cuda:2'), covar=tensor([0.1009, 0.1838, 0.2169, 0.1703, 0.2060, 0.0834, 0.1470, 0.2006], device='cuda:2'), in_proj_covar=tensor([0.0430, 0.0641, 0.0708, 0.0520, 0.0694, 0.0729, 0.0551, 0.0695], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 19:15:53,492 INFO [train.py:904] (2/8) Epoch 29, batch 7750, loss[loss=0.2066, simple_loss=0.2972, pruned_loss=0.05802, over 16734.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.286, pruned_loss=0.05625, over 3103369.77 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:16:55,756 INFO [optim.py:368] (2/8) 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 19:17:14,568 INFO [train.py:904] (2/8) Epoch 29, batch 7800, loss[loss=0.1978, simple_loss=0.2896, pruned_loss=0.05297, over 16689.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.287, pruned_loss=0.05703, over 3095520.15 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:17:27,065 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3182, 4.3222, 4.2391, 3.3452, 4.2697, 1.8205, 4.0465, 3.9002], device='cuda:2'), covar=tensor([0.0173, 0.0169, 0.0229, 0.0413, 0.0132, 0.2878, 0.0186, 0.0309], device='cuda:2'), in_proj_covar=tensor([0.0182, 0.0175, 0.0215, 0.0186, 0.0190, 0.0218, 0.0201, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:18:24,777 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 19:18:27,413 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9019, 4.1767, 3.9937, 4.0473, 3.7227, 3.7959, 3.8247, 4.1566], device='cuda:2'), covar=tensor([0.1315, 0.1089, 0.1123, 0.0965, 0.0901, 0.1748, 0.1069, 0.1153], device='cuda:2'), in_proj_covar=tensor([0.0723, 0.0867, 0.0715, 0.0675, 0.0554, 0.0556, 0.0726, 0.0683], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:18:30,015 INFO [train.py:904] (2/8) Epoch 29, batch 7850, loss[loss=0.2393, simple_loss=0.3094, pruned_loss=0.08459, over 11155.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2873, pruned_loss=0.05618, over 3102204.46 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:18:30,526 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:19:30,094 INFO [optim.py:368] (2/8) 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 19:19:43,449 INFO [train.py:904] (2/8) Epoch 29, batch 7900, loss[loss=0.1997, simple_loss=0.2857, pruned_loss=0.05684, over 16633.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2863, pruned_loss=0.05568, over 3110148.62 frames. ], batch size: 57, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:19:59,932 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=292115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:20:26,269 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:20:49,295 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5739, 4.5512, 4.4390, 3.6703, 4.5012, 1.7884, 4.2464, 4.0742], device='cuda:2'), covar=tensor([0.0117, 0.0111, 0.0211, 0.0393, 0.0106, 0.2988, 0.0160, 0.0282], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0174, 0.0214, 0.0185, 0.0189, 0.0217, 0.0200, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:21:01,247 INFO [train.py:904] (2/8) Epoch 29, batch 7950, loss[loss=0.1961, simple_loss=0.2817, pruned_loss=0.05526, over 16898.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2859, pruned_loss=0.05573, over 3104257.59 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:21:46,360 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4931, 4.0934, 4.0684, 2.6602, 3.6528, 4.0999, 3.6412, 2.3399], device='cuda:2'), covar=tensor([0.0563, 0.0074, 0.0067, 0.0439, 0.0131, 0.0133, 0.0116, 0.0484], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0135, 0.0103, 0.0117, 0.0099, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 19:22:03,657 INFO [optim.py:368] (2/8) 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 19:22:18,473 INFO [train.py:904] (2/8) Epoch 29, batch 8000, loss[loss=0.1941, simple_loss=0.2908, pruned_loss=0.04874, over 16853.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2868, pruned_loss=0.05636, over 3101078.57 frames. ], batch size: 42, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:23:16,735 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0607, 2.3892, 2.3598, 2.9206, 1.8415, 3.1954, 1.8573, 2.7340], device='cuda:2'), covar=tensor([0.1121, 0.0669, 0.1071, 0.0222, 0.0116, 0.0380, 0.1511, 0.0707], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0183, 0.0203, 0.0206, 0.0209, 0.0220, 0.0212, 0.0201], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 19:23:31,106 INFO [train.py:904] (2/8) Epoch 29, batch 8050, loss[loss=0.2001, simple_loss=0.2933, pruned_loss=0.05349, over 16765.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2864, pruned_loss=0.05612, over 3108224.87 frames. ], batch size: 89, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:23:32,621 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5416, 3.6130, 3.3919, 3.0650, 3.2650, 3.5084, 3.3228, 3.3424], device='cuda:2'), covar=tensor([0.0649, 0.0833, 0.0334, 0.0300, 0.0505, 0.0580, 0.1341, 0.0543], device='cuda:2'), in_proj_covar=tensor([0.0310, 0.0470, 0.0362, 0.0363, 0.0359, 0.0418, 0.0250, 0.0435], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:23:35,410 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 19:24:32,478 INFO [optim.py:368] (2/8) 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 19:24:46,379 INFO [train.py:904] (2/8) Epoch 29, batch 8100, loss[loss=0.1714, simple_loss=0.2649, pruned_loss=0.03896, over 16830.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2853, pruned_loss=0.05485, over 3114087.37 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:25:40,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5639, 4.8619, 4.6722, 4.6713, 4.3853, 4.3277, 4.3014, 4.9286], device='cuda:2'), covar=tensor([0.1286, 0.0854, 0.1008, 0.0945, 0.0852, 0.1430, 0.1231, 0.0920], device='cuda:2'), in_proj_covar=tensor([0.0722, 0.0863, 0.0715, 0.0673, 0.0551, 0.0554, 0.0727, 0.0680], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:26:01,419 INFO [train.py:904] (2/8) Epoch 29, batch 8150, loss[loss=0.1908, simple_loss=0.2775, pruned_loss=0.05208, over 16638.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.283, pruned_loss=0.05416, over 3111844.78 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:26:27,169 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1239, 5.1252, 4.9595, 4.2373, 5.0272, 1.9165, 4.7708, 4.5953], device='cuda:2'), covar=tensor([0.0109, 0.0092, 0.0205, 0.0454, 0.0102, 0.2766, 0.0131, 0.0273], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0175, 0.0214, 0.0185, 0.0190, 0.0218, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:26:50,756 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6755, 2.4217, 2.2582, 3.4698, 2.3874, 3.6973, 1.4748, 2.6673], device='cuda:2'), covar=tensor([0.1491, 0.0930, 0.1467, 0.0186, 0.0173, 0.0408, 0.1983, 0.0962], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0183, 0.0203, 0.0206, 0.0210, 0.0220, 0.0213, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 19:26:55,562 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3014, 6.0150, 6.2144, 5.8258, 5.9429, 6.4360, 5.9686, 5.6902], device='cuda:2'), covar=tensor([0.0928, 0.1722, 0.2201, 0.1832, 0.1979, 0.0846, 0.1492, 0.2257], device='cuda:2'), in_proj_covar=tensor([0.0429, 0.0639, 0.0708, 0.0517, 0.0690, 0.0730, 0.0549, 0.0693], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 19:27:01,238 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 29, batch 8200, loss[loss=0.19, simple_loss=0.2728, pruned_loss=0.05359, over 17012.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2799, pruned_loss=0.05278, over 3130377.83 frames. ], batch size: 50, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:27:25,888 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292410.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:28:00,169 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292432.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:28:34,795 INFO [train.py:904] (2/8) Epoch 29, batch 8250, loss[loss=0.1765, simple_loss=0.2597, pruned_loss=0.0466, over 12140.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2788, pruned_loss=0.05055, over 3107817.68 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:29:17,989 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:29:30,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8639, 4.9333, 4.7196, 4.3429, 4.4192, 4.8273, 4.6912, 4.5076], device='cuda:2'), covar=tensor([0.0602, 0.0603, 0.0331, 0.0362, 0.0953, 0.0490, 0.0421, 0.0726], device='cuda:2'), in_proj_covar=tensor([0.0309, 0.0469, 0.0362, 0.0362, 0.0358, 0.0417, 0.0249, 0.0434], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:29:41,436 INFO [optim.py:368] (2/8) 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 19:29:55,813 INFO [train.py:904] (2/8) Epoch 29, batch 8300, loss[loss=0.16, simple_loss=0.2647, pruned_loss=0.02762, over 16891.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2763, pruned_loss=0.04798, over 3088819.36 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:29:59,604 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8356, 3.6976, 3.9004, 3.9986, 4.1038, 3.6954, 4.0505, 4.1445], device='cuda:2'), covar=tensor([0.1705, 0.1366, 0.1372, 0.0854, 0.0740, 0.1923, 0.0943, 0.0922], device='cuda:2'), in_proj_covar=tensor([0.0672, 0.0821, 0.0945, 0.0836, 0.0638, 0.0656, 0.0698, 0.0814], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:31:15,559 INFO [train.py:904] (2/8) Epoch 29, batch 8350, loss[loss=0.1779, simple_loss=0.2741, pruned_loss=0.04083, over 16888.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2756, pruned_loss=0.04582, over 3078706.65 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:32:16,316 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5696, 3.9402, 3.9211, 2.8355, 3.5056, 3.9937, 3.6603, 2.4157], device='cuda:2'), covar=tensor([0.0502, 0.0066, 0.0059, 0.0370, 0.0125, 0.0118, 0.0092, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0134, 0.0102, 0.0115, 0.0098, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 19:32:20,505 INFO [optim.py:368] (2/8) 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 19:32:36,081 INFO [train.py:904] (2/8) Epoch 29, batch 8400, loss[loss=0.1711, simple_loss=0.2665, pruned_loss=0.03779, over 16458.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2734, pruned_loss=0.04432, over 3057083.96 frames. ], batch size: 68, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:33:56,324 INFO [train.py:904] (2/8) Epoch 29, batch 8450, loss[loss=0.165, simple_loss=0.2627, pruned_loss=0.03368, over 16639.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2721, pruned_loss=0.043, over 3070633.30 frames. ], batch size: 57, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:33:59,322 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3386, 4.2904, 4.1276, 3.1989, 4.2157, 1.7913, 3.9382, 3.8019], device='cuda:2'), covar=tensor([0.0143, 0.0139, 0.0249, 0.0472, 0.0149, 0.3163, 0.0197, 0.0372], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0184, 0.0190, 0.0218, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 19:34:05,582 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5145, 3.5129, 2.8173, 2.1497, 2.2277, 2.4200, 3.6572, 3.1090], device='cuda:2'), covar=tensor([0.3143, 0.0616, 0.1812, 0.3359, 0.3110, 0.2401, 0.0443, 0.1448], device='cuda:2'), in_proj_covar=tensor([0.0333, 0.0272, 0.0310, 0.0325, 0.0304, 0.0276, 0.0302, 0.0347], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 19:34:08,531 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4158, 3.5070, 3.6870, 3.6679, 3.6778, 3.4992, 3.5318, 3.5819], device='cuda:2'), covar=tensor([0.0450, 0.0915, 0.0614, 0.0583, 0.0555, 0.0701, 0.0921, 0.0560], device='cuda:2'), in_proj_covar=tensor([0.0439, 0.0500, 0.0480, 0.0443, 0.0526, 0.0505, 0.0583, 0.0406], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 19:34:37,546 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 19:35:03,479 INFO [optim.py:368] (2/8) 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 19:35:19,269 INFO [train.py:904] (2/8) Epoch 29, batch 8500, loss[loss=0.16, simple_loss=0.2421, pruned_loss=0.03894, over 12121.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2686, pruned_loss=0.04105, over 3044112.89 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:35:28,949 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292710.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:36:42,994 INFO [train.py:904] (2/8) Epoch 29, batch 8550, loss[loss=0.1725, simple_loss=0.2554, pruned_loss=0.04486, over 11857.00 frames. ], tot_loss[loss=0.174, simple_loss=0.267, pruned_loss=0.04056, over 3013447.97 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:36:46,316 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 19:36:52,468 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292758.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:38:04,312 INFO [optim.py:368] (2/8) 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 19:38:21,503 INFO [train.py:904] (2/8) Epoch 29, batch 8600, loss[loss=0.1708, simple_loss=0.2726, pruned_loss=0.03449, over 15299.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2671, pruned_loss=0.03936, over 3027383.71 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:39:14,832 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 19:40:01,866 INFO [train.py:904] (2/8) Epoch 29, batch 8650, loss[loss=0.1467, simple_loss=0.2501, pruned_loss=0.02167, over 16270.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2652, pruned_loss=0.03782, over 3039246.93 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:40:33,526 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5280, 3.5238, 2.7785, 2.2281, 2.2343, 2.4601, 3.6403, 3.1617], device='cuda:2'), covar=tensor([0.3111, 0.0512, 0.1924, 0.3243, 0.3074, 0.2314, 0.0419, 0.1306], device='cuda:2'), in_proj_covar=tensor([0.0333, 0.0272, 0.0311, 0.0325, 0.0303, 0.0276, 0.0301, 0.0346], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 19:41:31,353 INFO [optim.py:368] (2/8) 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 19:41:48,477 INFO [train.py:904] (2/8) Epoch 29, batch 8700, loss[loss=0.1641, simple_loss=0.2587, pruned_loss=0.03472, over 16933.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2628, pruned_loss=0.03696, over 3033287.68 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:43:26,843 INFO [train.py:904] (2/8) Epoch 29, batch 8750, loss[loss=0.1913, simple_loss=0.2847, pruned_loss=0.04894, over 16722.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2627, pruned_loss=0.0364, over 3048745.07 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:44:01,915 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:45:00,944 INFO [optim.py:368] (2/8) 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 19:45:20,128 INFO [train.py:904] (2/8) Epoch 29, batch 8800, loss[loss=0.1925, simple_loss=0.2747, pruned_loss=0.05509, over 12691.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2612, pruned_loss=0.03546, over 3043978.88 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:46:11,590 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:46:19,864 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-05-02 19:47:05,839 INFO [train.py:904] (2/8) Epoch 29, batch 8850, loss[loss=0.15, simple_loss=0.2449, pruned_loss=0.02755, over 12368.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2637, pruned_loss=0.03513, over 3037956.67 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:47:35,085 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3900, 2.9814, 2.6849, 2.2650, 2.2094, 2.3264, 2.8745, 2.7770], device='cuda:2'), covar=tensor([0.2892, 0.0710, 0.1820, 0.3146, 0.2921, 0.2392, 0.0532, 0.1517], device='cuda:2'), in_proj_covar=tensor([0.0331, 0.0270, 0.0308, 0.0322, 0.0300, 0.0274, 0.0299, 0.0344], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 19:48:05,751 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:48:35,616 INFO [optim.py:368] (2/8) 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 19:48:55,720 INFO [train.py:904] (2/8) Epoch 29, batch 8900, loss[loss=0.1609, simple_loss=0.2554, pruned_loss=0.03317, over 13098.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2637, pruned_loss=0.0343, over 3030965.50 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:49:23,445 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293117.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:50:06,667 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0836, 3.2419, 3.5785, 1.9977, 3.0345, 2.2797, 3.5055, 3.4920], device='cuda:2'), covar=tensor([0.0255, 0.0982, 0.0548, 0.2290, 0.0808, 0.1083, 0.0626, 0.0970], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0167, 0.0168, 0.0155, 0.0145, 0.0131, 0.0143, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 19:50:34,635 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:51:01,957 INFO [train.py:904] (2/8) Epoch 29, batch 8950, loss[loss=0.1506, simple_loss=0.2479, pruned_loss=0.02666, over 16726.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2631, pruned_loss=0.03441, over 3058006.81 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:51:52,427 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293178.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:52:01,144 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-02 19:52:32,341 INFO [optim.py:368] (2/8) 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 19:52:53,145 INFO [train.py:904] (2/8) Epoch 29, batch 9000, loss[loss=0.1608, simple_loss=0.2486, pruned_loss=0.0365, over 12151.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2605, pruned_loss=0.03358, over 3072339.29 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:52:53,145 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 19:53:02,745 INFO [train.py:938] (2/8) Epoch 29, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01987, over 944034.00 frames. 2023-05-02 19:53:02,745 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 19:54:47,961 INFO [train.py:904] (2/8) Epoch 29, batch 9050, loss[loss=0.175, simple_loss=0.2673, pruned_loss=0.04129, over 16386.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2611, pruned_loss=0.03416, over 3071142.24 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:56:08,808 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5296, 3.0812, 3.2980, 1.8771, 3.4726, 3.5084, 2.9798, 2.8594], device='cuda:2'), covar=tensor([0.0671, 0.0298, 0.0261, 0.1222, 0.0107, 0.0192, 0.0427, 0.0442], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0108, 0.0099, 0.0136, 0.0085, 0.0128, 0.0126, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-02 19:56:15,025 INFO [optim.py:368] (2/8) 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 19:56:16,911 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 19:56:34,919 INFO [train.py:904] (2/8) Epoch 29, batch 9100, loss[loss=0.1619, simple_loss=0.2545, pruned_loss=0.03468, over 12337.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2604, pruned_loss=0.03444, over 3059525.62 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:57:17,187 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:34,110 INFO [train.py:904] (2/8) Epoch 29, batch 9150, loss[loss=0.1551, simple_loss=0.2443, pruned_loss=0.03299, over 12021.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2615, pruned_loss=0.03452, over 3056716.07 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:58:46,540 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0145, 2.6256, 2.9786, 2.0680, 2.7200, 2.1594, 2.7233, 2.8327], device='cuda:2'), covar=tensor([0.0288, 0.1023, 0.0466, 0.2006, 0.0797, 0.0965, 0.0578, 0.0854], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0166, 0.0167, 0.0154, 0.0145, 0.0131, 0.0142, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 19:58:52,007 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293361.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:56,758 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:00:04,477 INFO [optim.py:368] (2/8) 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 20:00:20,591 INFO [train.py:904] (2/8) Epoch 29, batch 9200, loss[loss=0.1682, simple_loss=0.2681, pruned_loss=0.03416, over 16685.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2574, pruned_loss=0.03337, over 3083100.77 frames. ], batch size: 62, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:00:55,040 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293422.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:00,125 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:10,965 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 20:01:23,104 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:58,342 INFO [train.py:904] (2/8) Epoch 29, batch 9250, loss[loss=0.1619, simple_loss=0.2704, pruned_loss=0.02669, over 16726.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2567, pruned_loss=0.03307, over 3069405.87 frames. ], batch size: 89, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:02:39,018 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:03:28,486 INFO [optim.py:368] (2/8) 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 20:03:40,177 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8447, 3.7866, 3.9398, 4.0164, 4.1052, 3.7073, 4.0835, 4.1511], device='cuda:2'), covar=tensor([0.1755, 0.1168, 0.1322, 0.0737, 0.0625, 0.1800, 0.0751, 0.0792], device='cuda:2'), in_proj_covar=tensor([0.0658, 0.0801, 0.0921, 0.0822, 0.0625, 0.0642, 0.0681, 0.0797], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:03:49,264 INFO [train.py:904] (2/8) Epoch 29, batch 9300, loss[loss=0.1327, simple_loss=0.2313, pruned_loss=0.01704, over 16876.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2555, pruned_loss=0.03283, over 3055140.07 frames. ], batch size: 90, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:05:33,314 INFO [train.py:904] (2/8) Epoch 29, batch 9350, loss[loss=0.1503, simple_loss=0.2552, pruned_loss=0.02265, over 16678.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2559, pruned_loss=0.03304, over 3059324.70 frames. ], batch size: 89, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:05:55,105 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 20:06:10,001 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5585, 3.7967, 3.7926, 2.7244, 3.3985, 3.8323, 3.5506, 2.2930], device='cuda:2'), covar=tensor([0.0481, 0.0056, 0.0059, 0.0367, 0.0124, 0.0096, 0.0091, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0102, 0.0114, 0.0098, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-02 20:06:28,229 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5933, 4.5800, 4.8800, 4.8619, 4.8690, 4.6432, 4.5665, 4.5569], device='cuda:2'), covar=tensor([0.0383, 0.0988, 0.0530, 0.0607, 0.0629, 0.0639, 0.0954, 0.0478], device='cuda:2'), in_proj_covar=tensor([0.0426, 0.0484, 0.0466, 0.0430, 0.0511, 0.0491, 0.0564, 0.0394], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 20:06:56,970 INFO [optim.py:368] (2/8) 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 20:07:15,905 INFO [train.py:904] (2/8) Epoch 29, batch 9400, loss[loss=0.1602, simple_loss=0.2585, pruned_loss=0.03094, over 15325.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2559, pruned_loss=0.03268, over 3064097.24 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:07:53,827 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:08:55,052 INFO [train.py:904] (2/8) Epoch 29, batch 9450, loss[loss=0.1626, simple_loss=0.26, pruned_loss=0.0326, over 15346.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2583, pruned_loss=0.03338, over 3067524.28 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:09:24,254 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7889, 2.5552, 2.4940, 3.8384, 1.9357, 3.8359, 1.5111, 3.0014], device='cuda:2'), covar=tensor([0.1488, 0.0913, 0.1186, 0.0198, 0.0118, 0.0381, 0.1885, 0.0722], device='cuda:2'), in_proj_covar=tensor([0.0172, 0.0178, 0.0197, 0.0198, 0.0201, 0.0213, 0.0208, 0.0196], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:09:29,418 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:09:45,136 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7503, 4.8335, 4.6448, 4.2600, 4.3418, 4.7383, 4.5218, 4.4526], device='cuda:2'), covar=tensor([0.0591, 0.0570, 0.0360, 0.0359, 0.0910, 0.0647, 0.0447, 0.0630], device='cuda:2'), in_proj_covar=tensor([0.0301, 0.0455, 0.0353, 0.0353, 0.0348, 0.0404, 0.0243, 0.0420], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:10:18,284 INFO [optim.py:368] (2/8) 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 20:10:34,115 INFO [train.py:904] (2/8) Epoch 29, batch 9500, loss[loss=0.1511, simple_loss=0.2469, pruned_loss=0.02768, over 16617.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2568, pruned_loss=0.0329, over 3060746.65 frames. ], batch size: 62, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:11:02,505 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:08,937 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:22,835 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6982, 4.7720, 4.5665, 4.1575, 4.2788, 4.6609, 4.4428, 4.3694], device='cuda:2'), covar=tensor([0.0611, 0.0566, 0.0356, 0.0392, 0.0975, 0.0580, 0.0476, 0.0682], device='cuda:2'), in_proj_covar=tensor([0.0302, 0.0455, 0.0353, 0.0353, 0.0348, 0.0405, 0.0243, 0.0420], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:11:25,225 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-02 20:11:42,042 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:12:17,511 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.8609, 3.8396, 4.1316, 4.1294, 4.1529, 3.9388, 3.9388, 3.9635], device='cuda:2'), covar=tensor([0.0439, 0.1085, 0.0595, 0.0541, 0.0555, 0.0564, 0.0925, 0.0512], device='cuda:2'), in_proj_covar=tensor([0.0425, 0.0482, 0.0465, 0.0427, 0.0508, 0.0489, 0.0561, 0.0393], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 20:12:19,483 INFO [train.py:904] (2/8) Epoch 29, batch 9550, loss[loss=0.1741, simple_loss=0.2833, pruned_loss=0.03243, over 15330.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2561, pruned_loss=0.03263, over 3079015.61 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:12:56,183 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3809, 4.2251, 4.4213, 4.5550, 4.6861, 4.2505, 4.6894, 4.7227], device='cuda:2'), covar=tensor([0.1877, 0.1208, 0.1539, 0.0772, 0.0629, 0.1135, 0.0668, 0.0766], device='cuda:2'), in_proj_covar=tensor([0.0654, 0.0798, 0.0916, 0.0816, 0.0620, 0.0639, 0.0680, 0.0791], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:12:59,187 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:24,954 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:43,637 INFO [optim.py:368] (2/8) 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 20:13:57,138 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4526, 3.4909, 3.6631, 3.6554, 3.6837, 3.5319, 3.5544, 3.5595], device='cuda:2'), covar=tensor([0.0415, 0.0991, 0.0641, 0.0558, 0.0553, 0.0628, 0.0758, 0.0548], device='cuda:2'), in_proj_covar=tensor([0.0422, 0.0480, 0.0463, 0.0424, 0.0505, 0.0486, 0.0558, 0.0390], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 20:13:59,133 INFO [train.py:904] (2/8) Epoch 29, batch 9600, loss[loss=0.1614, simple_loss=0.2585, pruned_loss=0.03218, over 16568.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2571, pruned_loss=0.03328, over 3069221.68 frames. ], batch size: 68, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:14:32,482 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:15:45,740 INFO [train.py:904] (2/8) Epoch 29, batch 9650, loss[loss=0.165, simple_loss=0.2641, pruned_loss=0.03292, over 16327.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2594, pruned_loss=0.03355, over 3066653.41 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:16:55,548 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2883, 3.5690, 3.9376, 2.2779, 3.2919, 2.6407, 3.7294, 3.7299], device='cuda:2'), covar=tensor([0.0241, 0.0806, 0.0480, 0.2020, 0.0715, 0.0943, 0.0539, 0.0891], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0154, 0.0144, 0.0130, 0.0142, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:17:16,475 INFO [optim.py:368] (2/8) 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 20:17:35,788 INFO [train.py:904] (2/8) Epoch 29, batch 9700, loss[loss=0.1657, simple_loss=0.2622, pruned_loss=0.03454, over 15363.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2587, pruned_loss=0.03354, over 3071896.15 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:18:06,873 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6675, 2.1326, 1.7382, 1.8721, 2.4293, 2.0938, 1.9787, 2.5521], device='cuda:2'), covar=tensor([0.0219, 0.0508, 0.0703, 0.0593, 0.0311, 0.0456, 0.0238, 0.0299], device='cuda:2'), in_proj_covar=tensor([0.0220, 0.0236, 0.0225, 0.0226, 0.0235, 0.0234, 0.0229, 0.0232], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:19:17,210 INFO [train.py:904] (2/8) Epoch 29, batch 9750, loss[loss=0.1724, simple_loss=0.2654, pruned_loss=0.03965, over 16949.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2572, pruned_loss=0.0338, over 3055476.40 frames. ], batch size: 109, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:20:24,618 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 20:20:52,019 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7125, 5.0069, 4.6382, 4.8767, 4.5047, 4.4459, 4.5250, 5.0736], device='cuda:2'), covar=tensor([0.1930, 0.1350, 0.1999, 0.1297, 0.1475, 0.1815, 0.2615, 0.1621], device='cuda:2'), in_proj_covar=tensor([0.0705, 0.0851, 0.0697, 0.0661, 0.0542, 0.0542, 0.0711, 0.0666], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:20:56,835 INFO [train.py:904] (2/8) Epoch 29, batch 9800, loss[loss=0.1665, simple_loss=0.2709, pruned_loss=0.03102, over 15472.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2576, pruned_loss=0.03321, over 3076497.10 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:21:23,014 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:24,919 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 20:21:27,404 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294020.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:47,014 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4270, 3.4852, 3.6588, 3.6557, 3.6735, 3.4748, 3.5254, 3.5724], device='cuda:2'), covar=tensor([0.0394, 0.0815, 0.0535, 0.0480, 0.0485, 0.0654, 0.0808, 0.0467], device='cuda:2'), in_proj_covar=tensor([0.0425, 0.0482, 0.0465, 0.0427, 0.0507, 0.0489, 0.0562, 0.0393], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 20:22:29,324 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:22:30,864 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7059, 2.6314, 1.9099, 2.7800, 2.1582, 2.8372, 2.1689, 2.3723], device='cuda:2'), covar=tensor([0.0343, 0.0368, 0.1330, 0.0281, 0.0722, 0.0514, 0.1259, 0.0671], device='cuda:2'), in_proj_covar=tensor([0.0171, 0.0175, 0.0190, 0.0166, 0.0175, 0.0212, 0.0201, 0.0178], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:22:38,994 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2361, 4.1295, 4.3352, 4.4221, 4.5588, 4.1478, 4.5463, 4.5792], device='cuda:2'), covar=tensor([0.1917, 0.1169, 0.1381, 0.0761, 0.0616, 0.1199, 0.0699, 0.0759], device='cuda:2'), in_proj_covar=tensor([0.0652, 0.0795, 0.0913, 0.0815, 0.0621, 0.0636, 0.0678, 0.0788], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:22:39,770 INFO [train.py:904] (2/8) Epoch 29, batch 9850, loss[loss=0.1666, simple_loss=0.262, pruned_loss=0.03565, over 16772.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.258, pruned_loss=0.03262, over 3073393.68 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 16.0 2023-05-02 20:23:02,391 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:23:09,662 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:24:14,913 INFO [optim.py:368] (2/8) 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 20:24:32,429 INFO [train.py:904] (2/8) Epoch 29, batch 9900, loss[loss=0.1617, simple_loss=0.2497, pruned_loss=0.03681, over 12399.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2577, pruned_loss=0.03242, over 3047763.31 frames. ], batch size: 250, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:24:40,050 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3626, 2.4528, 2.1077, 2.2380, 2.7972, 2.3506, 2.6619, 2.9626], device='cuda:2'), covar=tensor([0.0182, 0.0526, 0.0602, 0.0571, 0.0339, 0.0533, 0.0262, 0.0314], device='cuda:2'), in_proj_covar=tensor([0.0219, 0.0236, 0.0225, 0.0225, 0.0234, 0.0233, 0.0228, 0.0232], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:24:48,811 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4409, 3.3045, 2.7048, 2.1462, 2.1256, 2.2809, 3.4625, 2.9229], device='cuda:2'), covar=tensor([0.3111, 0.0637, 0.1862, 0.3281, 0.3185, 0.2422, 0.0441, 0.1641], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0268, 0.0307, 0.0322, 0.0297, 0.0273, 0.0297, 0.0342], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 20:25:38,229 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 20:26:27,773 INFO [train.py:904] (2/8) Epoch 29, batch 9950, loss[loss=0.1533, simple_loss=0.2525, pruned_loss=0.02702, over 16800.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2597, pruned_loss=0.03262, over 3051431.02 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:27:06,738 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:28:08,137 INFO [optim.py:368] (2/8) 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 20:28:27,669 INFO [train.py:904] (2/8) Epoch 29, batch 10000, loss[loss=0.1661, simple_loss=0.2691, pruned_loss=0.03154, over 16633.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2585, pruned_loss=0.03229, over 3064858.85 frames. ], batch size: 76, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:29:21,440 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294231.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:30:08,530 INFO [train.py:904] (2/8) Epoch 29, batch 10050, loss[loss=0.1626, simple_loss=0.2568, pruned_loss=0.03419, over 17124.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2595, pruned_loss=0.03282, over 3057058.76 frames. ], batch size: 49, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:30:24,660 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0558, 3.0519, 1.9309, 3.2986, 2.2766, 3.3031, 2.1028, 2.5590], device='cuda:2'), covar=tensor([0.0382, 0.0445, 0.1681, 0.0273, 0.1007, 0.0626, 0.1612, 0.0876], device='cuda:2'), in_proj_covar=tensor([0.0170, 0.0173, 0.0189, 0.0164, 0.0174, 0.0210, 0.0199, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:31:08,347 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 20:31:27,888 INFO [optim.py:368] (2/8) 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 20:31:41,194 INFO [train.py:904] (2/8) Epoch 29, batch 10100, loss[loss=0.1642, simple_loss=0.2577, pruned_loss=0.03534, over 16323.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2601, pruned_loss=0.03342, over 3037814.17 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:32:42,438 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:32:46,797 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 30, batch 0, loss[loss=0.211, simple_loss=0.2834, pruned_loss=0.06926, over 16781.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2834, pruned_loss=0.06926, over 16781.00 frames. ], batch size: 83, lr: 2.26e-03, grad_scale: 8.0 2023-05-02 20:33:20,747 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 20:33:28,206 INFO [train.py:938] (2/8) Epoch 30, validation: loss=0.1426, simple_loss=0.2458, pruned_loss=0.01974, over 944034.00 frames. 2023-05-02 20:33:28,207 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 20:34:28,809 INFO [optim.py:368] (2/8) 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 20:34:29,660 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294400.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:34:36,207 INFO [train.py:904] (2/8) Epoch 30, batch 50, loss[loss=0.2096, simple_loss=0.287, pruned_loss=0.06606, over 16266.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2603, pruned_loss=0.04488, over 749951.22 frames. ], batch size: 165, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:35:02,131 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294423.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:35:43,885 INFO [train.py:904] (2/8) Epoch 30, batch 100, loss[loss=0.179, simple_loss=0.2677, pruned_loss=0.04512, over 15569.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2591, pruned_loss=0.04224, over 1312521.00 frames. ], batch size: 190, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:36:24,497 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294484.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:32,223 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:46,221 INFO [optim.py:368] (2/8) 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 20:36:51,464 INFO [train.py:904] (2/8) Epoch 30, batch 150, loss[loss=0.1825, simple_loss=0.2854, pruned_loss=0.03982, over 17013.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2587, pruned_loss=0.04177, over 1755801.21 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:36:57,722 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8596, 1.3939, 1.6945, 1.7008, 1.8906, 1.9369, 1.6817, 1.9050], device='cuda:2'), covar=tensor([0.0277, 0.0465, 0.0245, 0.0325, 0.0312, 0.0235, 0.0472, 0.0161], device='cuda:2'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0188, 0.0206, 0.0163, 0.0201, 0.0164], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:37:20,777 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294526.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:40,157 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 20:37:50,485 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2683, 3.3725, 3.7209, 2.2030, 3.1280, 2.3820, 3.6216, 3.7558], device='cuda:2'), covar=tensor([0.0272, 0.0975, 0.0563, 0.2109, 0.0860, 0.1075, 0.0579, 0.0968], device='cuda:2'), in_proj_covar=tensor([0.0158, 0.0166, 0.0168, 0.0155, 0.0145, 0.0131, 0.0143, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:37:53,392 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:59,954 INFO [train.py:904] (2/8) Epoch 30, batch 200, loss[loss=0.1691, simple_loss=0.2673, pruned_loss=0.03548, over 17136.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2597, pruned_loss=0.04275, over 2098499.14 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:38:23,476 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:38:25,712 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.2336, 5.1614, 5.0428, 4.5621, 4.7053, 5.0836, 5.0976, 4.6570], device='cuda:2'), covar=tensor([0.0660, 0.0663, 0.0406, 0.0399, 0.1292, 0.0601, 0.0336, 0.0899], device='cuda:2'), in_proj_covar=tensor([0.0307, 0.0463, 0.0359, 0.0359, 0.0354, 0.0412, 0.0247, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:39:01,688 INFO [optim.py:368] (2/8) 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 20:39:06,663 INFO [train.py:904] (2/8) Epoch 30, batch 250, loss[loss=0.1588, simple_loss=0.2432, pruned_loss=0.0372, over 15551.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2575, pruned_loss=0.04224, over 2372433.02 frames. ], batch size: 191, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:39:45,364 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:01,698 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:15,312 INFO [train.py:904] (2/8) Epoch 30, batch 300, loss[loss=0.1803, simple_loss=0.26, pruned_loss=0.05033, over 16918.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2547, pruned_loss=0.04082, over 2584639.79 frames. ], batch size: 109, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:41:06,778 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:07,555 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 20:41:12,839 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:19,992 INFO [optim.py:368] (2/8) 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 20:41:25,715 INFO [train.py:904] (2/8) Epoch 30, batch 350, loss[loss=0.1981, simple_loss=0.2631, pruned_loss=0.06655, over 16804.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2528, pruned_loss=0.03985, over 2749014.89 frames. ], batch size: 90, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:41:57,362 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 20:42:07,962 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7138, 3.8887, 2.5492, 4.4810, 3.0355, 4.4167, 2.7107, 3.2579], device='cuda:2'), covar=tensor([0.0413, 0.0444, 0.1693, 0.0373, 0.1021, 0.0574, 0.1503, 0.0843], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0181, 0.0196, 0.0173, 0.0180, 0.0220, 0.0206, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:42:22,844 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4984, 3.5591, 2.7205, 2.2329, 2.2655, 2.2995, 3.6496, 3.0799], device='cuda:2'), covar=tensor([0.3182, 0.0694, 0.1988, 0.3175, 0.3008, 0.2421, 0.0556, 0.1828], device='cuda:2'), in_proj_covar=tensor([0.0336, 0.0274, 0.0314, 0.0328, 0.0304, 0.0279, 0.0303, 0.0351], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 20:42:33,013 INFO [train.py:904] (2/8) Epoch 30, batch 400, loss[loss=0.1584, simple_loss=0.2375, pruned_loss=0.0397, over 12203.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2515, pruned_loss=0.03915, over 2871931.74 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:42:50,068 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6422, 2.4181, 1.8757, 2.2500, 2.7801, 2.5345, 2.7516, 2.8465], device='cuda:2'), covar=tensor([0.0285, 0.0498, 0.0692, 0.0490, 0.0278, 0.0407, 0.0248, 0.0363], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0246, 0.0234, 0.0235, 0.0244, 0.0243, 0.0240, 0.0243], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:43:06,902 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294779.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:43:12,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3197, 5.2826, 5.0648, 4.5702, 5.0972, 2.1439, 4.9116, 4.9075], device='cuda:2'), covar=tensor([0.0086, 0.0088, 0.0226, 0.0394, 0.0112, 0.2756, 0.0137, 0.0248], device='cuda:2'), in_proj_covar=tensor([0.0180, 0.0173, 0.0210, 0.0181, 0.0188, 0.0217, 0.0199, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:43:34,529 INFO [optim.py:368] (2/8) 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 20:43:41,854 INFO [train.py:904] (2/8) Epoch 30, batch 450, loss[loss=0.1509, simple_loss=0.2323, pruned_loss=0.03473, over 16747.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2501, pruned_loss=0.0385, over 2980612.65 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:44:12,197 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294826.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:36,118 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:49,592 INFO [train.py:904] (2/8) Epoch 30, batch 500, loss[loss=0.153, simple_loss=0.2368, pruned_loss=0.03459, over 16813.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2491, pruned_loss=0.03839, over 3059333.25 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:45:17,739 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:45:52,585 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 30, batch 550, loss[loss=0.1715, simple_loss=0.2512, pruned_loss=0.04585, over 16298.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2486, pruned_loss=0.03802, over 3118279.44 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:46:29,020 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:47:00,919 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 20:47:05,460 INFO [train.py:904] (2/8) Epoch 30, batch 600, loss[loss=0.1396, simple_loss=0.2254, pruned_loss=0.02693, over 16962.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2474, pruned_loss=0.0372, over 3172272.59 frames. ], batch size: 41, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:47:52,936 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:01,956 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:07,369 INFO [optim.py:368] (2/8) 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 20:48:13,113 INFO [train.py:904] (2/8) Epoch 30, batch 650, loss[loss=0.1496, simple_loss=0.2255, pruned_loss=0.03691, over 16803.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2466, pruned_loss=0.03732, over 3207427.72 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:48:45,385 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 20:49:04,906 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295043.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:15,108 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295050.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 20:49:20,779 INFO [train.py:904] (2/8) Epoch 30, batch 700, loss[loss=0.1564, simple_loss=0.2397, pruned_loss=0.03655, over 16454.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2457, pruned_loss=0.03647, over 3231543.11 frames. ], batch size: 68, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:49:39,330 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:49,991 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-02 20:49:54,481 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:50:23,317 INFO [optim.py:368] (2/8) 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 20:50:27,886 INFO [train.py:904] (2/8) Epoch 30, batch 750, loss[loss=0.1682, simple_loss=0.2444, pruned_loss=0.04601, over 16488.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2462, pruned_loss=0.03676, over 3240436.08 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:50:31,016 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-02 20:51:00,101 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:03,383 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:24,557 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:38,201 INFO [train.py:904] (2/8) Epoch 30, batch 800, loss[loss=0.1752, simple_loss=0.2709, pruned_loss=0.0398, over 16570.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2467, pruned_loss=0.03701, over 3261764.10 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:52:29,858 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:52:30,316 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 20:52:42,511 INFO [optim.py:368] (2/8) 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 20:52:44,581 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4838, 4.4651, 4.3784, 3.8663, 4.4248, 1.8762, 4.1577, 3.8969], device='cuda:2'), covar=tensor([0.0130, 0.0105, 0.0196, 0.0311, 0.0110, 0.2924, 0.0157, 0.0304], device='cuda:2'), in_proj_covar=tensor([0.0183, 0.0175, 0.0213, 0.0183, 0.0191, 0.0219, 0.0201, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:52:48,753 INFO [train.py:904] (2/8) Epoch 30, batch 850, loss[loss=0.174, simple_loss=0.2465, pruned_loss=0.05078, over 16767.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2461, pruned_loss=0.03682, over 3275110.17 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:53:18,785 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295225.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:21,234 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295227.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:39,160 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3177, 5.6226, 5.4309, 5.4795, 5.1220, 5.0998, 5.0014, 5.7432], device='cuda:2'), covar=tensor([0.1302, 0.0977, 0.1079, 0.0849, 0.0894, 0.0826, 0.1337, 0.0958], device='cuda:2'), in_proj_covar=tensor([0.0730, 0.0878, 0.0724, 0.0684, 0.0561, 0.0559, 0.0741, 0.0692], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:53:48,677 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 20:53:56,732 INFO [train.py:904] (2/8) Epoch 30, batch 900, loss[loss=0.1295, simple_loss=0.2183, pruned_loss=0.02035, over 16485.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.245, pruned_loss=0.03621, over 3285627.02 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:54:27,486 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:33,474 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295279.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:43,643 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295286.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:53,565 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1386, 4.8709, 5.1698, 5.3376, 5.5486, 4.8668, 5.5265, 5.5275], device='cuda:2'), covar=tensor([0.2033, 0.1542, 0.1867, 0.0884, 0.0600, 0.0889, 0.0561, 0.0661], device='cuda:2'), in_proj_covar=tensor([0.0694, 0.0845, 0.0975, 0.0861, 0.0654, 0.0677, 0.0717, 0.0834], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 20:55:02,819 INFO [optim.py:368] (2/8) 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 20:55:08,080 INFO [train.py:904] (2/8) Epoch 30, batch 950, loss[loss=0.1483, simple_loss=0.2268, pruned_loss=0.03484, over 16495.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2448, pruned_loss=0.03629, over 3293014.05 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:55:43,400 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-05-02 20:55:49,202 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:59,232 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:06,344 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295345.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 20:56:17,471 INFO [train.py:904] (2/8) Epoch 30, batch 1000, loss[loss=0.1646, simple_loss=0.242, pruned_loss=0.0436, over 16434.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.2435, pruned_loss=0.03614, over 3297626.10 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:56:24,318 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:33,961 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7604, 3.5053, 3.8297, 1.9870, 3.9279, 4.0305, 3.2892, 2.9521], device='cuda:2'), covar=tensor([0.0773, 0.0287, 0.0231, 0.1277, 0.0123, 0.0193, 0.0400, 0.0497], device='cuda:2'), in_proj_covar=tensor([0.0148, 0.0111, 0.0102, 0.0139, 0.0087, 0.0132, 0.0130, 0.0130], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:56:44,319 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7674, 3.9839, 2.6610, 4.5716, 3.1557, 4.5115, 2.7975, 3.2931], device='cuda:2'), covar=tensor([0.0395, 0.0509, 0.1595, 0.0364, 0.0843, 0.0648, 0.1461, 0.0799], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0176, 0.0181, 0.0222, 0.0207, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 20:57:12,951 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295394.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:20,046 INFO [optim.py:368] (2/8) 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 20:57:26,447 INFO [train.py:904] (2/8) Epoch 30, batch 1050, loss[loss=0.1532, simple_loss=0.2362, pruned_loss=0.03511, over 11989.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.2442, pruned_loss=0.03624, over 3293529.06 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:57:43,144 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9629, 2.2535, 2.6618, 3.0035, 2.9196, 3.5265, 2.5681, 3.5479], device='cuda:2'), covar=tensor([0.0348, 0.0582, 0.0385, 0.0388, 0.0399, 0.0229, 0.0538, 0.0198], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0200, 0.0188, 0.0194, 0.0212, 0.0169, 0.0206, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 20:57:48,289 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295420.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:53,564 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:58:34,208 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3352, 5.9628, 6.0837, 5.7074, 5.8919, 6.4114, 5.9517, 5.6018], device='cuda:2'), covar=tensor([0.0967, 0.1999, 0.2609, 0.2450, 0.2585, 0.1013, 0.1596, 0.2517], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0651, 0.0727, 0.0527, 0.0704, 0.0744, 0.0557, 0.0701], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 20:58:36,216 INFO [train.py:904] (2/8) Epoch 30, batch 1100, loss[loss=0.1608, simple_loss=0.252, pruned_loss=0.03475, over 17078.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2436, pruned_loss=0.03582, over 3297248.73 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:59:38,542 INFO [optim.py:368] (2/8) 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 20:59:43,307 INFO [train.py:904] (2/8) Epoch 30, batch 1150, loss[loss=0.1557, simple_loss=0.238, pruned_loss=0.03671, over 11873.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2436, pruned_loss=0.03583, over 3296873.43 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:00:11,583 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9002, 2.0604, 2.4630, 2.7733, 2.7883, 2.8661, 2.1369, 3.0413], device='cuda:2'), covar=tensor([0.0240, 0.0590, 0.0403, 0.0344, 0.0355, 0.0321, 0.0640, 0.0203], device='cuda:2'), in_proj_covar=tensor([0.0200, 0.0199, 0.0187, 0.0193, 0.0211, 0.0168, 0.0205, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:00:52,300 INFO [train.py:904] (2/8) Epoch 30, batch 1200, loss[loss=0.1495, simple_loss=0.2327, pruned_loss=0.0331, over 15522.00 frames. ], tot_loss[loss=0.1569, simple_loss=0.2437, pruned_loss=0.03509, over 3313673.59 frames. ], batch size: 190, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:00:55,027 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:00:57,929 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295558.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:01:04,376 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0732, 2.0825, 2.6319, 2.9664, 2.9278, 3.1464, 2.0480, 3.2352], device='cuda:2'), covar=tensor([0.0230, 0.0635, 0.0388, 0.0296, 0.0347, 0.0246, 0.0711, 0.0204], device='cuda:2'), in_proj_covar=tensor([0.0201, 0.0200, 0.0188, 0.0194, 0.0212, 0.0169, 0.0206, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 21:01:29,886 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295581.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:01:32,762 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 21:01:55,307 INFO [optim.py:368] (2/8) 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 21:02:00,821 INFO [train.py:904] (2/8) Epoch 30, batch 1250, loss[loss=0.1768, simple_loss=0.2463, pruned_loss=0.05363, over 16775.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2434, pruned_loss=0.03562, over 3318328.43 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:02:18,653 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295617.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:02:18,841 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 21:02:21,351 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295619.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:02:24,417 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9954, 3.1109, 2.8217, 2.9111, 3.3362, 2.9815, 3.5713, 3.4747], device='cuda:2'), covar=tensor([0.0198, 0.0520, 0.0586, 0.0551, 0.0398, 0.0470, 0.0405, 0.0371], device='cuda:2'), in_proj_covar=tensor([0.0239, 0.0251, 0.0239, 0.0240, 0.0250, 0.0249, 0.0247, 0.0250], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:02:34,277 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295629.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:02:38,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.7803, 4.7296, 4.6921, 4.3227, 4.3963, 4.7726, 4.5383, 4.4744], device='cuda:2'), covar=tensor([0.0723, 0.1101, 0.0420, 0.0389, 0.0989, 0.0614, 0.0551, 0.0779], device='cuda:2'), in_proj_covar=tensor([0.0322, 0.0487, 0.0377, 0.0379, 0.0374, 0.0436, 0.0260, 0.0452], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:02:39,515 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4789, 5.8806, 5.6586, 5.7226, 5.2796, 5.3734, 5.2632, 6.0075], device='cuda:2'), covar=tensor([0.1552, 0.1001, 0.1028, 0.0924, 0.0958, 0.0734, 0.1355, 0.1020], device='cuda:2'), in_proj_covar=tensor([0.0744, 0.0896, 0.0737, 0.0698, 0.0573, 0.0570, 0.0754, 0.0705], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:02:42,475 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:02:58,030 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295645.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:08,456 INFO [train.py:904] (2/8) Epoch 30, batch 1300, loss[loss=0.1568, simple_loss=0.2463, pruned_loss=0.03362, over 16769.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2431, pruned_loss=0.03586, over 3322708.16 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:03:40,927 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2259, 2.7065, 2.2844, 2.5097, 2.9686, 2.6987, 3.0197, 3.0826], device='cuda:2'), covar=tensor([0.0283, 0.0487, 0.0642, 0.0504, 0.0358, 0.0459, 0.0328, 0.0365], device='cuda:2'), in_proj_covar=tensor([0.0239, 0.0252, 0.0239, 0.0240, 0.0250, 0.0248, 0.0247, 0.0250], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:03:52,611 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:57,719 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:59,070 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295690.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:04:02,885 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295693.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:11,931 INFO [optim.py:368] (2/8) 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 21:04:18,058 INFO [train.py:904] (2/8) Epoch 30, batch 1350, loss[loss=0.184, simple_loss=0.2557, pruned_loss=0.05612, over 16754.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.2442, pruned_loss=0.03574, over 3332815.95 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:04:33,097 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:46,286 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295724.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:17,181 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295746.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:26,448 INFO [train.py:904] (2/8) Epoch 30, batch 1400, loss[loss=0.1752, simple_loss=0.2444, pruned_loss=0.05301, over 16869.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2442, pruned_loss=0.03611, over 3333548.99 frames. ], batch size: 116, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:05:52,089 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295772.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:06:11,771 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0059, 3.6520, 4.0855, 2.2032, 4.2730, 4.3327, 3.2638, 3.3567], device='cuda:2'), covar=tensor([0.0691, 0.0324, 0.0263, 0.1174, 0.0113, 0.0255, 0.0455, 0.0446], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0140, 0.0088, 0.0134, 0.0131, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:06:15,775 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9907, 4.9157, 4.8817, 4.4366, 4.5902, 4.9197, 4.8139, 4.5992], device='cuda:2'), covar=tensor([0.0681, 0.0853, 0.0377, 0.0369, 0.0979, 0.0556, 0.0433, 0.0807], device='cuda:2'), in_proj_covar=tensor([0.0324, 0.0490, 0.0379, 0.0381, 0.0376, 0.0438, 0.0261, 0.0454], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:06:30,028 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 30, batch 1450, loss[loss=0.1342, simple_loss=0.218, pruned_loss=0.02518, over 16205.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2432, pruned_loss=0.03595, over 3336537.32 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:06:57,421 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:07:44,322 INFO [train.py:904] (2/8) Epoch 30, batch 1500, loss[loss=0.1766, simple_loss=0.2691, pruned_loss=0.04207, over 17072.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.2433, pruned_loss=0.03549, over 3339353.48 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:08:11,988 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:21,933 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:22,037 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:47,908 INFO [optim.py:368] (2/8) 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 21:08:54,105 INFO [train.py:904] (2/8) Epoch 30, batch 1550, loss[loss=0.1569, simple_loss=0.2527, pruned_loss=0.03057, over 17277.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2446, pruned_loss=0.03676, over 3324901.00 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:09:05,572 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295912.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:09:08,344 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295914.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:09:29,567 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295929.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:09:36,309 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295934.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:09:37,301 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295935.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:10:02,943 INFO [train.py:904] (2/8) Epoch 30, batch 1600, loss[loss=0.1712, simple_loss=0.251, pruned_loss=0.04573, over 16516.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.247, pruned_loss=0.03717, over 3334263.30 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:10:42,972 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:10:45,976 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295985.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:10:50,697 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:11:05,442 INFO [optim.py:368] (2/8) 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 21:11:11,748 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-05-02 21:11:14,063 INFO [train.py:904] (2/8) Epoch 30, batch 1650, loss[loss=0.1599, simple_loss=0.2553, pruned_loss=0.03227, over 17135.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2478, pruned_loss=0.03726, over 3337226.56 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:11:30,017 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296015.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:11:48,031 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4842, 3.1005, 3.4472, 1.9366, 3.5279, 3.5606, 2.9674, 2.6759], device='cuda:2'), covar=tensor([0.0799, 0.0295, 0.0241, 0.1163, 0.0150, 0.0220, 0.0512, 0.0492], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0113, 0.0104, 0.0141, 0.0089, 0.0135, 0.0132, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:12:00,954 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:06,307 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:16,940 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:24,046 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 21:12:24,439 INFO [train.py:904] (2/8) Epoch 30, batch 1700, loss[loss=0.1343, simple_loss=0.2192, pruned_loss=0.02472, over 16951.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2487, pruned_loss=0.0377, over 3340489.69 frames. ], batch size: 41, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:12:36,067 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296063.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:49,174 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-05-02 21:12:53,868 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-05-02 21:13:13,776 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-05-02 21:13:26,988 INFO [optim.py:368] (2/8) 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 21:13:30,414 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7652, 2.8000, 2.4261, 2.7039, 3.0665, 2.8005, 3.3166, 3.2390], device='cuda:2'), covar=tensor([0.0220, 0.0527, 0.0626, 0.0518, 0.0378, 0.0503, 0.0303, 0.0412], device='cuda:2'), in_proj_covar=tensor([0.0242, 0.0253, 0.0240, 0.0242, 0.0253, 0.0250, 0.0250, 0.0253], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:13:32,339 INFO [train.py:904] (2/8) Epoch 30, batch 1750, loss[loss=0.1793, simple_loss=0.26, pruned_loss=0.04927, over 16745.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2497, pruned_loss=0.03754, over 3341215.03 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:13:40,762 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:13:43,365 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 21:13:50,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4806, 3.6143, 4.0585, 2.2193, 3.3055, 2.6231, 3.8592, 3.8457], device='cuda:2'), covar=tensor([0.0327, 0.1100, 0.0545, 0.2355, 0.0906, 0.1078, 0.0719, 0.1281], device='cuda:2'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 21:13:57,270 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1670, 3.8940, 4.3164, 2.2456, 4.4913, 4.5692, 3.4082, 3.5822], device='cuda:2'), covar=tensor([0.0686, 0.0273, 0.0219, 0.1156, 0.0102, 0.0192, 0.0449, 0.0415], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0113, 0.0104, 0.0141, 0.0089, 0.0135, 0.0132, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:14:08,127 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 21:14:41,096 INFO [train.py:904] (2/8) Epoch 30, batch 1800, loss[loss=0.1696, simple_loss=0.2629, pruned_loss=0.03815, over 15521.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2506, pruned_loss=0.03745, over 3335002.30 frames. ], batch size: 190, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:14:50,584 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2768, 4.3496, 4.6285, 4.5978, 4.6718, 4.3687, 4.3813, 4.2825], device='cuda:2'), covar=tensor([0.0409, 0.0603, 0.0444, 0.0448, 0.0610, 0.0478, 0.0912, 0.0698], device='cuda:2'), in_proj_covar=tensor([0.0449, 0.0511, 0.0489, 0.0452, 0.0538, 0.0520, 0.0595, 0.0417], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-02 21:15:12,948 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:15:47,287 INFO [optim.py:368] (2/8) 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 21:15:51,605 INFO [train.py:904] (2/8) Epoch 30, batch 1850, loss[loss=0.1701, simple_loss=0.2478, pruned_loss=0.04623, over 16523.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2508, pruned_loss=0.03736, over 3341863.98 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:16:03,742 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296212.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:16:06,020 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296214.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:16:27,906 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296229.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:17:02,018 INFO [train.py:904] (2/8) Epoch 30, batch 1900, loss[loss=0.1708, simple_loss=0.2507, pruned_loss=0.04544, over 16734.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2503, pruned_loss=0.03715, over 3337837.82 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:17:09,977 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296260.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:17:12,890 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296262.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:17:45,501 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296285.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:18:06,421 INFO [optim.py:368] (2/8) 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 21:18:10,598 INFO [train.py:904] (2/8) Epoch 30, batch 1950, loss[loss=0.1823, simple_loss=0.2664, pruned_loss=0.04909, over 16251.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2501, pruned_loss=0.03696, over 3338473.35 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:18:26,867 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4229, 4.6990, 4.5536, 4.5482, 4.3056, 4.1867, 4.2285, 4.7613], device='cuda:2'), covar=tensor([0.1153, 0.0863, 0.0917, 0.0838, 0.0743, 0.1654, 0.1090, 0.0771], device='cuda:2'), in_proj_covar=tensor([0.0740, 0.0890, 0.0733, 0.0695, 0.0570, 0.0564, 0.0749, 0.0701], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:18:26,987 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:18:44,643 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9120, 2.8719, 2.7082, 4.5033, 3.4175, 4.1762, 1.8212, 3.1549], device='cuda:2'), covar=tensor([0.1442, 0.0800, 0.1237, 0.0214, 0.0208, 0.0419, 0.1737, 0.0835], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0184, 0.0203, 0.0209, 0.0209, 0.0221, 0.0214, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:18:52,565 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296333.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:19:02,384 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296341.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:19:21,053 INFO [train.py:904] (2/8) Epoch 30, batch 2000, loss[loss=0.1228, simple_loss=0.2084, pruned_loss=0.01863, over 16997.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2502, pruned_loss=0.03662, over 3328187.71 frames. ], batch size: 41, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:19:30,343 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9413, 2.5650, 2.0743, 2.3089, 2.8899, 2.6457, 2.9537, 3.0112], device='cuda:2'), covar=tensor([0.0271, 0.0463, 0.0618, 0.0496, 0.0288, 0.0418, 0.0234, 0.0342], device='cuda:2'), in_proj_covar=tensor([0.0241, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0249, 0.0252], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:19:30,367 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7356, 2.6626, 2.6627, 4.2150, 3.5448, 4.1364, 1.7017, 2.9808], device='cuda:2'), covar=tensor([0.1685, 0.0818, 0.1230, 0.0239, 0.0180, 0.0370, 0.1832, 0.0915], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0185, 0.0204, 0.0210, 0.0209, 0.0222, 0.0214, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:19:51,079 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:09,535 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296389.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:25,443 INFO [optim.py:368] (2/8) 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 21:20:30,259 INFO [train.py:904] (2/8) Epoch 30, batch 2050, loss[loss=0.1353, simple_loss=0.2255, pruned_loss=0.02251, over 16814.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2492, pruned_loss=0.03645, over 3326545.30 frames. ], batch size: 42, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:20:32,245 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:12,406 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:36,293 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.6874, 4.7850, 4.9779, 4.6862, 4.7745, 5.3944, 4.8752, 4.5478], device='cuda:2'), covar=tensor([0.1607, 0.2208, 0.2495, 0.2340, 0.2788, 0.1141, 0.1959, 0.2764], device='cuda:2'), in_proj_covar=tensor([0.0438, 0.0654, 0.0729, 0.0532, 0.0709, 0.0748, 0.0562, 0.0706], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:21:39,734 INFO [train.py:904] (2/8) Epoch 30, batch 2100, loss[loss=0.168, simple_loss=0.2475, pruned_loss=0.04428, over 16685.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2499, pruned_loss=0.03711, over 3317295.91 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:22:09,538 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296476.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:33,152 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:36,807 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:38,999 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296497.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:46,673 INFO [optim.py:368] (2/8) 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 21:22:48,891 INFO [train.py:904] (2/8) Epoch 30, batch 2150, loss[loss=0.1814, simple_loss=0.2661, pruned_loss=0.04835, over 16866.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.251, pruned_loss=0.03734, over 3314672.85 frames. ], batch size: 96, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:22:49,448 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2023-05-02 21:23:03,725 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296524.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:23,346 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296529.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:23:41,321 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7656, 2.6855, 2.3014, 2.5896, 2.9891, 2.7261, 3.3065, 3.2193], device='cuda:2'), covar=tensor([0.0193, 0.0552, 0.0629, 0.0552, 0.0371, 0.0491, 0.0292, 0.0349], device='cuda:2'), in_proj_covar=tensor([0.0243, 0.0254, 0.0241, 0.0243, 0.0254, 0.0251, 0.0251, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:23:56,771 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296553.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:23:57,526 INFO [train.py:904] (2/8) Epoch 30, batch 2200, loss[loss=0.1748, simple_loss=0.2726, pruned_loss=0.03847, over 17147.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2524, pruned_loss=0.03817, over 3315010.19 frames. ], batch size: 49, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:24:04,177 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:10,383 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-02 21:24:27,054 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:30,753 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:47,358 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9741, 4.7790, 5.0341, 5.2133, 5.3978, 4.7658, 5.3835, 5.3969], device='cuda:2'), covar=tensor([0.2246, 0.1521, 0.1909, 0.0836, 0.0632, 0.1017, 0.0665, 0.0704], device='cuda:2'), in_proj_covar=tensor([0.0716, 0.0869, 0.1004, 0.0886, 0.0672, 0.0697, 0.0737, 0.0859], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:24:47,385 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.3012, 5.3157, 5.1831, 4.6382, 4.8451, 5.2204, 5.2022, 4.8010], device='cuda:2'), covar=tensor([0.0679, 0.0633, 0.0376, 0.0405, 0.1212, 0.0542, 0.0345, 0.0946], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0498, 0.0385, 0.0387, 0.0381, 0.0444, 0.0265, 0.0460], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:25:04,056 INFO [optim.py:368] (2/8) 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 21:25:06,331 INFO [train.py:904] (2/8) Epoch 30, batch 2250, loss[loss=0.1611, simple_loss=0.2603, pruned_loss=0.03091, over 17020.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2528, pruned_loss=0.03816, over 3318612.92 frames. ], batch size: 55, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:25:20,061 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:26:17,732 INFO [train.py:904] (2/8) Epoch 30, batch 2300, loss[loss=0.1586, simple_loss=0.2412, pruned_loss=0.038, over 16525.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2533, pruned_loss=0.03862, over 3313346.20 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:26:29,225 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 21:26:41,377 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:26:45,128 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:26:48,163 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4180, 2.9843, 3.3296, 1.9271, 3.4020, 3.3843, 2.8606, 2.6824], device='cuda:2'), covar=tensor([0.0832, 0.0331, 0.0231, 0.1203, 0.0162, 0.0296, 0.0506, 0.0499], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0113, 0.0104, 0.0141, 0.0089, 0.0136, 0.0133, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:26:49,306 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6992, 6.0615, 5.8155, 5.8796, 5.4946, 5.4718, 5.4569, 6.2010], device='cuda:2'), covar=tensor([0.1636, 0.0974, 0.1174, 0.0891, 0.0893, 0.0661, 0.1364, 0.0915], device='cuda:2'), in_proj_covar=tensor([0.0741, 0.0892, 0.0734, 0.0695, 0.0570, 0.0565, 0.0749, 0.0702], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:27:05,621 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:08,281 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 21:27:24,078 INFO [optim.py:368] (2/8) 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 21:27:27,341 INFO [train.py:904] (2/8) Epoch 30, batch 2350, loss[loss=0.1871, simple_loss=0.2683, pruned_loss=0.05301, over 16681.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2534, pruned_loss=0.03922, over 3304043.27 frames. ], batch size: 76, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:27:28,738 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:30,501 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:33,600 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1516, 3.1084, 3.5198, 2.2744, 3.0484, 2.3695, 3.6329, 3.4931], device='cuda:2'), covar=tensor([0.0274, 0.1075, 0.0631, 0.2009, 0.0893, 0.1121, 0.0565, 0.1039], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0173, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 21:28:36,423 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296753.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:37,278 INFO [train.py:904] (2/8) Epoch 30, batch 2400, loss[loss=0.1789, simple_loss=0.2748, pruned_loss=0.0415, over 17010.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2532, pruned_loss=0.03895, over 3307890.95 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:29:25,494 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.6487, 2.8259, 3.1155, 2.0543, 2.7357, 2.1288, 3.2620, 3.2093], device='cuda:2'), covar=tensor([0.0286, 0.1075, 0.0616, 0.2050, 0.0952, 0.1122, 0.0611, 0.0847], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0173, 0.0173, 0.0160, 0.0150, 0.0134, 0.0148, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 21:29:27,128 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:29:30,272 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4831, 5.8542, 5.5944, 5.6847, 5.3168, 5.3109, 5.1829, 5.9814], device='cuda:2'), covar=tensor([0.1599, 0.1009, 0.1311, 0.0940, 0.0914, 0.0713, 0.1337, 0.0985], device='cuda:2'), in_proj_covar=tensor([0.0744, 0.0895, 0.0737, 0.0697, 0.0572, 0.0567, 0.0751, 0.0705], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:29:39,881 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5948, 4.5416, 4.4974, 3.9480, 4.5367, 1.8640, 4.2980, 4.0596], device='cuda:2'), covar=tensor([0.0150, 0.0138, 0.0205, 0.0327, 0.0117, 0.2726, 0.0166, 0.0255], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0179, 0.0217, 0.0187, 0.0195, 0.0222, 0.0206, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:29:40,134 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 21:29:43,967 INFO [optim.py:368] (2/8) 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 21:29:46,361 INFO [train.py:904] (2/8) Epoch 30, batch 2450, loss[loss=0.1549, simple_loss=0.2551, pruned_loss=0.02735, over 17231.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2544, pruned_loss=0.03886, over 3316773.61 frames. ], batch size: 45, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:30:05,756 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1967, 4.2138, 4.4950, 4.4975, 4.5399, 4.2768, 4.3002, 4.1988], device='cuda:2'), covar=tensor([0.0375, 0.0676, 0.0446, 0.0407, 0.0497, 0.0458, 0.0787, 0.0649], device='cuda:2'), in_proj_covar=tensor([0.0448, 0.0510, 0.0487, 0.0450, 0.0532, 0.0515, 0.0592, 0.0414], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 21:30:48,229 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296848.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:30:53,679 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3343, 2.5967, 2.5623, 4.3817, 3.3152, 3.9892, 1.9431, 2.9288], device='cuda:2'), covar=tensor([0.1123, 0.0922, 0.1332, 0.0300, 0.0269, 0.0660, 0.1538, 0.0966], device='cuda:2'), in_proj_covar=tensor([0.0176, 0.0185, 0.0204, 0.0210, 0.0209, 0.0222, 0.0214, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:30:55,423 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 30, batch 2500, loss[loss=0.1607, simple_loss=0.2433, pruned_loss=0.03904, over 16656.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.254, pruned_loss=0.03864, over 3312347.27 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:31:19,655 INFO [zipformer.py:625] (2/8) 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] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:32:04,913 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 30, batch 2550, loss[loss=0.1511, simple_loss=0.2447, pruned_loss=0.02879, over 17211.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2541, pruned_loss=0.03854, over 3308385.94 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:32:14,196 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8978, 4.5094, 2.9792, 2.4431, 2.9227, 2.5934, 4.7945, 3.6453], device='cuda:2'), covar=tensor([0.3255, 0.0585, 0.2184, 0.3090, 0.3070, 0.2360, 0.0436, 0.1574], device='cuda:2'), in_proj_covar=tensor([0.0339, 0.0278, 0.0316, 0.0330, 0.0309, 0.0282, 0.0307, 0.0357], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:32:49,068 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296934.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:32:54,103 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:15,497 INFO [train.py:904] (2/8) Epoch 30, batch 2600, loss[loss=0.1781, simple_loss=0.2558, pruned_loss=0.05019, over 16920.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.254, pruned_loss=0.03851, over 3311919.52 frames. ], batch size: 109, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:33:36,661 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:39,717 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:17,879 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:22,262 INFO [optim.py:368] (2/8) 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 21:34:24,417 INFO [train.py:904] (2/8) Epoch 30, batch 2650, loss[loss=0.1721, simple_loss=0.2694, pruned_loss=0.0374, over 16753.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2546, pruned_loss=0.03792, over 3309688.93 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:34:46,003 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297019.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:35:11,960 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1729, 2.6414, 2.2393, 2.4437, 2.9530, 2.7115, 2.9427, 3.1094], device='cuda:2'), covar=tensor([0.0294, 0.0483, 0.0635, 0.0517, 0.0308, 0.0412, 0.0311, 0.0306], device='cuda:2'), in_proj_covar=tensor([0.0242, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0250, 0.0252], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:35:20,545 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:35:34,417 INFO [train.py:904] (2/8) Epoch 30, batch 2700, loss[loss=0.1972, simple_loss=0.2776, pruned_loss=0.05843, over 16894.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2547, pruned_loss=0.0374, over 3318590.75 frames. ], batch size: 109, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:35:51,464 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3065, 2.4460, 2.5000, 4.1677, 2.3369, 2.7684, 2.5062, 2.5321], device='cuda:2'), covar=tensor([0.1518, 0.3837, 0.3261, 0.0636, 0.4198, 0.2738, 0.3825, 0.3749], device='cuda:2'), in_proj_covar=tensor([0.0430, 0.0483, 0.0394, 0.0345, 0.0451, 0.0555, 0.0456, 0.0566], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:36:23,941 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:36:42,009 INFO [optim.py:368] (2/8) 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 21:36:43,157 INFO [train.py:904] (2/8) Epoch 30, batch 2750, loss[loss=0.1658, simple_loss=0.2503, pruned_loss=0.0406, over 15653.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2557, pruned_loss=0.03755, over 3324958.30 frames. ], batch size: 191, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:37:03,161 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7160, 2.7669, 2.6734, 4.8652, 3.8470, 4.2400, 1.6868, 3.0199], device='cuda:2'), covar=tensor([0.1474, 0.0880, 0.1300, 0.0203, 0.0233, 0.0426, 0.1749, 0.0907], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0210, 0.0210, 0.0222, 0.0214, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:37:17,888 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.7635, 3.7499, 2.4385, 4.3111, 3.0279, 4.2355, 2.4819, 3.1779], device='cuda:2'), covar=tensor([0.0301, 0.0400, 0.1622, 0.0343, 0.0820, 0.0555, 0.1625, 0.0750], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0186, 0.0199, 0.0180, 0.0184, 0.0226, 0.0209, 0.0187], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:37:30,682 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:44,442 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297148.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:50,746 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:51,162 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 21:37:51,587 INFO [train.py:904] (2/8) Epoch 30, batch 2800, loss[loss=0.1538, simple_loss=0.2488, pruned_loss=0.0294, over 16698.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2553, pruned_loss=0.0374, over 3323772.43 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:38:14,210 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:14,684 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 21:38:49,782 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:50,354 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 21:38:57,320 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:00,128 INFO [optim.py:368] (2/8) 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 21:39:01,278 INFO [train.py:904] (2/8) Epoch 30, batch 2850, loss[loss=0.1566, simple_loss=0.244, pruned_loss=0.03466, over 16976.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2545, pruned_loss=0.03753, over 3322190.78 frames. ], batch size: 41, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:39:21,041 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:26,923 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:37,163 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 30, batch 2900, loss[loss=0.1618, simple_loss=0.2572, pruned_loss=0.0332, over 16687.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2536, pruned_loss=0.03744, over 3321065.83 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:40:28,351 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7520, 4.2737, 2.6610, 2.3454, 2.4202, 2.4598, 4.5872, 3.3196], device='cuda:2'), covar=tensor([0.3300, 0.0673, 0.2449, 0.3193, 0.3439, 0.2523, 0.0506, 0.1711], device='cuda:2'), in_proj_covar=tensor([0.0343, 0.0280, 0.0319, 0.0333, 0.0312, 0.0285, 0.0310, 0.0361], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:40:33,060 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:40:41,559 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 21:40:45,386 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8858, 2.0654, 2.3248, 3.4047, 2.1086, 2.2715, 2.2332, 2.2152], device='cuda:2'), covar=tensor([0.1991, 0.4428, 0.3203, 0.1000, 0.4698, 0.3206, 0.4038, 0.4150], device='cuda:2'), in_proj_covar=tensor([0.0432, 0.0485, 0.0395, 0.0346, 0.0452, 0.0556, 0.0457, 0.0568], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:40:51,716 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=297283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:08,333 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297294.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:11,857 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1991, 5.7091, 5.8532, 5.4648, 5.6342, 6.1946, 5.6674, 5.3614], device='cuda:2'), covar=tensor([0.0912, 0.2055, 0.2710, 0.2076, 0.2614, 0.0911, 0.1520, 0.2403], device='cuda:2'), in_proj_covar=tensor([0.0441, 0.0657, 0.0732, 0.0532, 0.0712, 0.0750, 0.0563, 0.0710], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:41:22,344 INFO [optim.py:368] (2/8) 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] (2/8) Epoch 30, batch 2950, loss[loss=0.1674, simple_loss=0.2486, pruned_loss=0.04313, over 16249.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2535, pruned_loss=0.03811, over 3313665.84 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:41:41,509 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:04,303 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.5414, 5.5720, 5.4398, 4.9887, 5.0765, 5.4744, 5.4055, 5.0693], device='cuda:2'), covar=tensor([0.0544, 0.0379, 0.0301, 0.0308, 0.1001, 0.0364, 0.0256, 0.0739], device='cuda:2'), in_proj_covar=tensor([0.0329, 0.0499, 0.0385, 0.0389, 0.0382, 0.0446, 0.0265, 0.0462], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:42:18,972 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:33,298 INFO [train.py:904] (2/8) Epoch 30, batch 3000, loss[loss=0.1403, simple_loss=0.2294, pruned_loss=0.02556, over 17240.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2532, pruned_loss=0.03853, over 3316759.75 frames. ], batch size: 45, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:42:33,298 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 21:42:40,766 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2542, 3.2458, 3.4106, 2.4074, 3.1669, 3.4595, 3.3152, 2.2408], device='cuda:2'), covar=tensor([0.0492, 0.0132, 0.0080, 0.0429, 0.0144, 0.0098, 0.0110, 0.0472], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0136, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 21:42:42,087 INFO [train.py:938] (2/8) Epoch 30, validation: loss=0.1331, simple_loss=0.238, pruned_loss=0.01415, over 944034.00 frames. 2023-05-02 21:42:42,088 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 21:43:04,966 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-05-02 21:43:35,186 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297392.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:43:51,743 INFO [optim.py:368] (2/8) 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 21:43:51,758 INFO [train.py:904] (2/8) Epoch 30, batch 3050, loss[loss=0.1582, simple_loss=0.2424, pruned_loss=0.03705, over 16891.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2529, pruned_loss=0.03836, over 3321445.58 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:44:20,936 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.4827, 5.8761, 5.6587, 5.7229, 5.3107, 5.4108, 5.2724, 6.0504], device='cuda:2'), covar=tensor([0.1627, 0.1036, 0.1047, 0.0907, 0.1062, 0.0744, 0.1331, 0.0899], device='cuda:2'), in_proj_covar=tensor([0.0747, 0.0898, 0.0741, 0.0699, 0.0576, 0.0568, 0.0755, 0.0706], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:44:43,097 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1202, 5.1528, 5.5461, 5.5152, 5.5446, 5.2235, 5.1610, 4.9905], device='cuda:2'), covar=tensor([0.0344, 0.0595, 0.0381, 0.0422, 0.0513, 0.0387, 0.0992, 0.0469], device='cuda:2'), in_proj_covar=tensor([0.0451, 0.0514, 0.0490, 0.0453, 0.0536, 0.0518, 0.0598, 0.0417], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-02 21:45:01,424 INFO [train.py:904] (2/8) Epoch 30, batch 3100, loss[loss=0.1674, simple_loss=0.2437, pruned_loss=0.04551, over 16688.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2528, pruned_loss=0.0385, over 3330171.55 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:25,629 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 21:46:07,373 INFO [optim.py:368] (2/8) 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 21:46:07,396 INFO [train.py:904] (2/8) Epoch 30, batch 3150, loss[loss=0.1723, simple_loss=0.2484, pruned_loss=0.04811, over 16865.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2522, pruned_loss=0.0387, over 3321891.41 frames. ], batch size: 116, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:46:10,875 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8010, 2.1052, 2.5937, 2.8721, 2.7273, 3.4419, 2.4752, 3.4238], device='cuda:2'), covar=tensor([0.0399, 0.0619, 0.0431, 0.0459, 0.0483, 0.0256, 0.0555, 0.0223], device='cuda:2'), in_proj_covar=tensor([0.0206, 0.0205, 0.0194, 0.0199, 0.0218, 0.0174, 0.0211, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 21:46:42,038 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297529.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:47:17,204 INFO [train.py:904] (2/8) Epoch 30, batch 3200, loss[loss=0.1583, simple_loss=0.2418, pruned_loss=0.03743, over 16273.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2522, pruned_loss=0.03885, over 3311411.06 frames. ], batch size: 36, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:47:49,838 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:47:51,696 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:48:13,022 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:48:27,238 INFO [optim.py:368] (2/8) 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 21:48:27,253 INFO [train.py:904] (2/8) Epoch 30, batch 3250, loss[loss=0.1414, simple_loss=0.2329, pruned_loss=0.02497, over 16851.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2512, pruned_loss=0.03817, over 3308371.19 frames. ], batch size: 42, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:48:55,607 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7933, 2.8069, 2.7008, 5.0412, 3.9187, 4.3996, 1.6384, 3.2284], device='cuda:2'), covar=tensor([0.1367, 0.0844, 0.1228, 0.0178, 0.0189, 0.0341, 0.1700, 0.0750], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0210, 0.0210, 0.0222, 0.0214, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 21:49:20,231 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297642.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:49:36,041 INFO [train.py:904] (2/8) Epoch 30, batch 3300, loss[loss=0.1902, simple_loss=0.2653, pruned_loss=0.0576, over 16916.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2511, pruned_loss=0.03793, over 3316636.19 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:50:46,297 INFO [optim.py:368] (2/8) 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 21:50:46,312 INFO [train.py:904] (2/8) Epoch 30, batch 3350, loss[loss=0.1944, simple_loss=0.2884, pruned_loss=0.05021, over 16669.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2522, pruned_loss=0.03801, over 3314602.88 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:50:48,863 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-05-02 21:50:56,858 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3205, 2.4217, 2.4673, 4.0055, 2.3019, 2.7692, 2.4684, 2.5698], device='cuda:2'), covar=tensor([0.1483, 0.3753, 0.3329, 0.0718, 0.4333, 0.2620, 0.3784, 0.3553], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0486, 0.0396, 0.0347, 0.0454, 0.0558, 0.0459, 0.0570], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 21:51:55,689 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 21:51:56,062 INFO [train.py:904] (2/8) Epoch 30, batch 3400, loss[loss=0.1407, simple_loss=0.227, pruned_loss=0.02722, over 16819.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2514, pruned_loss=0.03718, over 3322446.58 frames. ], batch size: 39, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:53:02,949 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 21:53:07,131 INFO [optim.py:368] (2/8) 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 21:53:07,146 INFO [train.py:904] (2/8) Epoch 30, batch 3450, loss[loss=0.1381, simple_loss=0.222, pruned_loss=0.02708, over 15935.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2502, pruned_loss=0.03674, over 3313744.16 frames. ], batch size: 35, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:54:17,197 INFO [train.py:904] (2/8) Epoch 30, batch 3500, loss[loss=0.145, simple_loss=0.2316, pruned_loss=0.02918, over 16876.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2498, pruned_loss=0.03657, over 3316207.64 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:54:51,987 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:55:28,082 INFO [train.py:904] (2/8) Epoch 30, batch 3550, loss[loss=0.1512, simple_loss=0.2439, pruned_loss=0.02925, over 17208.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2489, pruned_loss=0.03615, over 3322183.00 frames. ], batch size: 45, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:55:29,302 INFO [optim.py:368] (2/8) 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 21:55:58,182 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.7407, 2.3675, 1.7948, 1.9502, 2.6618, 2.4009, 2.7475, 2.7564], device='cuda:2'), covar=tensor([0.0335, 0.0588, 0.0852, 0.0767, 0.0384, 0.0577, 0.0331, 0.0464], device='cuda:2'), in_proj_covar=tensor([0.0243, 0.0252, 0.0240, 0.0242, 0.0252, 0.0251, 0.0251, 0.0253], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:55:59,066 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:56:38,331 INFO [train.py:904] (2/8) Epoch 30, batch 3600, loss[loss=0.1389, simple_loss=0.2244, pruned_loss=0.02671, over 16987.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2478, pruned_loss=0.03616, over 3321296.05 frames. ], batch size: 41, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:56:42,424 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:57:07,255 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.6252, 5.9748, 5.7465, 5.8220, 5.4634, 5.4563, 5.2924, 6.1257], device='cuda:2'), covar=tensor([0.1452, 0.0958, 0.0968, 0.0919, 0.0882, 0.0691, 0.1424, 0.0870], device='cuda:2'), in_proj_covar=tensor([0.0750, 0.0904, 0.0743, 0.0703, 0.0580, 0.0569, 0.0760, 0.0711], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:57:07,310 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0531, 4.8323, 5.0967, 5.2823, 5.4479, 4.7799, 5.4616, 5.4554], device='cuda:2'), covar=tensor([0.2041, 0.1384, 0.1775, 0.0786, 0.0577, 0.1024, 0.0555, 0.0652], device='cuda:2'), in_proj_covar=tensor([0.0726, 0.0882, 0.1022, 0.0902, 0.0682, 0.0713, 0.0750, 0.0873], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:57:48,774 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6721, 1.7857, 1.6659, 1.5381, 1.9261, 1.5762, 1.6099, 1.8760], device='cuda:2'), covar=tensor([0.0286, 0.0356, 0.0514, 0.0409, 0.0266, 0.0328, 0.0215, 0.0247], device='cuda:2'), in_proj_covar=tensor([0.0243, 0.0252, 0.0241, 0.0243, 0.0252, 0.0251, 0.0252, 0.0254], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 21:57:53,330 INFO [train.py:904] (2/8) Epoch 30, batch 3650, loss[loss=0.1679, simple_loss=0.2407, pruned_loss=0.04754, over 16755.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.247, pruned_loss=0.03702, over 3306254.01 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:57:55,114 INFO [optim.py:368] (2/8) 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 21:58:14,201 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:59:06,224 INFO [train.py:904] (2/8) Epoch 30, batch 3700, loss[loss=0.1865, simple_loss=0.2701, pruned_loss=0.05146, over 15605.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2459, pruned_loss=0.03858, over 3288020.90 frames. ], batch size: 191, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:59:23,661 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 22:00:17,693 INFO [train.py:904] (2/8) Epoch 30, batch 3750, loss[loss=0.1796, simple_loss=0.2747, pruned_loss=0.04228, over 17246.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2469, pruned_loss=0.04027, over 3272220.78 frames. ], batch size: 52, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:19,704 INFO [optim.py:368] (2/8) 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 22:00:39,482 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8401, 1.9942, 2.4835, 2.6955, 2.7486, 2.7390, 2.0509, 2.9406], device='cuda:2'), covar=tensor([0.0212, 0.0613, 0.0407, 0.0341, 0.0397, 0.0430, 0.0662, 0.0222], device='cuda:2'), in_proj_covar=tensor([0.0205, 0.0204, 0.0193, 0.0199, 0.0217, 0.0174, 0.0209, 0.0174], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 22:00:52,421 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.9458, 3.0314, 3.2855, 2.1864, 2.9185, 2.2828, 3.5057, 3.5288], device='cuda:2'), covar=tensor([0.0247, 0.0972, 0.0712, 0.2005, 0.0912, 0.1107, 0.0536, 0.0825], device='cuda:2'), in_proj_covar=tensor([0.0164, 0.0175, 0.0173, 0.0160, 0.0150, 0.0134, 0.0147, 0.0188], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 22:01:30,455 INFO [train.py:904] (2/8) Epoch 30, batch 3800, loss[loss=0.1739, simple_loss=0.2521, pruned_loss=0.04789, over 16523.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2487, pruned_loss=0.04189, over 3265845.79 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:01:39,515 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 22:02:43,875 INFO [train.py:904] (2/8) Epoch 30, batch 3850, loss[loss=0.149, simple_loss=0.2311, pruned_loss=0.03342, over 16491.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2497, pruned_loss=0.04266, over 3259406.41 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:44,948 INFO [optim.py:368] (2/8) 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 22:03:53,265 INFO [train.py:904] (2/8) Epoch 30, batch 3900, loss[loss=0.1888, simple_loss=0.2671, pruned_loss=0.05527, over 15642.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2489, pruned_loss=0.04283, over 3265556.31 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:05:04,298 INFO [train.py:904] (2/8) Epoch 30, batch 3950, loss[loss=0.1711, simple_loss=0.2438, pruned_loss=0.04919, over 16903.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2487, pruned_loss=0.04336, over 3275228.79 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:05:05,533 INFO [optim.py:368] (2/8) 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 22:05:17,507 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:05:32,468 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 22:05:35,226 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0034, 3.9404, 3.9629, 2.9277, 3.9297, 1.7431, 3.6769, 3.2965], device='cuda:2'), covar=tensor([0.0223, 0.0189, 0.0260, 0.0536, 0.0148, 0.3592, 0.0230, 0.0485], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0180, 0.0219, 0.0190, 0.0197, 0.0223, 0.0209, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 22:06:15,411 INFO [train.py:904] (2/8) Epoch 30, batch 4000, loss[loss=0.1909, simple_loss=0.2714, pruned_loss=0.05522, over 12259.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.248, pruned_loss=0.04316, over 3274628.52 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:06:44,269 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 22:07:25,658 INFO [train.py:904] (2/8) Epoch 30, batch 4050, loss[loss=0.1685, simple_loss=0.2557, pruned_loss=0.04066, over 16897.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2495, pruned_loss=0.04267, over 3273397.22 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:07:27,598 INFO [optim.py:368] (2/8) 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 22:08:37,711 INFO [train.py:904] (2/8) Epoch 30, batch 4100, loss[loss=0.1637, simple_loss=0.2528, pruned_loss=0.03732, over 16564.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2515, pruned_loss=0.0426, over 3265808.09 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:09:19,614 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:09:53,959 INFO [train.py:904] (2/8) Epoch 30, batch 4150, loss[loss=0.1815, simple_loss=0.269, pruned_loss=0.04699, over 16562.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2585, pruned_loss=0.04496, over 3246459.24 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:09:56,054 INFO [optim.py:368] (2/8) 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 22:10:35,420 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:10:55,714 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298543.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:11:12,253 INFO [train.py:904] (2/8) Epoch 30, batch 4200, loss[loss=0.1897, simple_loss=0.2834, pruned_loss=0.04801, over 16706.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2648, pruned_loss=0.04609, over 3228753.38 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:08,277 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:12:22,602 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6501, 3.9117, 2.9462, 2.3607, 2.5924, 2.7106, 4.1588, 3.4087], device='cuda:2'), covar=tensor([0.3075, 0.0696, 0.1965, 0.3051, 0.2898, 0.1996, 0.0564, 0.1387], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0277, 0.0316, 0.0330, 0.0311, 0.0281, 0.0307, 0.0356], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 22:12:26,984 INFO [train.py:904] (2/8) Epoch 30, batch 4250, loss[loss=0.1822, simple_loss=0.2598, pruned_loss=0.05236, over 12185.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.268, pruned_loss=0.04592, over 3188428.51 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:28,280 INFO [optim.py:368] (2/8) 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 22:12:40,792 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=298613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:13:04,659 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298629.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:13:29,353 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-02 22:13:41,911 INFO [train.py:904] (2/8) Epoch 30, batch 4300, loss[loss=0.1719, simple_loss=0.2695, pruned_loss=0.03715, over 16869.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2691, pruned_loss=0.04488, over 3185181.18 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:13:52,426 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=298661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:35,037 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:36,367 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298690.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:14:41,147 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9021, 4.9049, 4.6299, 3.2127, 4.0092, 4.7110, 3.9900, 2.8660], device='cuda:2'), covar=tensor([0.0554, 0.0023, 0.0055, 0.0412, 0.0113, 0.0077, 0.0109, 0.0406], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0138, 0.0106, 0.0118, 0.0101, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 22:14:56,148 INFO [train.py:904] (2/8) Epoch 30, batch 4350, loss[loss=0.1771, simple_loss=0.2717, pruned_loss=0.04121, over 16757.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2721, pruned_loss=0.04555, over 3185675.65 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:14:57,381 INFO [optim.py:368] (2/8) 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 22:15:15,706 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0150, 3.2743, 3.1708, 2.1050, 2.9946, 3.2468, 3.0291, 1.9187], device='cuda:2'), covar=tensor([0.0597, 0.0064, 0.0088, 0.0493, 0.0135, 0.0116, 0.0132, 0.0525], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0106, 0.0118, 0.0101, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 22:16:03,927 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298750.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:16:09,615 INFO [train.py:904] (2/8) Epoch 30, batch 4400, loss[loss=0.1887, simple_loss=0.2812, pruned_loss=0.04808, over 16243.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2741, pruned_loss=0.0466, over 3166188.11 frames. ], batch size: 35, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:16:22,226 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 22:17:21,872 INFO [train.py:904] (2/8) Epoch 30, batch 4450, loss[loss=0.1838, simple_loss=0.2731, pruned_loss=0.04731, over 16372.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2776, pruned_loss=0.04783, over 3186560.06 frames. ], batch size: 35, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:23,559 INFO [optim.py:368] (2/8) 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 22:18:01,227 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 22:18:07,604 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1124, 5.1010, 4.8524, 4.1664, 5.0632, 1.9196, 4.7822, 4.3479], device='cuda:2'), covar=tensor([0.0064, 0.0058, 0.0154, 0.0324, 0.0058, 0.3056, 0.0093, 0.0282], device='cuda:2'), in_proj_covar=tensor([0.0188, 0.0180, 0.0219, 0.0190, 0.0197, 0.0223, 0.0208, 0.0186], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 22:18:09,765 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298838.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:18:32,505 INFO [train.py:904] (2/8) Epoch 30, batch 4500, loss[loss=0.1962, simple_loss=0.2838, pruned_loss=0.05426, over 16591.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2783, pruned_loss=0.04873, over 3185349.15 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:15,884 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 22:19:19,727 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:19:29,751 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.1939, 2.1687, 2.9219, 3.1579, 3.0919, 3.6402, 2.4198, 3.6826], device='cuda:2'), covar=tensor([0.0216, 0.0560, 0.0315, 0.0300, 0.0313, 0.0196, 0.0565, 0.0142], device='cuda:2'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0198, 0.0216, 0.0172, 0.0208, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 22:19:33,016 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3464, 2.2994, 3.0951, 3.2822, 3.2955, 3.8939, 2.5779, 3.8775], device='cuda:2'), covar=tensor([0.0221, 0.0557, 0.0306, 0.0289, 0.0279, 0.0140, 0.0569, 0.0126], device='cuda:2'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0198, 0.0216, 0.0172, 0.0208, 0.0173], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 22:19:44,877 INFO [train.py:904] (2/8) Epoch 30, batch 4550, loss[loss=0.2054, simple_loss=0.2807, pruned_loss=0.06506, over 11744.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.279, pruned_loss=0.04989, over 3191527.36 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:46,102 INFO [optim.py:368] (2/8) 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 22:20:27,773 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3715, 2.1574, 1.8080, 1.9561, 2.4614, 2.0931, 2.1235, 2.5115], device='cuda:2'), covar=tensor([0.0229, 0.0445, 0.0564, 0.0499, 0.0278, 0.0371, 0.0237, 0.0284], device='cuda:2'), in_proj_covar=tensor([0.0237, 0.0248, 0.0237, 0.0238, 0.0248, 0.0246, 0.0247, 0.0248], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 22:20:57,286 INFO [train.py:904] (2/8) Epoch 30, batch 4600, loss[loss=0.1823, simple_loss=0.2758, pruned_loss=0.04441, over 16495.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2797, pruned_loss=0.05019, over 3187319.81 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:21:09,745 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-02 22:21:42,885 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298985.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:22:09,581 INFO [train.py:904] (2/8) Epoch 30, batch 4650, loss[loss=0.2035, simple_loss=0.2774, pruned_loss=0.06475, over 11743.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2793, pruned_loss=0.05071, over 3188576.75 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:22:10,879 INFO [optim.py:368] (2/8) 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 22:22:41,792 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8085, 5.0698, 5.2535, 4.8961, 5.0469, 5.6040, 4.9943, 4.6989], device='cuda:2'), covar=tensor([0.1107, 0.1867, 0.1922, 0.2144, 0.2445, 0.0859, 0.1512, 0.2523], device='cuda:2'), in_proj_covar=tensor([0.0433, 0.0638, 0.0710, 0.0520, 0.0692, 0.0731, 0.0548, 0.0691], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 22:23:09,773 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299045.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:23:11,075 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5481, 2.3705, 2.4120, 4.3192, 2.9726, 3.8425, 1.5394, 2.8493], device='cuda:2'), covar=tensor([0.1759, 0.1142, 0.1526, 0.0225, 0.0300, 0.0430, 0.2143, 0.0971], device='cuda:2'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0209, 0.0209, 0.0220, 0.0213, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 22:23:22,504 INFO [train.py:904] (2/8) Epoch 30, batch 4700, loss[loss=0.1574, simple_loss=0.2523, pruned_loss=0.03123, over 16832.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2766, pruned_loss=0.0494, over 3191405.56 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:23:25,666 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 22:23:39,699 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4613, 3.5367, 2.2012, 4.0668, 2.7507, 3.9668, 2.3004, 2.8079], device='cuda:2'), covar=tensor([0.0378, 0.0476, 0.1888, 0.0174, 0.0937, 0.0617, 0.1751, 0.0974], device='cuda:2'), in_proj_covar=tensor([0.0179, 0.0184, 0.0197, 0.0176, 0.0182, 0.0224, 0.0206, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 22:24:36,507 INFO [train.py:904] (2/8) Epoch 30, batch 4750, loss[loss=0.1724, simple_loss=0.2663, pruned_loss=0.0392, over 16836.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2724, pruned_loss=0.0471, over 3198390.81 frames. ], batch size: 39, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:24:37,717 INFO [optim.py:368] (2/8) 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 22:25:05,779 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-05-02 22:25:07,724 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:25,747 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:25,904 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9817, 2.1750, 2.6792, 2.9256, 2.8931, 3.4423, 2.2793, 3.3549], device='cuda:2'), covar=tensor([0.0270, 0.0609, 0.0388, 0.0409, 0.0381, 0.0197, 0.0664, 0.0214], device='cuda:2'), in_proj_covar=tensor([0.0202, 0.0202, 0.0190, 0.0197, 0.0215, 0.0171, 0.0207, 0.0172], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 22:25:49,328 INFO [train.py:904] (2/8) Epoch 30, batch 4800, loss[loss=0.1895, simple_loss=0.2799, pruned_loss=0.04951, over 16917.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2686, pruned_loss=0.04552, over 3185753.21 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:26:37,232 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:37,310 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:38,485 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299187.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:45,052 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299191.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:49,636 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1691, 3.7245, 3.7445, 2.4365, 3.3312, 3.7570, 3.4443, 2.0255], device='cuda:2'), covar=tensor([0.0637, 0.0073, 0.0066, 0.0470, 0.0123, 0.0145, 0.0131, 0.0562], device='cuda:2'), in_proj_covar=tensor([0.0140, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 22:27:04,357 INFO [train.py:904] (2/8) Epoch 30, batch 4850, loss[loss=0.1897, simple_loss=0.2762, pruned_loss=0.05166, over 16531.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2699, pruned_loss=0.04507, over 3161298.38 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:27:06,457 INFO [optim.py:368] (2/8) 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 22:27:49,336 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:02,992 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299243.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:16,414 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299252.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:18,993 INFO [train.py:904] (2/8) Epoch 30, batch 4900, loss[loss=0.1765, simple_loss=0.2598, pruned_loss=0.04662, over 12116.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2688, pruned_loss=0.04364, over 3149678.34 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:28:20,371 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 22:28:21,399 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 22:29:00,292 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-05-02 22:29:05,977 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299285.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:29:10,813 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.5509, 2.5038, 2.3513, 3.8122, 2.3322, 3.7633, 1.4647, 2.7507], device='cuda:2'), covar=tensor([0.1553, 0.0908, 0.1434, 0.0170, 0.0139, 0.0356, 0.1975, 0.0894], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0207, 0.0207, 0.0218, 0.0211, 0.0201], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 22:29:33,775 INFO [train.py:904] (2/8) Epoch 30, batch 4950, loss[loss=0.1994, simple_loss=0.2922, pruned_loss=0.05329, over 16884.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2686, pruned_loss=0.04296, over 3161624.70 frames. ], batch size: 116, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:29:34,228 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299304.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:29:34,847 INFO [optim.py:368] (2/8) 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 22:30:02,255 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:30:17,898 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299333.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:30:36,115 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:30:47,899 INFO [train.py:904] (2/8) Epoch 30, batch 5000, loss[loss=0.1639, simple_loss=0.2618, pruned_loss=0.03302, over 16829.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2697, pruned_loss=0.04252, over 3178892.77 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:30:51,138 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 22:31:33,612 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:31:46,453 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:32:01,705 INFO [train.py:904] (2/8) Epoch 30, batch 5050, loss[loss=0.1966, simple_loss=0.2778, pruned_loss=0.05776, over 12050.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2705, pruned_loss=0.04274, over 3172498.26 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:32:02,875 INFO [optim.py:368] (2/8) 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 22:32:06,439 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0387, 2.1000, 2.6985, 3.0496, 2.9528, 3.5673, 2.2689, 3.5413], device='cuda:2'), covar=tensor([0.0265, 0.0581, 0.0385, 0.0339, 0.0354, 0.0176, 0.0645, 0.0148], device='cuda:2'), in_proj_covar=tensor([0.0201, 0.0200, 0.0189, 0.0196, 0.0214, 0.0170, 0.0207, 0.0171], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-02 22:33:13,772 INFO [train.py:904] (2/8) Epoch 30, batch 5100, loss[loss=0.1732, simple_loss=0.2693, pruned_loss=0.03853, over 15295.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2692, pruned_loss=0.04229, over 3174595.84 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:33:17,251 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9088, 3.2716, 3.2020, 2.0895, 2.9936, 3.2323, 3.0493, 1.8342], device='cuda:2'), covar=tensor([0.0654, 0.0073, 0.0082, 0.0497, 0.0127, 0.0143, 0.0138, 0.0563], device='cuda:2'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 22:33:58,041 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:34:30,232 INFO [train.py:904] (2/8) Epoch 30, batch 5150, loss[loss=0.1786, simple_loss=0.2804, pruned_loss=0.03838, over 16767.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2697, pruned_loss=0.04178, over 3168131.08 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:34:31,338 INFO [optim.py:368] (2/8) 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 22:34:34,023 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 22:34:57,232 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0358, 3.5258, 3.5277, 2.2736, 3.2430, 3.5022, 3.2398, 1.9243], device='cuda:2'), covar=tensor([0.0675, 0.0066, 0.0072, 0.0493, 0.0122, 0.0149, 0.0134, 0.0575], device='cuda:2'), in_proj_covar=tensor([0.0142, 0.0093, 0.0095, 0.0138, 0.0106, 0.0119, 0.0102, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-02 22:35:33,202 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:35:43,065 INFO [train.py:904] (2/8) Epoch 30, batch 5200, loss[loss=0.196, simple_loss=0.2751, pruned_loss=0.05849, over 16693.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2677, pruned_loss=0.0413, over 3171612.02 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:35:58,598 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0715, 4.0294, 3.9646, 3.1200, 3.9609, 1.7096, 3.7759, 3.4810], device='cuda:2'), covar=tensor([0.0127, 0.0137, 0.0194, 0.0370, 0.0119, 0.3170, 0.0158, 0.0333], device='cuda:2'), in_proj_covar=tensor([0.0187, 0.0178, 0.0217, 0.0188, 0.0195, 0.0221, 0.0205, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 22:36:48,493 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299599.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:36:55,727 INFO [train.py:904] (2/8) Epoch 30, batch 5250, loss[loss=0.1586, simple_loss=0.2578, pruned_loss=0.02969, over 16367.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.265, pruned_loss=0.0407, over 3183563.51 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:36:56,956 INFO [optim.py:368] (2/8) 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 22:37:18,280 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-05-02 22:38:06,895 INFO [train.py:904] (2/8) Epoch 30, batch 5300, loss[loss=0.1531, simple_loss=0.2367, pruned_loss=0.03471, over 16644.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2608, pruned_loss=0.03926, over 3195253.43 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:38:42,825 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:38:57,924 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0592, 2.3057, 2.2075, 3.7340, 2.1640, 2.6709, 2.3791, 2.4517], device='cuda:2'), covar=tensor([0.1590, 0.3806, 0.3347, 0.0665, 0.4291, 0.2549, 0.3928, 0.3257], device='cuda:2'), in_proj_covar=tensor([0.0428, 0.0482, 0.0391, 0.0344, 0.0448, 0.0554, 0.0455, 0.0565], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 22:39:12,305 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.8937, 2.5602, 2.3029, 3.8596, 2.2199, 3.7642, 1.6574, 2.7913], device='cuda:2'), covar=tensor([0.1362, 0.0936, 0.1451, 0.0185, 0.0172, 0.0429, 0.1810, 0.0890], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0208, 0.0208, 0.0219, 0.0213, 0.0202], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 22:39:18,175 INFO [train.py:904] (2/8) Epoch 30, batch 5350, loss[loss=0.1689, simple_loss=0.2628, pruned_loss=0.03746, over 16877.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2594, pruned_loss=0.03881, over 3200625.04 frames. ], batch size: 116, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:39:19,344 INFO [optim.py:368] (2/8) 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 22:40:07,308 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 22:40:29,714 INFO [train.py:904] (2/8) Epoch 30, batch 5400, loss[loss=0.2055, simple_loss=0.2926, pruned_loss=0.05914, over 11907.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2623, pruned_loss=0.04008, over 3183913.87 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:40:51,607 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299769.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:10,081 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:34,563 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299798.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:41:46,646 INFO [train.py:904] (2/8) Epoch 30, batch 5450, loss[loss=0.2028, simple_loss=0.2909, pruned_loss=0.05731, over 16633.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2648, pruned_loss=0.04108, over 3193455.48 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:41:47,795 INFO [optim.py:368] (2/8) 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 22:42:27,030 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:27,229 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:45,614 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.4276, 4.4713, 4.7528, 4.7099, 4.7884, 4.4647, 4.4499, 4.3688], device='cuda:2'), covar=tensor([0.0353, 0.0554, 0.0445, 0.0463, 0.0428, 0.0453, 0.0965, 0.0516], device='cuda:2'), in_proj_covar=tensor([0.0439, 0.0502, 0.0480, 0.0445, 0.0525, 0.0508, 0.0585, 0.0410], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 22:42:51,977 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299847.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:43:02,907 INFO [train.py:904] (2/8) Epoch 30, batch 5500, loss[loss=0.1901, simple_loss=0.2852, pruned_loss=0.04752, over 16831.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2718, pruned_loss=0.04491, over 3169615.65 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:43:11,033 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299859.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:44:07,184 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4587, 3.0460, 2.6220, 2.2691, 2.3383, 2.2557, 3.0746, 2.9128], device='cuda:2'), covar=tensor([0.2744, 0.0683, 0.1759, 0.2481, 0.2621, 0.2410, 0.0564, 0.1278], device='cuda:2'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0330, 0.0307, 0.0281, 0.0305, 0.0353], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 22:44:09,163 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299895.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:14,135 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:21,842 INFO [train.py:904] (2/8) Epoch 30, batch 5550, loss[loss=0.1996, simple_loss=0.2899, pruned_loss=0.0546, over 15451.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2778, pruned_loss=0.04887, over 3153739.12 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:44:23,799 INFO [optim.py:368] (2/8) 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 22:44:31,946 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 22:45:31,925 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299947.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:45:43,005 INFO [train.py:904] (2/8) Epoch 30, batch 5600, loss[loss=0.2772, simple_loss=0.3387, pruned_loss=0.1079, over 10939.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2823, pruned_loss=0.05267, over 3118004.72 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:46:24,525 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4014, 3.3851, 3.4196, 3.5183, 3.5332, 3.2923, 3.5105, 3.5762], device='cuda:2'), covar=tensor([0.1233, 0.0994, 0.1107, 0.0725, 0.0748, 0.2341, 0.1257, 0.0924], device='cuda:2'), in_proj_covar=tensor([0.0679, 0.0827, 0.0957, 0.0846, 0.0642, 0.0668, 0.0701, 0.0817], device='cuda:2'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 22:46:27,193 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:46:47,548 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-02 22:47:11,368 INFO [train.py:904] (2/8) Epoch 30, batch 5650, loss[loss=0.2379, simple_loss=0.3126, pruned_loss=0.08156, over 11588.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2876, pruned_loss=0.05715, over 3063243.39 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:47:13,271 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:48:27,593 INFO [train.py:904] (2/8) Epoch 30, batch 5700, loss[loss=0.2678, simple_loss=0.3348, pruned_loss=0.1004, over 11248.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.289, pruned_loss=0.05882, over 3052647.35 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:48:57,316 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:49:45,341 INFO [train.py:904] (2/8) Epoch 30, batch 5750, loss[loss=0.1985, simple_loss=0.2955, pruned_loss=0.0508, over 16780.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2921, pruned_loss=0.06053, over 3051916.10 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:49:49,226 INFO [optim.py:368] (2/8) 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 22:50:18,716 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300125.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:50:34,995 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:51:06,734 INFO [train.py:904] (2/8) Epoch 30, batch 5800, loss[loss=0.1676, simple_loss=0.2595, pruned_loss=0.03789, over 16252.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2914, pruned_loss=0.05891, over 3069330.84 frames. ], batch size: 35, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:51:07,776 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300154.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:52:25,619 INFO [train.py:904] (2/8) Epoch 30, batch 5850, loss[loss=0.2076, simple_loss=0.2996, pruned_loss=0.05773, over 16212.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2891, pruned_loss=0.05725, over 3088972.12 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:52:28,950 INFO [optim.py:368] (2/8) 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 22:53:48,214 INFO [train.py:904] (2/8) Epoch 30, batch 5900, loss[loss=0.1833, simple_loss=0.2799, pruned_loss=0.04329, over 16509.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2897, pruned_loss=0.05734, over 3098573.26 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:54:19,145 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 22:54:38,406 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4542, 2.0765, 1.7872, 1.9172, 2.4111, 2.1295, 2.0478, 2.5362], device='cuda:2'), covar=tensor([0.0267, 0.0513, 0.0612, 0.0538, 0.0297, 0.0437, 0.0237, 0.0337], device='cuda:2'), in_proj_covar=tensor([0.0234, 0.0244, 0.0234, 0.0235, 0.0245, 0.0242, 0.0242, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 22:55:09,235 INFO [train.py:904] (2/8) Epoch 30, batch 5950, loss[loss=0.204, simple_loss=0.2855, pruned_loss=0.06125, over 15350.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2898, pruned_loss=0.05571, over 3102562.60 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:55:12,894 INFO [optim.py:368] (2/8) 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 22:56:23,971 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 22:56:29,059 INFO [train.py:904] (2/8) Epoch 30, batch 6000, loss[loss=0.198, simple_loss=0.2776, pruned_loss=0.05919, over 15277.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2884, pruned_loss=0.05497, over 3119463.70 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:56:29,059 INFO [train.py:929] (2/8) Computing validation loss 2023-05-02 22:56:39,785 INFO [train.py:938] (2/8) Epoch 30, validation: loss=0.1471, simple_loss=0.2591, pruned_loss=0.01755, over 944034.00 frames. 2023-05-02 22:56:39,786 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-02 22:57:57,800 INFO [train.py:904] (2/8) Epoch 30, batch 6050, loss[loss=0.2015, simple_loss=0.2949, pruned_loss=0.05402, over 16927.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2869, pruned_loss=0.05418, over 3124602.33 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:58:01,296 INFO [optim.py:368] (2/8) 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 22:58:29,367 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:58:35,425 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:59:15,657 INFO [train.py:904] (2/8) Epoch 30, batch 6100, loss[loss=0.1771, simple_loss=0.275, pruned_loss=0.03956, over 16798.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2864, pruned_loss=0.05309, over 3143704.24 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:59:16,322 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300454.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:59:42,191 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2954, 3.0196, 3.3084, 1.8088, 3.4232, 3.4423, 2.8083, 2.6666], device='cuda:2'), covar=tensor([0.0862, 0.0317, 0.0219, 0.1270, 0.0113, 0.0225, 0.0496, 0.0538], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 22:59:47,465 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:59:51,339 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1778, 5.6765, 5.8466, 5.5384, 5.6245, 6.1623, 5.6527, 5.4323], device='cuda:2'), covar=tensor([0.0866, 0.1661, 0.2419, 0.1762, 0.2195, 0.0801, 0.1551, 0.2316], device='cuda:2'), in_proj_covar=tensor([0.0433, 0.0642, 0.0717, 0.0520, 0.0693, 0.0732, 0.0551, 0.0692], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 23:00:32,802 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300502.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 23:00:34,966 INFO [train.py:904] (2/8) Epoch 30, batch 6150, loss[loss=0.1811, simple_loss=0.2752, pruned_loss=0.04352, over 15342.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2849, pruned_loss=0.05291, over 3138204.00 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:00:38,227 INFO [optim.py:368] (2/8) 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 23:01:53,689 INFO [train.py:904] (2/8) Epoch 30, batch 6200, loss[loss=0.1647, simple_loss=0.2635, pruned_loss=0.03298, over 16871.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2837, pruned_loss=0.05305, over 3114389.16 frames. ], batch size: 96, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:02:24,923 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8337, 3.6553, 4.2817, 1.9982, 4.4700, 4.4636, 3.2629, 3.2634], device='cuda:2'), covar=tensor([0.0796, 0.0324, 0.0203, 0.1331, 0.0071, 0.0168, 0.0442, 0.0499], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 23:03:13,162 INFO [train.py:904] (2/8) Epoch 30, batch 6250, loss[loss=0.2496, simple_loss=0.3114, pruned_loss=0.09386, over 11760.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2834, pruned_loss=0.05272, over 3117928.85 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:16,361 INFO [optim.py:368] (2/8) 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 23:03:21,834 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2732, 3.3453, 2.0716, 3.5871, 2.5788, 3.5966, 2.1894, 2.6767], device='cuda:2'), covar=tensor([0.0333, 0.0437, 0.1746, 0.0275, 0.0884, 0.0649, 0.1708, 0.0931], device='cuda:2'), in_proj_covar=tensor([0.0178, 0.0184, 0.0197, 0.0175, 0.0181, 0.0223, 0.0206, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 23:03:34,841 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:03:53,401 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3765, 4.4486, 4.6739, 4.6543, 4.6823, 4.4415, 4.3965, 4.3425], device='cuda:2'), covar=tensor([0.0346, 0.0605, 0.0473, 0.0445, 0.0467, 0.0473, 0.0844, 0.0549], device='cuda:2'), in_proj_covar=tensor([0.0441, 0.0503, 0.0481, 0.0447, 0.0527, 0.0509, 0.0585, 0.0412], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 23:04:31,361 INFO [train.py:904] (2/8) Epoch 30, batch 6300, loss[loss=0.2219, simple_loss=0.3192, pruned_loss=0.06233, over 16706.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2832, pruned_loss=0.05228, over 3120622.60 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:07,625 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1936, 2.4377, 2.5457, 1.9554, 2.6997, 2.7364, 2.4764, 2.3563], device='cuda:2'), covar=tensor([0.0639, 0.0279, 0.0267, 0.0931, 0.0143, 0.0283, 0.0438, 0.0458], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0088, 0.0135, 0.0132, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 23:05:10,947 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300678.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:05:50,440 INFO [train.py:904] (2/8) Epoch 30, batch 6350, loss[loss=0.188, simple_loss=0.2855, pruned_loss=0.04522, over 16804.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2844, pruned_loss=0.05378, over 3106084.48 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:53,979 INFO [optim.py:368] (2/8) 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 23:06:21,474 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 23:06:29,151 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:07:07,467 INFO [train.py:904] (2/8) Epoch 30, batch 6400, loss[loss=0.1565, simple_loss=0.2494, pruned_loss=0.0318, over 16872.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2857, pruned_loss=0.05585, over 3066539.46 frames. ], batch size: 90, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:07:42,791 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300777.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:08:04,681 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 23:08:19,382 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:08:21,338 INFO [train.py:904] (2/8) Epoch 30, batch 6450, loss[loss=0.1873, simple_loss=0.2819, pruned_loss=0.04632, over 16674.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2851, pruned_loss=0.05461, over 3087674.08 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:08:24,279 INFO [optim.py:368] (2/8) 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 23:08:39,269 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 23:09:38,121 INFO [train.py:904] (2/8) Epoch 30, batch 6500, loss[loss=0.2052, simple_loss=0.293, pruned_loss=0.05874, over 16733.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.283, pruned_loss=0.05424, over 3083937.68 frames. ], batch size: 76, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:09:52,525 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300863.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:09:53,731 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6903, 3.8896, 2.8453, 2.3193, 2.5574, 2.4874, 4.1848, 3.4135], device='cuda:2'), covar=tensor([0.3103, 0.0669, 0.1964, 0.2908, 0.2810, 0.2229, 0.0484, 0.1324], device='cuda:2'), in_proj_covar=tensor([0.0337, 0.0275, 0.0314, 0.0329, 0.0306, 0.0280, 0.0305, 0.0352], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 23:10:55,636 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.0144, 5.4815, 5.6104, 5.3383, 5.4430, 5.9998, 5.4924, 5.2206], device='cuda:2'), covar=tensor([0.0936, 0.1733, 0.2603, 0.1975, 0.2313, 0.0854, 0.1506, 0.2357], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0643, 0.0720, 0.0525, 0.0695, 0.0736, 0.0554, 0.0699], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 23:10:58,875 INFO [train.py:904] (2/8) Epoch 30, batch 6550, loss[loss=0.2646, simple_loss=0.3277, pruned_loss=0.1007, over 11653.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.286, pruned_loss=0.05608, over 3066353.45 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:11:01,652 INFO [optim.py:368] (2/8) 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 23:12:19,421 INFO [train.py:904] (2/8) Epoch 30, batch 6600, loss[loss=0.2489, simple_loss=0.3188, pruned_loss=0.08953, over 11674.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.288, pruned_loss=0.05659, over 3064823.35 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:12:48,783 INFO [zipformer.py:625] (2/8) 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] (2/8) Epoch 30, batch 6650, loss[loss=0.194, simple_loss=0.2829, pruned_loss=0.05254, over 16254.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2882, pruned_loss=0.05691, over 3072779.08 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:13:43,931 INFO [optim.py:368] (2/8) 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 23:14:55,111 INFO [train.py:904] (2/8) Epoch 30, batch 6700, loss[loss=0.1953, simple_loss=0.2804, pruned_loss=0.05508, over 16528.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2868, pruned_loss=0.05708, over 3059548.13 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:11,407 INFO [train.py:904] (2/8) Epoch 30, batch 6750, loss[loss=0.1915, simple_loss=0.282, pruned_loss=0.05053, over 15292.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2855, pruned_loss=0.05645, over 3072523.06 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:15,728 INFO [optim.py:368] (2/8) 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 23:16:35,390 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 23:17:07,597 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:17:28,471 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 23:17:29,020 INFO [train.py:904] (2/8) Epoch 30, batch 6800, loss[loss=0.1997, simple_loss=0.2891, pruned_loss=0.05514, over 16594.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2853, pruned_loss=0.05613, over 3085590.80 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:17:35,474 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301158.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:18:18,027 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.0771, 3.1583, 2.8343, 2.9371, 3.4881, 3.1004, 3.5609, 3.6528], device='cuda:2'), covar=tensor([0.0105, 0.0423, 0.0496, 0.0428, 0.0265, 0.0375, 0.0277, 0.0271], device='cuda:2'), in_proj_covar=tensor([0.0231, 0.0242, 0.0232, 0.0233, 0.0244, 0.0240, 0.0239, 0.0242], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:18:43,924 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:18:47,466 INFO [train.py:904] (2/8) Epoch 30, batch 6850, loss[loss=0.1933, simple_loss=0.2967, pruned_loss=0.0449, over 16492.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2863, pruned_loss=0.05585, over 3116674.15 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:18:51,707 INFO [optim.py:368] (2/8) 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 23:19:29,035 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2671, 2.5805, 2.2491, 2.2489, 2.9570, 2.5549, 2.8430, 3.1279], device='cuda:2'), covar=tensor([0.0206, 0.0562, 0.0670, 0.0611, 0.0332, 0.0473, 0.0336, 0.0372], device='cuda:2'), in_proj_covar=tensor([0.0233, 0.0243, 0.0234, 0.0234, 0.0245, 0.0241, 0.0241, 0.0243], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:20:04,219 INFO [train.py:904] (2/8) Epoch 30, batch 6900, loss[loss=0.1975, simple_loss=0.2889, pruned_loss=0.05308, over 16870.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2877, pruned_loss=0.0543, over 3142888.73 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:20:35,499 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:21:22,699 INFO [train.py:904] (2/8) Epoch 30, batch 6950, loss[loss=0.2005, simple_loss=0.2818, pruned_loss=0.05955, over 15264.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2886, pruned_loss=0.05524, over 3128508.43 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:21:26,934 INFO [optim.py:368] (2/8) 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 23:21:50,650 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:22:38,931 INFO [train.py:904] (2/8) Epoch 30, batch 7000, loss[loss=0.1967, simple_loss=0.3042, pruned_loss=0.04456, over 16696.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.289, pruned_loss=0.05491, over 3108405.43 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:23:55,570 INFO [train.py:904] (2/8) Epoch 30, batch 7050, loss[loss=0.2039, simple_loss=0.2966, pruned_loss=0.05555, over 15304.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2893, pruned_loss=0.05466, over 3103306.73 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:23:58,814 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5340, 3.5122, 3.5080, 2.6820, 3.3538, 2.1471, 3.1850, 2.8153], device='cuda:2'), covar=tensor([0.0177, 0.0147, 0.0202, 0.0246, 0.0122, 0.2473, 0.0157, 0.0325], device='cuda:2'), in_proj_covar=tensor([0.0186, 0.0176, 0.0216, 0.0188, 0.0193, 0.0220, 0.0204, 0.0181], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:24:00,750 INFO [optim.py:368] (2/8) 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 23:24:38,118 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:25:14,075 INFO [train.py:904] (2/8) Epoch 30, batch 7100, loss[loss=0.1758, simple_loss=0.2664, pruned_loss=0.04258, over 16795.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2878, pruned_loss=0.05466, over 3084707.03 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:25:20,947 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301458.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:14,363 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301491.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:21,069 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:33,761 INFO [train.py:904] (2/8) Epoch 30, batch 7150, loss[loss=0.2126, simple_loss=0.3004, pruned_loss=0.06243, over 16750.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2867, pruned_loss=0.05493, over 3099980.62 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:26:36,531 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301506.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:37,391 INFO [optim.py:368] (2/8) 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 23:26:48,749 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1088, 2.4221, 2.5322, 1.9683, 2.6899, 2.7513, 2.3956, 2.3479], device='cuda:2'), covar=tensor([0.0704, 0.0301, 0.0277, 0.0922, 0.0155, 0.0326, 0.0500, 0.0461], device='cuda:2'), in_proj_covar=tensor([0.0151, 0.0112, 0.0104, 0.0140, 0.0089, 0.0135, 0.0133, 0.0133], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 23:26:51,303 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-05-02 23:27:38,409 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([5.1011, 5.6296, 5.8101, 5.5002, 5.6162, 6.1315, 5.5970, 5.3581], device='cuda:2'), covar=tensor([0.0888, 0.1632, 0.2045, 0.1851, 0.2082, 0.0865, 0.1449, 0.2185], device='cuda:2'), in_proj_covar=tensor([0.0434, 0.0644, 0.0718, 0.0526, 0.0696, 0.0735, 0.0555, 0.0701], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 23:27:49,037 INFO [train.py:904] (2/8) Epoch 30, batch 7200, loss[loss=0.1743, simple_loss=0.2676, pruned_loss=0.04048, over 16666.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2852, pruned_loss=0.05407, over 3088422.92 frames. ], batch size: 62, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:10,980 INFO [train.py:904] (2/8) Epoch 30, batch 7250, loss[loss=0.1904, simple_loss=0.2747, pruned_loss=0.05303, over 16185.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2824, pruned_loss=0.05285, over 3070707.90 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:15,161 INFO [optim.py:368] (2/8) 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 23:29:39,690 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5295, 4.5991, 4.9104, 4.8771, 4.9020, 4.6091, 4.5707, 4.4742], device='cuda:2'), covar=tensor([0.0386, 0.0574, 0.0411, 0.0413, 0.0472, 0.0452, 0.0974, 0.0559], device='cuda:2'), in_proj_covar=tensor([0.0443, 0.0507, 0.0484, 0.0448, 0.0528, 0.0513, 0.0589, 0.0412], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 23:30:27,540 INFO [train.py:904] (2/8) Epoch 30, batch 7300, loss[loss=0.1826, simple_loss=0.2772, pruned_loss=0.04396, over 17219.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2816, pruned_loss=0.05256, over 3080170.69 frames. ], batch size: 44, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:31:37,113 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2606, 3.3111, 2.0964, 3.7259, 2.5493, 3.7241, 2.1767, 2.6841], device='cuda:2'), covar=tensor([0.0346, 0.0453, 0.1783, 0.0196, 0.0933, 0.0514, 0.1732, 0.0893], device='cuda:2'), in_proj_covar=tensor([0.0177, 0.0182, 0.0196, 0.0173, 0.0180, 0.0221, 0.0204, 0.0184], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 23:31:44,861 INFO [train.py:904] (2/8) Epoch 30, batch 7350, loss[loss=0.2057, simple_loss=0.2867, pruned_loss=0.06236, over 16721.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2834, pruned_loss=0.05414, over 3051163.17 frames. ], batch size: 76, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:31:50,961 INFO [optim.py:368] (2/8) 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 23:32:44,237 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301743.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:33:00,865 INFO [train.py:904] (2/8) Epoch 30, batch 7400, loss[loss=0.1919, simple_loss=0.2886, pruned_loss=0.04763, over 16765.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2842, pruned_loss=0.05443, over 3066707.79 frames. ], batch size: 39, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:33:53,667 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:10,907 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:23,455 INFO [train.py:904] (2/8) Epoch 30, batch 7450, loss[loss=0.1957, simple_loss=0.2934, pruned_loss=0.04902, over 16170.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2853, pruned_loss=0.05514, over 3085745.16 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:34:24,242 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:29,849 INFO [optim.py:368] (2/8) 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 23:34:43,545 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3630, 3.3228, 3.7064, 1.8074, 3.8555, 3.9141, 3.0278, 2.8981], device='cuda:2'), covar=tensor([0.0885, 0.0303, 0.0244, 0.1358, 0.0104, 0.0222, 0.0455, 0.0503], device='cuda:2'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-02 23:34:54,598 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3373, 2.9338, 2.6651, 2.2440, 2.2209, 2.2817, 2.9707, 2.8078], device='cuda:2'), covar=tensor([0.2760, 0.0751, 0.1722, 0.2684, 0.2427, 0.2360, 0.0507, 0.1486], device='cuda:2'), in_proj_covar=tensor([0.0338, 0.0276, 0.0316, 0.0330, 0.0307, 0.0281, 0.0305, 0.0353], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:2') 2023-05-02 23:35:30,252 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:35:45,203 INFO [train.py:904] (2/8) Epoch 30, batch 7500, loss[loss=0.1873, simple_loss=0.2743, pruned_loss=0.05017, over 16900.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2857, pruned_loss=0.05513, over 3054296.29 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:36:45,647 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8093, 5.0606, 4.8407, 4.8445, 4.6120, 4.5525, 4.4988, 5.1570], device='cuda:2'), covar=tensor([0.1180, 0.0897, 0.1075, 0.0921, 0.0796, 0.1171, 0.1218, 0.0839], device='cuda:2'), in_proj_covar=tensor([0.0717, 0.0864, 0.0712, 0.0675, 0.0549, 0.0552, 0.0725, 0.0678], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:36:52,215 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.2369, 2.5023, 2.5454, 4.0332, 2.4385, 2.8720, 2.5407, 2.6747], device='cuda:2'), covar=tensor([0.1448, 0.3326, 0.2842, 0.0591, 0.3863, 0.2317, 0.3347, 0.2994], device='cuda:2'), in_proj_covar=tensor([0.0426, 0.0479, 0.0390, 0.0340, 0.0449, 0.0551, 0.0453, 0.0562], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:36:57,901 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:05,267 INFO [train.py:904] (2/8) Epoch 30, batch 7550, loss[loss=0.1905, simple_loss=0.2757, pruned_loss=0.05268, over 16684.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2848, pruned_loss=0.05532, over 3044483.38 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:37:07,777 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301905.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:11,523 INFO [optim.py:368] (2/8) 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 23:37:37,735 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3918, 4.5509, 4.7286, 4.4748, 4.5721, 5.0659, 4.5645, 4.3373], device='cuda:2'), covar=tensor([0.1540, 0.1995, 0.2322, 0.1899, 0.2285, 0.0987, 0.1964, 0.2615], device='cuda:2'), in_proj_covar=tensor([0.0431, 0.0640, 0.0715, 0.0522, 0.0692, 0.0733, 0.0554, 0.0697], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-02 23:38:23,955 INFO [train.py:904] (2/8) Epoch 30, batch 7600, loss[loss=0.2103, simple_loss=0.2917, pruned_loss=0.0645, over 15193.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2843, pruned_loss=0.05531, over 3074366.06 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:38:34,144 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301960.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:38:43,755 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:38:52,163 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.2115, 4.4583, 4.1355, 3.9658, 3.5538, 4.4007, 4.0665, 4.0224], device='cuda:2'), covar=tensor([0.0968, 0.0702, 0.0540, 0.0455, 0.1627, 0.0588, 0.0931, 0.0820], device='cuda:2'), in_proj_covar=tensor([0.0311, 0.0472, 0.0363, 0.0365, 0.0360, 0.0421, 0.0251, 0.0434], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:39:47,151 INFO [train.py:904] (2/8) Epoch 30, batch 7650, loss[loss=0.1945, simple_loss=0.2751, pruned_loss=0.05693, over 16325.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.285, pruned_loss=0.0559, over 3079570.22 frames. ], batch size: 35, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:39:53,135 INFO [optim.py:368] (2/8) 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 23:40:59,090 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3949, 2.5257, 2.1708, 2.2667, 2.9056, 2.5276, 2.8721, 3.1026], device='cuda:2'), covar=tensor([0.0180, 0.0485, 0.0606, 0.0535, 0.0320, 0.0440, 0.0257, 0.0283], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0243, 0.0234, 0.0235, 0.0245, 0.0241, 0.0241, 0.0243], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:41:06,720 INFO [train.py:904] (2/8) Epoch 30, batch 7700, loss[loss=0.1845, simple_loss=0.2793, pruned_loss=0.04487, over 16751.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.285, pruned_loss=0.05626, over 3078963.23 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:41:57,516 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:18,119 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:26,381 INFO [train.py:904] (2/8) Epoch 30, batch 7750, loss[loss=0.2428, simple_loss=0.314, pruned_loss=0.08579, over 11588.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2848, pruned_loss=0.05572, over 3082187.45 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:42:33,832 INFO [optim.py:368] (2/8) 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 23:43:12,804 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:43:41,534 INFO [train.py:904] (2/8) Epoch 30, batch 7800, loss[loss=0.2026, simple_loss=0.2877, pruned_loss=0.05873, over 16462.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2864, pruned_loss=0.05736, over 3054112.73 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:44:55,078 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302200.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:45:00,017 INFO [train.py:904] (2/8) Epoch 30, batch 7850, loss[loss=0.2608, simple_loss=0.3303, pruned_loss=0.09566, over 11703.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.287, pruned_loss=0.0574, over 3045123.22 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:45:06,745 INFO [optim.py:368] (2/8) 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 23:46:15,231 INFO [train.py:904] (2/8) Epoch 30, batch 7900, loss[loss=0.2089, simple_loss=0.3029, pruned_loss=0.05747, over 16372.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2864, pruned_loss=0.0567, over 3061110.62 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:46:16,927 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302255.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:46:25,210 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:46:25,401 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:47:22,453 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302296.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:47:32,971 INFO [train.py:904] (2/8) Epoch 30, batch 7950, loss[loss=0.1682, simple_loss=0.261, pruned_loss=0.03774, over 17178.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.286, pruned_loss=0.05655, over 3071562.31 frames. ], batch size: 46, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:47:41,024 INFO [optim.py:368] (2/8) 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 23:48:26,670 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7825, 1.8852, 1.6590, 1.5287, 2.0044, 1.6692, 1.6379, 1.9821], device='cuda:2'), covar=tensor([0.0227, 0.0310, 0.0454, 0.0372, 0.0243, 0.0283, 0.0192, 0.0218], device='cuda:2'), in_proj_covar=tensor([0.0233, 0.0244, 0.0234, 0.0235, 0.0246, 0.0242, 0.0242, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:48:48,631 INFO [train.py:904] (2/8) Epoch 30, batch 8000, loss[loss=0.1872, simple_loss=0.2824, pruned_loss=0.04604, over 16491.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2867, pruned_loss=0.057, over 3074416.02 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:48:54,512 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:49:57,499 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302399.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:50:03,889 INFO [train.py:904] (2/8) Epoch 30, batch 8050, loss[loss=0.1922, simple_loss=0.2842, pruned_loss=0.05013, over 15418.00 frames. ], tot_loss[loss=0.199, simple_loss=0.286, pruned_loss=0.05599, over 3094493.38 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:50:09,897 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4062, 2.7952, 3.0613, 1.9761, 2.7807, 2.0250, 3.0310, 3.1384], device='cuda:2'), covar=tensor([0.0304, 0.0950, 0.0613, 0.2250, 0.0864, 0.1140, 0.0660, 0.0876], device='cuda:2'), in_proj_covar=tensor([0.0160, 0.0172, 0.0170, 0.0157, 0.0149, 0.0134, 0.0146, 0.0185], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-02 23:50:11,661 INFO [optim.py:368] (2/8) 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 23:51:10,082 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302447.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:51:21,030 INFO [train.py:904] (2/8) Epoch 30, batch 8100, loss[loss=0.1953, simple_loss=0.2784, pruned_loss=0.05605, over 16893.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2852, pruned_loss=0.0554, over 3079794.85 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:51:56,150 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:52:06,372 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8931, 2.4337, 2.0649, 2.2029, 2.7403, 2.4034, 2.5205, 2.8942], device='cuda:2'), covar=tensor([0.0258, 0.0469, 0.0579, 0.0507, 0.0309, 0.0404, 0.0266, 0.0295], device='cuda:2'), in_proj_covar=tensor([0.0232, 0.0244, 0.0234, 0.0235, 0.0246, 0.0242, 0.0242, 0.0244], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-02 23:52:35,993 INFO [train.py:904] (2/8) Epoch 30, batch 8150, loss[loss=0.2124, simple_loss=0.2882, pruned_loss=0.06831, over 11750.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2829, pruned_loss=0.05455, over 3088508.75 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:52:43,486 INFO [optim.py:368] (2/8) 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 23:53:28,207 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302538.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:53:29,757 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 23:53:52,239 INFO [train.py:904] (2/8) Epoch 30, batch 8200, loss[loss=0.2061, simple_loss=0.2794, pruned_loss=0.06638, over 11900.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2807, pruned_loss=0.05446, over 3076872.07 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:53:53,932 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:53:55,105 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:54:03,195 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:05,089 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 23:55:12,501 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:13,852 INFO [train.py:904] (2/8) Epoch 30, batch 8250, loss[loss=0.1631, simple_loss=0.2526, pruned_loss=0.03679, over 11899.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2795, pruned_loss=0.0522, over 3046512.62 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:55:22,071 INFO [optim.py:368] (2/8) 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] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:02,584 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:35,914 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:38,704 INFO [train.py:904] (2/8) Epoch 30, batch 8300, loss[loss=0.1787, simple_loss=0.2578, pruned_loss=0.04976, over 11902.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2766, pruned_loss=0.04896, over 3038648.73 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:57:36,004 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 23:57:43,531 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302693.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:58:01,686 INFO [train.py:904] (2/8) Epoch 30, batch 8350, loss[loss=0.1875, simple_loss=0.2827, pruned_loss=0.04612, over 16288.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2756, pruned_loss=0.04714, over 3018145.33 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:58:09,437 INFO [optim.py:368] (2/8) 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 23:58:20,149 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.3545, 3.4618, 3.6273, 3.6050, 3.6269, 3.4789, 3.4896, 3.5246], device='cuda:2'), covar=tensor([0.0538, 0.0915, 0.0672, 0.0604, 0.0649, 0.0738, 0.1014, 0.0669], device='cuda:2'), in_proj_covar=tensor([0.0439, 0.0500, 0.0478, 0.0443, 0.0522, 0.0507, 0.0583, 0.0408], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-02 23:59:22,769 INFO [train.py:904] (2/8) Epoch 30, batch 8400, loss[loss=0.1755, simple_loss=0.2652, pruned_loss=0.04291, over 15318.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2732, pruned_loss=0.04498, over 3022472.33 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:59:45,804 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8888, 2.6941, 2.9535, 2.1639, 2.7432, 2.1827, 2.7330, 2.8855], device='cuda:2'), covar=tensor([0.0313, 0.1115, 0.0488, 0.2132, 0.0859, 0.1066, 0.0633, 0.0880], device='cuda:2'), in_proj_covar=tensor([0.0159, 0.0170, 0.0169, 0.0156, 0.0148, 0.0132, 0.0145, 0.0183], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:2') 2023-05-03 00:00:37,352 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.5923, 4.7173, 4.9297, 4.6898, 4.7920, 5.2972, 4.7744, 4.4814], device='cuda:2'), covar=tensor([0.1218, 0.1877, 0.1921, 0.1841, 0.2067, 0.0817, 0.1532, 0.2329], device='cuda:2'), in_proj_covar=tensor([0.0426, 0.0634, 0.0707, 0.0517, 0.0685, 0.0725, 0.0549, 0.0690], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:2') 2023-05-03 00:00:44,411 INFO [train.py:904] (2/8) Epoch 30, batch 8450, loss[loss=0.1632, simple_loss=0.2614, pruned_loss=0.03245, over 16671.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2718, pruned_loss=0.04326, over 3044722.65 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:52,080 INFO [optim.py:368] (2/8) 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 00:01:31,127 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302833.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:02:04,571 INFO [train.py:904] (2/8) Epoch 30, batch 8500, loss[loss=0.1646, simple_loss=0.2435, pruned_loss=0.04288, over 11995.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2683, pruned_loss=0.04109, over 3047706.36 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:02:08,473 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302856.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:02:19,443 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5654, 3.6121, 3.3929, 3.0809, 3.2340, 3.4943, 3.3328, 3.3547], device='cuda:2'), covar=tensor([0.0562, 0.0706, 0.0304, 0.0286, 0.0419, 0.0549, 0.1324, 0.0517], device='cuda:2'), in_proj_covar=tensor([0.0312, 0.0474, 0.0364, 0.0365, 0.0358, 0.0421, 0.0252, 0.0435], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:02:48,192 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302881.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:03:26,596 INFO [train.py:904] (2/8) Epoch 30, batch 8550, loss[loss=0.159, simple_loss=0.2438, pruned_loss=0.0371, over 11822.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2659, pruned_loss=0.04027, over 3013480.52 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:03:27,650 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302904.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:03:38,834 INFO [optim.py:368] (2/8) 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 00:04:15,838 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-03 00:04:45,023 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302942.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:05:04,946 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302952.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:05:08,027 INFO [train.py:904] (2/8) Epoch 30, batch 8600, loss[loss=0.1678, simple_loss=0.2694, pruned_loss=0.03309, over 15354.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2662, pruned_loss=0.03941, over 3025972.48 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:05:23,980 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-03 00:06:16,016 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302988.0, num_to_drop=1, layers_to_drop={3} 2023-05-03 00:06:39,315 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303000.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:06:44,212 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.4325, 3.7247, 3.7429, 2.5465, 3.3103, 3.7826, 3.5038, 2.2159], device='cuda:2'), covar=tensor([0.0504, 0.0072, 0.0062, 0.0404, 0.0146, 0.0095, 0.0104, 0.0478], device='cuda:2'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0135, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:2') 2023-05-03 00:06:46,604 INFO [train.py:904] (2/8) Epoch 30, batch 8650, loss[loss=0.1758, simple_loss=0.2591, pruned_loss=0.04625, over 11852.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2651, pruned_loss=0.03821, over 3036361.35 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:06:58,835 INFO [optim.py:368] (2/8) 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 00:08:05,617 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-05-03 00:08:31,191 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303052.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:08:34,579 INFO [train.py:904] (2/8) Epoch 30, batch 8700, loss[loss=0.1531, simple_loss=0.2546, pruned_loss=0.02579, over 16868.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.263, pruned_loss=0.03725, over 3049484.16 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:09:57,840 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.2260, 1.6711, 1.9680, 2.2561, 2.3228, 2.5631, 1.8271, 2.5182], device='cuda:2'), covar=tensor([0.0319, 0.0605, 0.0437, 0.0420, 0.0451, 0.0251, 0.0663, 0.0204], device='cuda:2'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0193, 0.0211, 0.0168, 0.0205, 0.0169], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:2') 2023-05-03 00:10:14,424 INFO [train.py:904] (2/8) Epoch 30, batch 8750, loss[loss=0.1707, simple_loss=0.2719, pruned_loss=0.03474, over 16568.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2622, pruned_loss=0.03649, over 3045157.04 frames. ], batch size: 62, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:25,189 INFO [optim.py:368] (2/8) 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 00:10:28,949 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.6006, 2.7583, 2.3406, 4.2059, 2.4935, 3.9383, 1.4724, 3.0251], device='cuda:2'), covar=tensor([0.1503, 0.0775, 0.1381, 0.0190, 0.0139, 0.0393, 0.1845, 0.0723], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0204, 0.0204, 0.0216, 0.0210, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-03 00:10:38,719 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303113.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:11:23,564 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303133.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:12:06,340 INFO [train.py:904] (2/8) Epoch 30, batch 8800, loss[loss=0.168, simple_loss=0.2687, pruned_loss=0.03367, over 16849.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2614, pruned_loss=0.03571, over 3068147.27 frames. ], batch size: 96, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:13:02,844 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303181.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:13:26,388 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-03 00:13:50,924 INFO [train.py:904] (2/8) Epoch 30, batch 8850, loss[loss=0.1469, simple_loss=0.2426, pruned_loss=0.02559, over 12698.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2646, pruned_loss=0.0354, over 3063387.62 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:14:00,695 INFO [optim.py:368] (2/8) 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 00:14:49,609 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-03 00:15:03,333 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303237.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:15:38,681 INFO [train.py:904] (2/8) Epoch 30, batch 8900, loss[loss=0.1741, simple_loss=0.2716, pruned_loss=0.03834, over 16966.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.264, pruned_loss=0.03454, over 3040270.29 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:15:49,465 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303258.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:16:15,045 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.9623, 2.2646, 2.3263, 3.1396, 1.8173, 3.2873, 1.7718, 2.7775], device='cuda:2'), covar=tensor([0.1201, 0.0723, 0.1098, 0.0191, 0.0077, 0.0329, 0.1623, 0.0709], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0203, 0.0203, 0.0215, 0.0209, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-03 00:17:03,859 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303288.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:17:43,148 INFO [train.py:904] (2/8) Epoch 30, batch 8950, loss[loss=0.1534, simple_loss=0.2459, pruned_loss=0.03045, over 16959.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2634, pruned_loss=0.03489, over 3046865.26 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:17:53,309 INFO [optim.py:368] (2/8) 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 00:18:16,175 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303319.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:18:23,901 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4460, 2.7604, 3.1728, 2.0119, 2.8285, 2.1462, 3.0348, 3.0573], device='cuda:2'), covar=tensor([0.0283, 0.0955, 0.0544, 0.2241, 0.0810, 0.1027, 0.0729, 0.0919], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0168, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-03 00:18:53,964 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303336.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:19:32,192 INFO [train.py:904] (2/8) Epoch 30, batch 9000, loss[loss=0.1452, simple_loss=0.2408, pruned_loss=0.0248, over 16704.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2602, pruned_loss=0.0337, over 3043727.66 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:19:32,193 INFO [train.py:929] (2/8) Computing validation loss 2023-05-03 00:19:42,075 INFO [train.py:938] (2/8) Epoch 30, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01984, over 944034.00 frames. 2023-05-03 00:19:42,076 INFO [train.py:939] (2/8) Maximum memory allocated so far is 17820MB 2023-05-03 00:20:36,957 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.1614, 4.3040, 4.0731, 3.8117, 3.6456, 4.2010, 3.8988, 3.9338], device='cuda:2'), covar=tensor([0.0745, 0.0725, 0.0453, 0.0422, 0.1052, 0.0635, 0.0940, 0.0727], device='cuda:2'), in_proj_covar=tensor([0.0308, 0.0466, 0.0360, 0.0361, 0.0354, 0.0415, 0.0248, 0.0428], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:20:37,009 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.3770, 3.0594, 3.3476, 1.7858, 3.4313, 3.5127, 2.9122, 2.7973], device='cuda:2'), covar=tensor([0.0734, 0.0304, 0.0201, 0.1203, 0.0105, 0.0188, 0.0429, 0.0472], device='cuda:2'), in_proj_covar=tensor([0.0145, 0.0108, 0.0099, 0.0135, 0.0085, 0.0129, 0.0127, 0.0127], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:2') 2023-05-03 00:20:44,495 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.9202, 5.2817, 5.0805, 5.1058, 4.8314, 4.8333, 4.7028, 5.3715], device='cuda:2'), covar=tensor([0.1265, 0.0876, 0.0902, 0.0809, 0.0797, 0.0881, 0.1216, 0.0783], device='cuda:2'), in_proj_covar=tensor([0.0712, 0.0854, 0.0704, 0.0666, 0.0544, 0.0545, 0.0715, 0.0672], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:20:44,624 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.4513, 3.5697, 2.0390, 3.9423, 2.6228, 3.8631, 2.3457, 2.8621], device='cuda:2'), covar=tensor([0.0336, 0.0397, 0.1821, 0.0274, 0.0952, 0.0612, 0.1556, 0.0822], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0177, 0.0191, 0.0168, 0.0176, 0.0215, 0.0200, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-03 00:21:25,802 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.3613, 3.5353, 3.8482, 2.3040, 3.2774, 2.5708, 3.6727, 3.7373], device='cuda:2'), covar=tensor([0.0249, 0.0866, 0.0501, 0.2094, 0.0775, 0.0923, 0.0664, 0.1051], device='cuda:2'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0180], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:2') 2023-05-03 00:21:28,473 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303403.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:29,165 INFO [train.py:904] (2/8) Epoch 30, batch 9050, loss[loss=0.1598, simple_loss=0.2468, pruned_loss=0.03636, over 16492.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.261, pruned_loss=0.03408, over 3051594.62 frames. ], batch size: 62, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:21:38,011 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303408.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:39,555 INFO [optim.py:368] (2/8) 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 00:23:13,940 INFO [train.py:904] (2/8) Epoch 30, batch 9100, loss[loss=0.183, simple_loss=0.2772, pruned_loss=0.04438, over 16941.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2605, pruned_loss=0.03454, over 3054171.13 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:23:33,690 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303464.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:24:43,203 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.5213, 1.8311, 2.1984, 2.5775, 2.5101, 2.9852, 2.0129, 2.8739], device='cuda:2'), covar=tensor([0.0307, 0.0676, 0.0424, 0.0400, 0.0481, 0.0198, 0.0645, 0.0184], device='cuda:2'), in_proj_covar=tensor([0.0197, 0.0198, 0.0187, 0.0192, 0.0209, 0.0166, 0.0205, 0.0167], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:25:12,849 INFO [train.py:904] (2/8) Epoch 30, batch 9150, loss[loss=0.1521, simple_loss=0.2497, pruned_loss=0.02727, over 16986.00 frames. ], tot_loss[loss=0.165, simple_loss=0.261, pruned_loss=0.03455, over 3032429.07 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:25:25,336 INFO [optim.py:368] (2/8) 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 00:25:48,560 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5746, 3.5202, 3.5167, 2.5725, 3.4019, 1.9266, 3.2483, 2.8754], device='cuda:2'), covar=tensor([0.0306, 0.0244, 0.0270, 0.0339, 0.0197, 0.3138, 0.0210, 0.0407], device='cuda:2'), in_proj_covar=tensor([0.0183, 0.0173, 0.0212, 0.0182, 0.0189, 0.0217, 0.0200, 0.0177], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:25:54,996 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.8129, 1.9861, 2.3578, 3.1056, 2.1065, 2.1299, 2.2063, 2.0937], device='cuda:2'), covar=tensor([0.1831, 0.4353, 0.3078, 0.0972, 0.5599, 0.3632, 0.4069, 0.4760], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0472, 0.0386, 0.0332, 0.0441, 0.0541, 0.0446, 0.0551], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:26:23,102 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303537.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:26:54,942 INFO [train.py:904] (2/8) Epoch 30, batch 9200, loss[loss=0.1534, simple_loss=0.2439, pruned_loss=0.03144, over 17055.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2569, pruned_loss=0.03396, over 3020545.37 frames. ], batch size: 50, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:27:31,972 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303574.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:27:51,756 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303585.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:28:17,435 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.6760, 2.6722, 1.7182, 2.7959, 2.0594, 2.7903, 1.9715, 2.3269], device='cuda:2'), covar=tensor([0.0357, 0.0381, 0.1569, 0.0358, 0.0789, 0.0603, 0.1576, 0.0679], device='cuda:2'), in_proj_covar=tensor([0.0174, 0.0178, 0.0191, 0.0168, 0.0176, 0.0215, 0.0201, 0.0179], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-03 00:28:29,157 INFO [train.py:904] (2/8) Epoch 30, batch 9250, loss[loss=0.1433, simple_loss=0.2488, pruned_loss=0.01889, over 16816.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2574, pruned_loss=0.03412, over 3038500.53 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:28:41,899 INFO [optim.py:368] (2/8) 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 00:28:49,554 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303614.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:29:38,958 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303635.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:30:19,432 INFO [train.py:904] (2/8) Epoch 30, batch 9300, loss[loss=0.1472, simple_loss=0.2334, pruned_loss=0.0305, over 12573.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.256, pruned_loss=0.03364, over 3029027.94 frames. ], batch size: 250, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:31:02,544 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([1.7272, 2.4670, 2.3890, 3.7926, 1.8253, 3.6726, 1.6482, 2.8234], device='cuda:2'), covar=tensor([0.1556, 0.0999, 0.1433, 0.0215, 0.0114, 0.0518, 0.1925, 0.0849], device='cuda:2'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0202, 0.0202, 0.0215, 0.0210, 0.0198], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:2') 2023-05-03 00:31:16,530 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-03 00:32:05,574 INFO [train.py:904] (2/8) Epoch 30, batch 9350, loss[loss=0.1735, simple_loss=0.2688, pruned_loss=0.03909, over 16483.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2554, pruned_loss=0.0333, over 3031271.27 frames. ], batch size: 147, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:32:13,791 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303708.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:32:16,696 INFO [optim.py:368] (2/8) 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 00:33:46,254 INFO [train.py:904] (2/8) Epoch 30, batch 9400, loss[loss=0.1608, simple_loss=0.2682, pruned_loss=0.02666, over 15447.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2548, pruned_loss=0.03344, over 3000358.59 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:33:50,139 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303756.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:33:56,164 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303759.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:35:26,329 INFO [train.py:904] (2/8) Epoch 30, batch 9450, loss[loss=0.1506, simple_loss=0.2419, pruned_loss=0.02963, over 12424.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2569, pruned_loss=0.03383, over 2988341.61 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:35:36,995 INFO [optim.py:368] (2/8) 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 00:35:38,563 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303810.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:36:33,620 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([4.8211, 4.8475, 5.1896, 5.1654, 5.1775, 4.9441, 4.8563, 4.7719], device='cuda:2'), covar=tensor([0.0325, 0.0634, 0.0372, 0.0402, 0.0469, 0.0377, 0.0876, 0.0431], device='cuda:2'), in_proj_covar=tensor([0.0427, 0.0485, 0.0466, 0.0432, 0.0511, 0.0493, 0.0563, 0.0398], device='cuda:2'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:2') 2023-05-03 00:37:08,363 INFO [train.py:904] (2/8) Epoch 30, batch 9500, loss[loss=0.1632, simple_loss=0.2524, pruned_loss=0.037, over 12520.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.256, pruned_loss=0.03341, over 3010157.99 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:37:43,313 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303871.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:36,347 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303896.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:44,685 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-03 00:38:54,494 INFO [train.py:904] (2/8) Epoch 30, batch 9550, loss[loss=0.1704, simple_loss=0.2706, pruned_loss=0.03507, over 15391.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2559, pruned_loss=0.03346, over 3025497.14 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:39:08,517 INFO [optim.py:368] (2/8) 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 00:39:18,199 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303914.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:39:49,633 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303930.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:40:36,089 INFO [train.py:904] (2/8) Epoch 30, batch 9600, loss[loss=0.1699, simple_loss=0.2626, pruned_loss=0.03857, over 16686.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2567, pruned_loss=0.03354, over 3034435.74 frames. ], batch size: 76, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:40:43,440 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303957.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:40:52,683 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303962.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:41:16,221 INFO [zipformer.py:625] (2/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303974.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:42:24,335 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.0769, 1.8007, 1.6396, 1.4694, 1.9682, 1.5324, 1.4921, 1.9616], device='cuda:2'), covar=tensor([0.0249, 0.0420, 0.0620, 0.0519, 0.0310, 0.0418, 0.0203, 0.0276], device='cuda:2'), in_proj_covar=tensor([0.0225, 0.0238, 0.0229, 0.0230, 0.0240, 0.0237, 0.0233, 0.0236], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:42:31,193 INFO [train.py:904] (2/8) Epoch 30, batch 9650, loss[loss=0.1529, simple_loss=0.2479, pruned_loss=0.02898, over 16611.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2584, pruned_loss=0.03374, over 3019644.63 frames. ], batch size: 57, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:42:48,098 INFO [optim.py:368] (2/8) 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 00:43:29,959 INFO [scaling.py:679] (2/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-03 00:43:40,591 INFO [zipformer.py:625] (2/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304035.0, num_to_drop=1, layers_to_drop={2} 2023-05-03 00:44:20,836 INFO [train.py:904] (2/8) Epoch 30, batch 9700, loss[loss=0.1681, simple_loss=0.2545, pruned_loss=0.04086, over 12135.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2575, pruned_loss=0.03338, over 3029294.26 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:44:30,050 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304059.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:45:15,257 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-03 00:45:26,565 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-03 00:45:48,818 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.1462, 3.5371, 3.6428, 2.3709, 3.2880, 3.6557, 3.4434, 2.0730], device='cuda:2'), covar=tensor([0.0679, 0.0087, 0.0066, 0.0497, 0.0140, 0.0106, 0.0094, 0.0548], device='cuda:2'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0134, 0.0103, 0.0114, 0.0097, 0.0129], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-03 00:46:04,106 INFO [train.py:904] (2/8) Epoch 30, batch 9750, loss[loss=0.1731, simple_loss=0.2577, pruned_loss=0.04421, over 12446.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2567, pruned_loss=0.03377, over 3034766.70 frames. ], batch size: 250, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:46:09,338 INFO [zipformer.py:625] (2/8) 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] (2/8) 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 00:47:41,831 INFO [train.py:904] (2/8) Epoch 30, batch 9800, loss[loss=0.1558, simple_loss=0.2666, pruned_loss=0.02247, over 16744.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2574, pruned_loss=0.03288, over 3052850.93 frames. ], batch size: 83, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:48:04,642 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304166.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:49:25,007 INFO [train.py:904] (2/8) Epoch 30, batch 9850, loss[loss=0.1701, simple_loss=0.2642, pruned_loss=0.03798, over 15324.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2584, pruned_loss=0.03258, over 3058248.10 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:49:39,349 INFO [optim.py:368] (2/8) 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 00:50:19,497 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304230.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:50:27,199 INFO [scaling.py:679] (2/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-03 00:51:13,312 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304252.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:51:16,011 INFO [train.py:904] (2/8) Epoch 30, batch 9900, loss[loss=0.1768, simple_loss=0.2612, pruned_loss=0.04623, over 12184.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2586, pruned_loss=0.03282, over 3039079.07 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:52:03,595 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9071, 2.0715, 2.2270, 3.3675, 2.0637, 2.3004, 2.2190, 2.1989], device='cuda:2'), covar=tensor([0.1586, 0.3989, 0.3391, 0.0802, 0.4888, 0.3003, 0.3871, 0.4147], device='cuda:2'), in_proj_covar=tensor([0.0419, 0.0471, 0.0385, 0.0332, 0.0441, 0.0539, 0.0446, 0.0550], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:52:13,219 INFO [zipformer.py:625] (2/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=304278.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:53:14,777 INFO [train.py:904] (2/8) Epoch 30, batch 9950, loss[loss=0.1996, simple_loss=0.2896, pruned_loss=0.05481, over 12641.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2617, pruned_loss=0.03344, over 3047027.90 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:53:31,713 INFO [optim.py:368] (2/8) 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 00:54:20,733 INFO [zipformer.py:625] (2/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304330.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:55:15,632 INFO [train.py:904] (2/8) Epoch 30, batch 10000, loss[loss=0.1705, simple_loss=0.2545, pruned_loss=0.04329, over 12818.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2603, pruned_loss=0.03314, over 3067120.85 frames. ], batch size: 250, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:56:57,636 INFO [train.py:904] (2/8) Epoch 30, batch 10050, loss[loss=0.1809, simple_loss=0.2795, pruned_loss=0.04111, over 15274.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2599, pruned_loss=0.03279, over 3079663.74 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:57:10,557 INFO [optim.py:368] (2/8) 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 00:57:42,890 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.0714, 4.4143, 4.4522, 3.2946, 3.7244, 4.4464, 3.9139, 2.7235], device='cuda:2'), covar=tensor([0.0416, 0.0061, 0.0042, 0.0321, 0.0140, 0.0086, 0.0091, 0.0421], device='cuda:2'), in_proj_covar=tensor([0.0136, 0.0089, 0.0090, 0.0133, 0.0102, 0.0112, 0.0096, 0.0128], device='cuda:2'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:2') 2023-05-03 00:57:52,062 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([2.9447, 2.1568, 2.2576, 3.3902, 2.1417, 2.4191, 2.3052, 2.2791], device='cuda:2'), covar=tensor([0.1544, 0.4053, 0.3398, 0.0739, 0.4518, 0.2600, 0.3713, 0.3862], device='cuda:2'), in_proj_covar=tensor([0.0417, 0.0469, 0.0384, 0.0330, 0.0439, 0.0536, 0.0443, 0.0547], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:58:34,282 INFO [train.py:904] (2/8) Epoch 30, batch 10100, loss[loss=0.1352, simple_loss=0.2327, pruned_loss=0.01887, over 16862.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2601, pruned_loss=0.03285, over 3085031.53 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:58:40,331 INFO [zipformer.py:1454] (2/8) attn_weights_entropy = tensor([3.5818, 3.4708, 3.5279, 2.8241, 3.2976, 2.0500, 3.0909, 2.8699], device='cuda:2'), covar=tensor([0.0240, 0.0311, 0.0255, 0.0393, 0.0196, 0.2643, 0.0235, 0.0391], device='cuda:2'), in_proj_covar=tensor([0.0181, 0.0172, 0.0209, 0.0179, 0.0187, 0.0216, 0.0198, 0.0175], device='cuda:2'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:2') 2023-05-03 00:58:57,145 INFO [zipformer.py:625] (2/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304466.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:59:55,830 INFO [train.py:1169] (2/8) Done!